Audio can change the whole feel of a short video.

The same clip can feel calm, messy, sad, big, or fast based on the sound under it. So I keep a small list of audios I can grab when I do not want to scroll for an hour.

If you came from Instagram and commented SOUNDS26, this is the free list.

Start here

Use these inside Instagram when you want the sound to match the mood.

1. Now or Never - motivational

Use this audio on Instagram

1. Good Man With a Broken Heart - cinematic

Use this audio on Instagram

1. Runaway - dramatic

Use this audio on Instagram

1. Timeless - powerful

Use this audio on Instagram

1. Snowfall - nostalgic

Use this audio on Instagram

1. Lost in the Fire - mysterious

Use this audio on Instagram

More background audios

This public list has more viral background audio links you can test:

Open the Viral Background Audios list

Quick tip

Do not build the whole Reel first.

Pick the mood. Add the sound. Then cut the b-roll to the sound.

And before you post, check the audio from your own account. Instagram can change what is shown by account, country, and account type.

I will keep adding clean links as I find them.

Do that six times and your AI bill beats your phone bill while your work looks exactly the same. Meanwhile the apps you already pay for shipped AI features you never switched on, so you are paying twice for things you own.

What you will be able to do

• Name the six jobs a working AI stack has to cover, and which tool you use for each.

• Apply one rule that tells you whether a subscription stays or goes.

• Turn on AI features inside apps you already pay for, without buying anything.

• Set up your main AI so it knows who you are and can reach your email, calendar, and files.

• Ignore most AI news on purpose and still not fall behind.

This is for anyone whose AI spending grew faster than their AI results. You need about five tools, and two of them are probably free.

The proof behind this

Liz Elliott runs 33 active social brands and her whole operation runs on AI automations across several computers. That is not 33 tools. It is a small stack, used hard, wired together.

What this lesson is

Lesson 2.5 of the free AI for Entrepreneurs course, in Module 2: Choosing Your AI.

There are thousands of AI tools. You need about five, and two of them are probably free.

The trap is collecting. You see a tool, it looks clever, you sign up, you use it twice, and now you pay eleven dollars a month forever for a thing you forgot about. Do that six times and your AI bill is bigger than your phone bill and your work has not changed.

The fix is to stop shopping for tools and start naming jobs. A working setup covers a small number of distinct jobs, and each tool owns exactly one of them. Getting information in. Thinking and building. Looking things up with sources. Capturing meetings. Making visuals. Keeping the output somewhere you can find it.

Once you think in jobs, the shopping question changes. Instead of "is this tool good," you ask "which job does this do, and what am I currently using for that job." Two tools fighting over one job is waste. A job with no tool on it is a hole.

Here is the rule that keeps the list short. A subscription earns its place if you can name the one specific thing it replaced. Not "it helps." What did it replace, and what would you have to go back to doing by hand if you cancelled it tomorrow? If you cannot answer in one sentence, it goes.

And before you buy anything, check what you already have. Most of the apps you already pay for quietly shipped AI features in the last couple of years. You may be paying twice for something you already own.

From Liz's playbook: Liz Elliott runs 33 active social brands and her whole operation runs on a set of AI automations across several computers. That is not thirty-three tools. It is a small stack, used hard, wired together.

New here? Use the free AI for Entrepreneurs course outline to read this series in the best order.

Do it

Part 1: Apply the keep-or-cut rule

Open your bank statement or your app store subscriptions and list every AI tool you pay for. For each one, finish this sentence out loud: "If I cancelled this tomorrow, I would have to go back to ___."

If you cannot finish the sentence, cancel it today. If you can, write the answer down. That sentence is why the tool exists, and it is the thing to check again in six months.

Do this before you add anything from the rest of this lesson.

Part 2: Set up the hub

One tool is the middle of the stack. Everything else either feeds it information or catches what comes out. For most people that is the main chat AI you already use.

Set it up properly once. It takes a couple of minutes and it changes every answer after it.

1. Create an account at claude.ai, or download the desktop or mobile app.

2. Go to Settings, then Memory, and turn it on.

3. Tell it your name, your job, and how you like things written.

4. Go to Settings, then Connectors.

5. Link your Google account so it can read email, check your calendar, and search your Drive.

Menu wording moves between versions. If those exact words are not there, look for anything called memory, personalization, connectors, or apps.

Three more things worth doing in the hub:

• Build a Project for your work. Load in your role, your goals, your clients, and the documents you keep re-explaining. Every chat inside that Project starts already knowing all of it.

• Set your standards once. Something like: Push back when I am wrong. No jargon. Keep it under 150 words. It holds those across everything after.

• Stop using it like a search engine. It works best when you hand it real context and ask for real work, not when you type five words and hope.

Connectors are the line between an AI that talks about your work and one that reaches in and does something. Before you link anything, know what you are giving access to. Email and Drive are your whole life. Grant one at a time and see how it goes.

Part 3: Cover the six jobs

Here is the stack by job. Pick one tool per row. Two tools in one row is waste.

The job / What it does for you / Common picks

• Getting information in

• What it does for you: Turn talking into clean written text anywhere you can type

• Common picks: Wispr Flow, or the voice mode built into your main AI

• Thinking and building

• What it does for you: Writing, analysis, planning, documents, tools, systems

• Common picks: Your main AI: Claude or ChatGPT

• Looking things up

• What it does for you: Web answers with real sources you can click

• Common picks: Perplexity

• Capturing meetings

• What it does for you: Notes, decisions, action items, who owns what

• Common picks: Granola, Fathom, or the AI already inside Zoom

• Making things you can see

• What it does for you: Decks, images, mockups, graphics

• Common picks: Gamma for decks, your AI's image tool for images

• Keeping the output

• What it does for you: One searchable place where the work lands

• Common picks: Notion, or any notes app you already use

A seventh job shows up if you study or research a lot: turning documents, papers, and videos into something you can actually absorb. NotebookLM does that. Upload a chapter, a report, or a video, and get flashcards, a quiz, a study guide, or a spoken overview with two AI hosts talking through it.

A few notes on the rows.

Getting information in is the cheapest big win and the one people skip. Most people talk far faster than they type, and when you type you unconsciously shorten what you were about to say. Less context in, worse answer out. Voice mode is built into the main AI apps and costs nothing. A dedicated tool like Wispr Flow goes further, because it works in every app, not just the AI one, and cleans up the filler words as it goes. Set an activation shortcut, hold it, talk, let go, and clean text appears where your cursor is. Free tiers here are usually word-capped, so if it clicks you will hit the cap quickly.

Looking things up is a genuinely different job from thinking. Your main AI is not a search engine. A search AI reads live pages and hands you an answer with links you can check. Use it for product comparisons, current facts, market and competitor questions, and anything where you need to see the source.

Capturing meetings comes with a rule that no tool will remind you of. Tell people you are recording, and get their agreement, before the tool starts. Recording laws differ by country and by state, and workplace policies differ too. This is on you, not on the app. Also read the notes before you share them. Meeting AI regularly misattributes who said what and invents action items nobody agreed to.

Keeping the output matters more than it sounds. Chat histories are where good work goes to die. You had a great planning session three weeks ago and now you cannot find it. Make one page per project and dump every useful output into it. Boring, and it compounds fast. Then connect that place back to your AI so it can read your own material instead of the internet's.

Part 4: Turn on the AI you already pay for

Before you buy a single new thing, check these five. Some are free, some are included in plans you already have.

App / The feature / What it does / Typical access

• Zoom

• The feature: AI Companion meeting notes

• What it does: Summary, action items, who owns each one, emailed to attendees

• Typical access: Included on paid Zoom plans

• Slack

• The feature: Channel recaps and thread summaries

• What it does: A morning digest of your noisiest channels, and a Summarize button on long threads

• Typical access: Included on paid Slack plans

• Notion

• The feature: Workspace search and autofill

• What it does: Ask a question and it reads across your own pages, notes, and databases to answer

• Typical access: Paid tiers only, and the cheap add-on has been discontinued before

• Google Docs

• The feature: Help Me Write

• What it does: Type what you need and it drafts it, using context from your other Google apps

• Typical access: Free with a Google account, more on Workspace plans

• Google Sheets

• The feature: Formulas and analysis in plain English

• What it does: Describe the formula in words and it writes it. Also makes charts and finds trends

• Typical access: Free with a Google account, more on Workspace plans

How to reach them:

• Zoom: click the sparkle icon during your next meeting. That turns on AI Companion. It listens to the call and writes the summary at the end.

• Slack: click Summarize on a long thread. For the daily digest, look in Slack's AI settings for channel recaps.

• Notion, Docs, Sheets: these move around. In Notion, press Space on an empty line, or highlight text and click Ask AI. In Docs and Sheets, look for the AI or Help Me Write option in the toolbar or side panel.

Do not turn on all five at once. Pick the app you spend the most time in, turn on its one feature, and use it once today. Come back for the rest later.

The same consent rule applies to Slack huddle transcription and any workspace-wide search. Know what your team has agreed to before you switch on something that reads across everyone's messages.

Part 5: Spend money like an adult

Four rules, in order.

1. Start on the free tier every time. Most of these have one. Use it until it stops being enough.

2. When you pay, start at the cheapest tier that does the job. The standard paid tier of a main AI has been around twenty dollars a month for a long time. Heavier tiers run several times that. Prices move, so check the vendor's own page before you buy. Nothing in this lesson is a current price.

3. Upgrade as a reaction, not as a hope. The right reason to move up a tier is that you keep hitting limits in the middle of real work. Wanting to be the kind of person who has the big plan is not a reason.

4. Run one AI for a month before you pay for a second. Notice where it keeps letting you down, and let that pick the second one. One paid plan plus one free account on the other side still gets you a second opinion for nothing.

Add the whole list up before you commit. Six tools at twenty dollars is a real bill, and the guides that recommend them almost never do the arithmetic for you.

Part 6: Ignore most of the news

New models and new features ship constantly. Reading every announcement feels like keeping up and is mostly a way to avoid doing the work. Almost all of it is a version number.

Here is the filter. A change is worth your attention only if it changes what you can hand over. Everything else is a footnote.

By that test, a small number of real shifts have happened, and each one you would feel the same day you switched:

1. AI finishes whole jobs now. The old pattern was ask a question, get an answer, still do the work. Now you hand over the outcome and walk away, and a finished file comes back. That is the biggest change and the one most people are behind on.

2. AI images are good enough to use. The warped hands and garbled text era is over. Images now go into ads, presentations, product mockups, and videos. The skill is describing the finished shot like a creative director instead of typing three words.

3. Talking beats typing. Covered above. Cheapest upgrade on the list.

4. Desktop apps beat browser tabs. In a browser tab, the AI only knows what you paste in, so your day becomes copy and paste. The desktop apps work with the files already on your machine, and access is scoped to the folder you grant.

5. Describing the result beats scripting the steps. Older models needed the steps spelled out. Now that advice makes your results worse. Give the situation and the finished outcome, then stop.

That last one is worth a moment. Old way: "Read this transcript, pull the key points, group them into themes, summarize each, add an intro and a conclusion." New way: "Turn this transcript into a one page brief my team can read in three minutes and know what we decided and who owns what." Shorter, and better, because it says what the thing is for.

Five habits for images, since that job trips people up most:

1. Say what you want, not what you do not want. "An empty street" works. "A street with no cars" often leaves the cars in.

2. Name materials and light. "Navy blue tweed" beats "jacket." "Golden hour backlight" beats "nice lighting."

3. Put exact words in quotes when you need text in the image, then add "appears once, no other text." That kills doubled and garbled words.

4. Feed it your own photos and give each one a job. That holds a face or a product steady across a set.

5. Change one thing at a time, and add "natural skin texture, no glossy filter, real photo feel" to kill the plastic look.

Copy and paste

Email triage

Run this in your main AI once your email is connected.

Summarize the emails I got today.
Group them by urgency:
- What needs a response now
- What can wait
- What I can ignore
Draft a reply for anything urgent.

After a meeting

Run this against your meeting notes, in whichever tool captured them.

What were the action items from this meeting and who owns each one?

Your standards line

Paste this into your main AI's settings or Project instructions once.

Push back when I am wrong. No jargon. Keep it under 150 words.

Build your context file by talking

The fastest way to make the file that makes every answer better. Two minutes of talking beats an hour of writing.

[Turn on voice input, hit record, and ramble for two minutes about
who you are, how you work, your clients, your goals, and what you
want this AI to do for you. Then paste that here and add:]

Turn this into a clean markdown file I can reuse as my context
document. Keep every fact I said. Organize it into sections:
who I am, what I sell, who my customers are, how I like to work,
my standards, and what I want you to do for me.

The second opinion pass

Paste finished work from one AI into a different one. This is lesson 2.1's review step, in one block.

Here is a finished piece of work I built with another AI.
[paste the work]

Tell me what is WRONG with it. Not what is good.
List the weaknesses, the wrong assumptions, the missing pieces,
and anything that would embarrass me if a customer saw it.
Be specific and rank them worst first.

Review mining

For anyone who sells something. Feed it real customer reviews.

Here are the reviews for [product or competitor].
[paste the reviews]

1. Group every complaint into themes and rank them by how often
   they come up.
2. Pull the exact words and phrases customers repeat about the
   product. Those become my ad copy.
3. Turn the top five complaints into product improvement ideas.

The outcome brief

The general shape for handing over a whole job instead of asking a question.

Here is my situation: [who you are, what you sell, who this is for].
Here is the raw material: [paste transcript, notes, or data].

Turn this into [the finished thing] that [the exact reader] can
[the exact action] in [time limit].
Done looks like: [what a great version contains].

Old way versus new way

Read these two side by side once. The second one gets better results.

Old way: "Read this transcript, pull the key points, group them into themes, summarize each, add an intro and a conclusion."

New way: "Turn this transcript into a one page brief my team can read in three minutes and know what we decided and who owns what."

The image brief

Fill in every line. This is the difference between a usable image and generic soup.

Create an image of [subject] in [setting].
Framing: [close up / wide shot / overhead / eye level].
Light: [golden hour backlight / soft window light / hard studio light].
Materials and detail: [navy blue tweed jacket, brushed brass, wet pavement].
Mood: [calm, warm, editorial].
Text in the image: "[EXACT WORDS]" - appears once, no other text.
Natural skin texture, no glossy filter, real photo feel.

Blending your own reference photos

For keeping a product or a face identical across a set of images.

Use photo one for [the product]. Use photo two for [the background].
Keep [the product] identical to photo one. Do not change its shape,
color, or label.
Match the lighting to [golden hour backlight].
Change only [one specific thing].

Image lines worth keeping

Short strings to paste into any image prompt.

appears once, no other text
natural skin texture, no glossy filter, real photo feel
Use photo one for the product, photo two for the background, keep the product identical.

One-job requests

The right size of ask. One clear job per request.

Make a product mockup for this candle.

Build a morning brief from my calendar and project list.

Ask your workspace a question

For Notion AI, or any tool that can search across your own notes.

What did the team decide about the launch date?

Spreadsheet formulas in plain English

Type this into the AI panel in Google Sheets instead of looking up the syntax.

Calculate the average of column B only for rows where column A says Q1.

what are the top trends in this data?

A personal image from your own photo library

Some photo tools can use people, pets, and places you have labeled. This is opt-in, so check the setting first.

Create a claymation version of me hiking

Money questions

Some money apps connect to your bank and answer questions like these. Only connect a bank if you are comfortable with that app holding that access, and read what it does with the data first.

How much did I spend on food this month?

Where am I wasting money?

Watch out

You buy the top tier first

The heavy plans solve a problem you may not have yet. Start at the cheapest paid tier. Move up only when limits interrupt real work in the middle of a task.

You pay for both big AIs on day one

Run one for a month. Notice where it keeps letting you down. Let that decide the second one, and keep the second one free if you can.

You collect tools instead of covering jobs

Two tools fighting over one job is money on fire. One tool per row in the table. If you cannot name what a tool replaced, cancel it.

You record people without telling them

Meeting AI records and transcribes a live call with other people on it. Recording laws vary by country and by state, and your employer may have its own policy. Say it out loud at the top of the call and get agreement. No app will do this for you.

You connect your bank to a chatty app

Money apps that read your accounts are useful and they are also a bank connection. Read what the app collects, what it keeps, and who it shares with. If you are unsure, skip it. Nothing in this lesson is financial advice.

You point a scraper at whatever you like

Tools that pull content off the web are powerful and they are not automatically allowed. Paid newsletters, subscription content, and other people's work come with terms of service and copyright. Check before you collect, and never republish what you pull.

You leave the work in chat history

Chats vanish into scroll and search badly. Move every output that matters into a real page or file. One page per project.

You chase every launch

Most announcements are a version number. A change matters only if it changes what you can hand over. Everything else can wait until you trip over it.

You skip the free layer

Zoom, Slack, Notion, Docs, and Sheets already have AI in them, and Docs and Sheets are free with a Google account. Check what you already own before you add a bill.

Check yourself

• [ ] I listed every AI tool I pay for and finished the "I would go back to ___" sentence for each.

• [ ] I cancelled anything I could not finish that sentence for.

• [ ] I turned on memory and connected at least one app in my main AI.

• [ ] I made a context file and saved it somewhere I can reuse it.

• [ ] I can name a tool for each of the six jobs, with no job double-covered.

• [ ] I turned on one AI feature inside an app I already pay for, and used it once.

• [ ] I told people before recording a meeting.

• [ ] I know what my whole AI stack costs per month, added up.

Do this now

Open your subscriptions list, pick the AI tool you are least sure about, and finish this sentence out loud: "If I cancelled this tomorrow, I would have to go back to ___." If you cannot finish it, cancel it before you close the tab.

Keep going

Next in AI for Entrepreneurs:

• Part 11: Day one: the five things to do first

• Part 12: Day two and three: the next ten things

• Part 13: The 60 minute setup

Every part of this course sits in reading order in the free AI for Entrepreneurs course outline.

This part is free, start to finish. Subscribing keeps the rest of the course coming and opens the paid parts.

Tell me in the comments the subscription you are about to cancel, and I will point you at the right next part.

While that argument runs, the real choice sits untouched. Both companies build the same three shapes. Picking by logo means you use the wrong shape for the job, then blame the brand for the result.

What you will be able to do

• Name the three shapes both companies build, and match them across brands.

• Read any model name and tell the generation apart from the size tier.

• Pick between the two tools using two questions, in about ten seconds.

• Hand over a whole job instead of asking a question, and get a finished file back.

• Spot the five quiet gates that make a request fail without an error message.

This is for anyone paying for one and wondering about the other, or paying for both and not sure why.

The proof behind this

Liz Elliott teaches beginners and non-technical people, and she runs 33 active social brands. She has no loyalty to a logo, and neither does this lesson.

What this lesson is

Lesson 2.4 of the free AI for Entrepreneurs course, in Module 2: Choosing Your AI.

The question people ask is "which one is better." That question has no answer, and asking it keeps you stuck.

Here is what is actually true. Both companies build the same three shapes, and they always have. A chat surface for thinking. A work surface that hands you a finished file or takes a real action in an app you connected. A build surface that makes something that runs.

Anthropic calls those Claude, Cowork, and Claude Code. OpenAI calls them Chat, Work, and Codex. Different names, same three shapes. When one company ships something new, the other usually ships its version within a few months. So whoever is ahead today is a snapshot, not a fact about the world.

Think of it like two supermarkets on the same street. Both sell bread, milk, and eggs. One has better bread this month. Next month the other one hires a new baker. If you have a loyalty card for one and refuse to walk into the other, you eat worse bread for no reason.

What does not flip every few months is the shape of your own work. Some of your work lives in cloud apps: research, docs, decks, reports. Some of it lives on your actual computer: downloads, spreadsheets, PDFs, the mess on your desktop. That difference is stable, and it is the best thing to choose on.

The second stable thing is how clear you are. When you already know exactly what you want, you want a tool that starts moving. When the job is fuzzy or the stakes are high, you want one that asks questions and pauses before it commits.

Those two questions pick your tool. Not the logo.

New here? Use the free AI for Entrepreneurs course outline to read this series in the best order.

Do it

Part 1: Learn the three shapes

Memorize this table and you can drop any new product into it.

The job / OpenAI calls it / Anthropic calls it / What comes out

• Thinking with you

• OpenAI calls it: Chat

• Anthropic calls it: Claude Chat

• What comes out: Answers, arguments, drafts, a second opinion. Nothing is saved or sent.

• Office work, finished files

• OpenAI calls it: Work

• Anthropic calls it: Cowork

• What comes out: A finished file, or a real action inside an app you connected.

• Building things that run

• OpenAI calls it: Codex

• Anthropic calls it: Claude Code

• What comes out: Something that runs. A site, a script, a change to your files.

One line sorts all of it: Chat gives you thinking. Work gives you a finished thing. The build surface gives you something that runs. Name which one you want before you type, and you are ahead of most people.

Inside ChatGPT the first two sit on a toggle right above the box you type in. The build surface sits one level up, in a product selector, not on that toggle. So switching to it means leaving the chat product, not flipping the switch. In Claude, Chat and Cowork are separate places in the desktop app sidebar, and Claude Code is its own thing.

Part 2: Read a model name

Model names carry two pieces of information and beginners mix them up.

The number is the generation. It tells you how new the underlying tech is.

The word is the tier. It tells you how much power you are using and paying for.

So a lineup might ship as one generation with three tier names on it. One example lineup used Sol for the strongest, Terra for the everyday one, and Luna for the fast cheap one. When the next generation shipped, the same three tier names came with it. Same three jobs, newer brains.

That is the whole trick. When you see a new name, ask which of the three sizes from lesson 2.1 it is. The tier names will keep coming back. The numbers will keep going up.

On prices. Guides love to quote model prices as two numbers, like "$5 in and $30 out per million tokens." Those are pay-as-you-go prices for people building software on the API. If you pay a monthly subscription and use the app, those numbers are not your bill. They are still useful for one thing: they tell you the ranking. The tier that costs five times more per token is the tier that costs you more of your monthly limit per message.

On what the tiers share. In a good lineup, all the tiers share the same limits: how much they can read at once, how much they can write at once, and how recent their training goes. The only things that change between tiers are speed, smarts, and price. Check that before you assume the cheap one is crippled.

On the knowledge cutoff. Every model was trained up to a date and knows nothing after it. Ask any of them about something recent and you get a guess, unless you paste in the source or turn search on. That is not a bug in one brand. It is how all of them work.

Part 3: Pick by where the work lives

This is the one that decides most jobs.

Cloud tools are happiest with cloud work. Research, analysis, documents, decks, charts, reports. Anything where the inputs are already online and the output is a polished file.

The Claude side sits closer to your actual machine. Cowork reads and writes the real files and folders on your computer, and it can drive your browser. That makes it strong on the messy stuff: downloads, spreadsheets, PDFs, dashboards, and cleaning up what is already sitting on your desktop.

So ask one question: is this work mostly in the cloud, or mostly on my machine? Cloud work goes to the cloud tool. Local files go to the local one.

If you run a team, there is a second factor. Cowork has had the stronger team controls: folder-level access so it only touches what you allow, approval steps before big actions, team permissions, and spend limits. That is what lets you hand a team an agent without losing oversight. The cloud side leans on its broader ecosystem instead, which is lighter to start and gives you less fine control. Check both before you commit a team, because this is exactly the kind of thing that changes.

Part 4: Pick by how sure you are

Second question: do I already know exactly what I want?

If yes, use the one that just goes. ChatGPT Work tends to start moving with less back and forth. That shows up most when building. Describe a tool or an automation and it makes a first version fast instead of asking ten questions.

If it is fuzzy or high stakes, use the one that checks in. Cowork tends to ask clarifying questions before it starts, explain what it is doing along the way, and pause before a bigger decision. When you want a second brain in the room, that is worth the extra minutes.

These are tendencies, not laws, and they shift with every release. Run one real job through each and trust what you see over what you read.

Part 5: Hand over a job, not a question

Both work surfaces are built to take a whole outcome. Most people still prompt them one question at a time and then wonder what the fuss is about.

The old way is ten prompts and you assemble the result. The new way is one brief and you get the result.

The test for whether a task belongs in a work surface: would you have opened another app to finish it? If yes, it belongs there.

Use the outcome handoff prompt below. It gives the tool the outcome, the context, the rules it cannot break, the deliverable, and what done looks like. Then let it run across as many steps as it takes.

Part 6: Know the five quiet gates

These are the reasons a request fails with no error message. Each one has burned a lot of people.

1. Connections do not work in plain Chat. The pieces that let an AI reach into your email, drive, chat app, or code live in the Work and build surfaces. Ask plain Chat to read your inbox and it cannot, and it will not always tell you why.

2. Writing to your apps is off by default. Even after you connect an app, sending and drafting usually starts switched off, and once on it still asks you every time. On business and enterprise accounts an admin turns it on per app in workspace settings. If your drafts never appear, that is why.

3. Normal folders default to read-only. A build surface can read your files and then quietly fail to change a single one. You have to grant write access when it asks. If it says it did the work and nothing changed on disk, this is it.

4. The work and build surfaces share one usage pool. They are the same engine underneath. A long building session on Tuesday can eat the budget you were saving for Friday's report, and nothing warns you at the time.

5. Features are gated by plan and by country. Some flagship features need a mid or top paid tier, and some are not available in every region. Check before you build a plan around one.

Two extras worth knowing. On Windows, the feature that lets an AI control your screen runs on your actual desktop, so you watch the cursor move and you cannot use the machine for something else. On a Mac you grant Screen Recording and Accessibility permissions first. And if you cannot find the Work toggle on the web, check the desktop app, which has been the more reliable place to find it.

Part 7: Run the comparison yourself

Reading comparisons is not the same as knowing. Do this once and you will stop asking which one is better.

1. Pick one real job you would normally break into ten prompts.

2. Decide where the work lives: cloud apps, or files on your machine.

3. Write one outcome brief using the handoff prompt below.

4. Run it in the matching tool. If it is fuzzy or high stakes, run it in the one that asks questions and answer them honestly.

5. Run the same brief in the other tool.

6. Compare the two finished deliverables. Set your default for that type of job. Not for everything, just that type.

Then keep both accounts. One paid plan and one free account still gets you a second opinion for nothing, which is the whole point of lesson 2.1's review step.

Copy and paste

The which-tool question

Run this in your head, or paste it somewhere you will see it, before any job.

Before starting any job, answer these three:

1. Does this work live in cloud apps, or in files on my machine?
   Cloud -> the cloud work surface.   Local files -> the local one.

2. Do I already know exactly what I want?
   Yes -> the tool that just executes.   Fuzzy or high stakes -> the one that checks in.

3. Does anyone besides me need approval steps or spend limits?
   Yes -> the one with folder access, approvals, permissions, and spend limits.

The model picker rule

Fill in your own current tier names and prices. Use it before you pick a model inside either brand.

Use this rule to pick a model before every task.

Ask: would a wrong answer here cost me real time or real money?
  YES -> the top tier [name: ____]

Ask: is this everyday business work I need a lot of, where speed
and smarts both matter?
  YES -> the middle tier [name: ____]

Ask: is this small, repetitive, and needed right now?
  YES -> the cheap fast tier [name: ____]

All tiers in one generation usually share the same reading limit,
writing limit, and knowledge cutoff. Only speed, smarts, and price change.

The outcome handoff

The main prompt for either work surface. Use it instead of asking a question.

Here is the outcome I want: [the finished thing, named exactly].

Context you need:
- Who this is for: [audience]
- What I already have: [files, apps, data, and where they live]
- Hard rules you cannot break: [budget, tone, format, deadline]
- What good looks like: [an example or a description]

Deliverable: [a spreadsheet / a slide deck / a doc / a small web app].
Done looks like: [what has to be true before you hand it back].

Work across as many steps as it takes. Do not hand me a list of
things for me to go do.

A one-line example of an outcome request

The shortest version of the same idea, for when the job is simple.

Analyze last quarter's sales and give me the three biggest opportunities.

Chat 1: pressure-test a decision

For the decision you keep flip-flopping on.

I keep going back and forth on a decision, and I want you to think it through with me, not just agree with me.

The decision:
[WHAT YOU'RE DECIDING]

What matters most to me here:
[YOUR PRIORITIES / CONSTRAINTS]

What I'm leaning toward, and why:
[YOUR CURRENT LEAN]

Do this:
1. Lay out the real options, including any I haven't named.
2. Give the strongest honest case FOR each one.
3. Tell me the biggest risk I'm underweighting.
4. Ask me up to 3 questions if something important is missing.
5. Then give me your recommendation, how confident you are, and what would change your mind.

Push back if my reasoning is weak. I would rather be challenged now than wrong later.

Chat 2: the message you are nervous to send

For the hard email you have rewritten five times.

Help me write something I'm nervous to send.

The situation:
[WHO IT'S TO AND WHAT'S GOING ON]

My rough draft or bullet points:
[PASTE / NOTES]

The tone I want:
[e.g. warm but firm, kind but honest, short and professional]

Give me:
1. A clean version that says the hard part clearly, with no passive-aggression and no over-apologizing.
2. A softer version and a more direct version, so I can choose.
3. One line I could add if they push back.

Keep it in my voice. Do not make it sound like a corporate template.

Chat 3: rehearse the conversation

For a raise, a negotiation, or a hard personal talk. Practice with nobody watching.

Role-play a conversation with me so I can practice before the real thing.

The conversation:
[WHAT IT IS, e.g. asking for a raise, renegotiating a contract, a hard personal talk]

Who you're playing:
[THE OTHER PERSON AND THEIR LIKELY MINDSET]

What I want to walk away with:
[YOUR GOAL]

Play them realistically and a little resistant. Raise one objection at a time and wait for my answer. Do not go easy on me.

After about 6 exchanges, stop and tell me: where I was convincing, where I got weak, and the two lines I should actually use on the day.

Chat 4: brain dump to plan

For the week where everything is in your head at once.

I'm overwhelmed and it's all in my head. I'm going to dump it, and I want you to turn it into a plan.

Everything on my plate right now:
[BRAIN DUMP, MESSY IS FINE]

Do this:
1. Sort it into: do this week, can wait, and should probably drop.
2. Flag anything that's actually urgent or has a real deadline.
3. Tell me the 3 things to do first, and why those three.
4. Point out anything I'm treating as one task that's really several.

Be honest about what I'm overcommitting to.

Chat 5: teach me this fast

For a contract clause, a health term, or a tool you do not understand yet.

Teach me [TOPIC / THE THING I'M TRYING TO UNDERSTAND] like I'm smart but completely new to it.

Do it in this order:
1. The single most important idea, the one everything else hangs on.
2. A plain, real-world analogy.
3. How it actually works, step by step, no jargon (or define any term you have to use).
4. The most common mistake beginners make.
5. One thing I could do or try today to make it click.

If I've given you something specific to react to, use my real example: [PASTE ANYTHING RELEVANT]

Work 1: research to a finished document

When you want the brief, not the notes you would turn into a brief at 11pm.

Research this and give me back a finished document I could actually use or send, not just notes.

What I need researched:
[TOPIC / QUESTION / LIST]

Who it's for and what they care about:
[AUDIENCE]

What I want at the end:
[e.g. a one-page brief, an 8-slide deck, a comparison table]

Rules:
- Put a short summary and your recommendation at the very top.
- Cite your sources so I can check them.
- Flag anything you're unsure about instead of guessing.
- Export it as [.docx / .pdf / slides] so it's ready to use.

Work 2: clean up a messy file

For the export nobody wants to touch.

Here's a messy [spreadsheet / export / list]: [ATTACH OR PASTE]

Clean it up and give it back to me usable:
1. Remove duplicates and fix the obvious formatting problems (dates, capitalization, blank rows).
2. Tell me exactly what you changed, so I can trust it.
3. Add [a summary at the top / totals / a simple breakdown by [CATEGORY]].
4. Give it back as an editable spreadsheet with real formulas, not a picture of a table.

If something looks wrong or inconsistent, point it out instead of silently fixing it.

Work 3: a recurring update that runs without you

Set once, arrives every week. Needs at least one app connected.

Set this up to run on a schedule so I don't have to remember it.

Every [DAY AND TIME], I want you to:
1. Pull [SOURCE, e.g. a channel, folder, or sheet you've connected].
2. Summarize [what changed / what needs my attention / the numbers that matter].
3. [Email it to me / drop it in a doc] as a short one-page update.

Keep it skimmable: the 3 things I actually need to know at the top, details below. Tell me now if you need me to connect an app first.

Work 4: combine scattered pieces

Six files and a notes doc into one thing you can send.

I have a bunch of scattered pieces and I need one clean [proposal / report / summary / plan] out of them.

Everything I have:
[ATTACH FILES / PASTE NOTES]

The format and tone I want:
[e.g. our house style, formal, friendly]

Who's going to read it:
[AUDIENCE]

Do this:
1. Pull it into one coherent document.
2. Where there are gaps, fill them with clearly-labeled assumptions I can check.
3. Flag anything that contradicts itself across the pieces.
4. Export as [.docx / .pdf] so it's ready to send.

Work 5: inbox triage with drafts written

Sorting is the easy half. The value is the drafted replies.

Go through [my inbox / messages] from [this week] that I haven't dealt with.

1. Sort them into: needs a real reply, can wait, and ignore.
2. For the ones that need a reply, draft a response in my voice, short and human.
3. Flag anything time-sensitive, or anything that could quietly become a problem.

Do not send anything. Just leave the drafts ready for me to review and send.

Build 1: clean up a folder of files

The lowest-risk way into the build surface. Copy the folder first.

Open this folder: [FOLDER]. I have a mess of files I was never going to fix by hand.

What I want:
1. [Rename / reorganize / convert] everything by this rule: [YOUR RULE, e.g. rename to YYYY-MM-Description, sort into subfolders by year].
2. Show me your plan first and let me approve it before you touch anything.
3. Then do it, and tell me exactly what changed.

Grant write access when it asks, or nothing will actually change on my computer.

Build 2: turn a monthly task into a button

For the same tedious hour you spend every month.

Every [WEEK / MONTH] I do this by hand and I'm tired of it: [DESCRIBE THE TASK, STEP BY STEP].

Build me something I can re-run that does it for me:
1. It takes [INPUT, e.g. this kind of file or export] and produces [OUTPUT, e.g. this formatted summary] the same way every time.
2. Explain in plain English how to run it, like I've never done this before.
3. Make it easy to tweak later when the inputs change.

Start by telling me what you'll build and what you need from me.

Build 3: a simple tool people can use

A tracker, checklist, or calculator, published to a link. Publishing may need a paid tier.

Build me a simple [tracker / checklist / calculator / dashboard] for [PURPOSE, e.g. tracking projects, chores, orders, a budget].

1. Let me [add / update / check off] things easily, no spreadsheet skills needed.
2. Keep it clean and obvious to use.
3. Publish it so [only me / my team / my family] can open it in a browser with a link.

Ask me what fields or columns it should have before you build, if that would help.

Build 4: a one-page site

A real page, live, with a form. Publishing may need a paid tier and may not be available in every country.

Build me a clean one-page [website / landing page / event page] for [PURPOSE, e.g. my new service, an event, a portfolio].

Include: [e.g. a headline and short description, an email sign-up form, a simple pricing section, an RSVP button].

Style: [e.g. minimal, warm, bold; match my colors: [COLORS]].

When it's ready, publish it to a link I can share, and walk me through anything I need to do on my end.

Build 5: explain what you own, then improve it

Use the build surface as a tutor on your own setup.

Open [this folder / these files]: [WHAT IT IS].

First, before changing anything: explain in plain English what each part does and how it fits together, like you're teaching me my own setup.

Then make this one improvement: [SPECIFIC CHANGE, e.g. make it work on phones, fix the thing that's broken, add this feature].

Go one step at a time and tell me what you're doing as you go, so I actually learn it instead of just getting a result.

Watch out

You ask which one is better

That question has no answer and it keeps you comparing instead of working. Ask where the work lives and how clear the outcome is.

You prompt a work surface one question at a time

These tools are built to take a whole job. A question gets you an answer and leaves the work with you. Name the finished deliverable instead.

You expect app connections to work in plain chat

They only run in the work and build surfaces. The request will fail quietly. If an AI cannot see your inbox, check which surface you are in before you blame the model.

You assume your files changed

Normal folders start read-only. Grant write access when it asks, then open the folder yourself and look.

You burn one budget on the other job

The work and build surfaces draw from the same pool. If Friday's report matters, do not spend Tuesday on a long building session.

You get loyal to a logo

Whoever is ahead this month may be behind next month. Keep one paid plan and one free account so you always have a second opinion available for nothing.

You read a tier name as a version number

The tier names carry forward into the next generation. The number is what tells you how new the tech is.

You point an agent at your whole computer

These tools can rename, move, and change real files. Grant access to one folder, not everything. Copy anything you care about before the first run. Show me your plan first is a line worth putting in every build prompt.

Check yourself

• [ ] I can name the three shapes and match them across both brands.

• [ ] I know which surface I am in when I type, and I choose it on purpose.

• [ ] I can point at a model name and say which part is the generation and which is the tier.

• [ ] I ran one real job as an outcome brief, not as ten questions.

• [ ] I ran the same brief in both tools and compared the deliverables.

• [ ] I checked whether write access was granted before assuming files changed.

• [ ] I have one paid plan and one free account on the other side.

Do this now

Pick one job from this week that you would normally break into ten prompts. Write it once using the outcome handoff prompt, and run it in whichever tool matches where the files actually live.

Keep going

Next in AI for Entrepreneurs:

• Part 10: Your starter stack, and what to ignore

• Part 11: Day one: the five things to do first

• Part 12: Day two and three: the next ten things

Every part of this course sits in reading order in the free AI for Entrepreneurs course outline.

This part is free, start to finish. Subscribing keeps the rest of the course coming and opens the paid parts.

Tell me in the comments which of the two you use, and for what, and I will point you at the right next part.

One of those settings quietly spends your limit on work that never needed it. The other hands you a shallow answer on the decision that mattered most this month. Both feel exactly the same while they are happening.

What you will be able to do

• Decide fast or deep on any task in about ten seconds, using three questions.

• Escalate to a bigger model only when the work has earned it, and know when it has.

• Set the effort level so simple work stays cheap and hard work gets real thought.

• Use your most expensive model to build things a cheaper model can run for months.

• Get an AI to flag what it is unsure about instead of guessing at you.

This is for anyone who wants a rule instead of a guess. Beginner friendly, and the model names are marked as examples, because names go stale in months.

The proof behind this

Liz Elliott runs 33 active social brands and runs her whole operation on AI automations across several computers. At that volume you cannot open the biggest model for everything, so the picking rule is not theory.

What this lesson is

Lesson 2.3 of the free AI for Entrepreneurs course, in Module 2: Choosing Your AI.

Lesson 2.1 gave you three sizes: fast, default, deep. This lesson is about picking one, on a real task, in a few seconds.

Here is the thing that makes it simple. Capability costs money. The deep model is not free. It costs more of your usage limit, it takes longer to answer, and on a small job it gives you nothing back for the extra spend. Using it for everything is like taking a taxi to the end of your own driveway.

Compare it to hiring help. For sorting a box of receipts you want somebody fast and cheap who will not stop to philosophize. For deciding whether to sign a five-year lease you want somebody careful, expensive, and slow, and you want them to think out loud so you can follow their reasoning. Same person for both jobs is the wrong answer twice.

So you match the model to the stakes, not to the mood. Most of your week is receipts. A small slice is the lease.

There is a second dial next to the size. Effort, or thinking, or reasoning depending on which app you are in. It tells the model how hard to work before it answers. Turn it up and the answer takes longer and comes back more careful. Turn it down and it comes back quick and cheap. This dial matters as much as which model you picked, and most people never touch it.

One more thing worth knowing before you start. The specific model names in every guide you read, including this one, go stale in months. Prices change. New models ship and old ones retire. The questions below do not change. Learn the questions. The names live in a reference table at the end, clearly marked as examples.

New here? Use the free AI for Entrepreneurs course outline to read this series in the best order.

Do it

Part 1: Ask three questions

Before you type, answer these. It takes about ten seconds.

Question 1: Does this need thinking, or just doing?

Thinking means weighing options, finding a root cause, planning an approach, making a judgment call. Doing means the decisions are already made and something needs writing, formatting, sorting, or summarizing.

Question 2: Is the job long, or short?

Long means a big document, a pile of files, many steps, or work that has to hold a lot in mind at once. Short means one thing, one answer.

Question 3: What happens if it is wrong?

If a wrong answer costs you nothing but thirty seconds, that is a low-stakes task. If it costs money, a client, a deadline, or a legal problem, it is not.

Now read the row that matches.

The job / Model size / Thinking turned up?

• Short and simple. Lookups, definitions, labels, tidying text, sorting a list

• Model size: Fast

• Thinking turned up?: No

• Bulk and repetitive. The same small task fifty times

• Model size: Fast

• Thinking turned up?: No

• Normal daily work. Emails, drafts, reports, summaries, everyday coding

• Model size: Default

• Thinking turned up?: No

• Long documents and long jobs, but the thinking is clear

• Model size: Default or deep

• Thinking turned up?: No

• Hard reasoning. Strategy, root causes, competing options, big plans

• Model size: Deep

• Thinking turned up?: Yes

• High stakes. Money, legal, medical, anything you will cite later

• Model size: Deep

• Thinking turned up?: Yes

• Anything you will not double-check yourself

• Model size: Deep

• Thinking turned up?: Yes

If you do not know, use the default model. That is what it is for. Around eighty to ninety percent of most people's work runs fine there.

Part 2: Escalate on evidence, not on a hunch

Do not open the deep model because the task feels important. Open it when the work has proven it needs it. The ladder:

1. Start on your default model. Every time.

2. If the answer is thin, add context before you switch. Most weak output comes from a weak ask. Give it more detail, say who the output is for, and say what a good version looks like. Try again on the same model.

3. If it is still thin after two honest tries, escalate. Two disappointing answers is your signal, not one.

4. Do not drag the mess with you. If you are twelve messages deep in a tangled thread, do not just switch the model picker. Start a clean chat on the bigger model and paste in a proper brief. A big model gets far more out of a clean, well-structured brief than out of a long messy thread.

5. Drop back down when the hard part is over. Once the plan exists, the writing and building can go to a cheaper model.

Part 3: Move the effort dial

Effort tells the model how hard to think before it answers. Most apps offer something like four levels, and the wording varies.

Effort level / Best for / Watch out for

• Low

• Best for: Quick rewrites, formatting, short summaries, one-off simple tasks

• Watch out for: Do not use it for real reasoning

• Medium

• Best for: Emails, content drafts, normal coding tasks

• Watch out for: A good baseline when you want to keep costs down

• High

• Best for: Analysis, multi-step work, code that has to work, anything where a better answer saves you time

• Watch out for: On some models this is already the default

• Max

• Best for: Big migrations, hard debugging, the genuinely gnarly problems

• Watch out for: Uses far more of your limit, so turn it on deliberately

The pattern to run: start at high, drop to medium when the task is routine, and save max for jobs that have already beaten you once. Do not leave it on max for chat-style work. It burns your limit fast for no gain.

Two notes. Some newer models handle this by themselves and adjust the effort to the question, so there is no dial to find. And in the terminal version of Claude, effort is a slash command: /effort high. In the app, look for a thinking or reasoning setting.

Part 4: Never hand a long document to the small fast model

This is the single most expensive mistake in this lesson, because it fails quietly.

Small fast models lose track of detail across a long file. They do not warn you. They give you a confident summary that is missing the thing in the middle you actually needed. Bulk jobs, yes. Long documents, no.

Long files go to the default or the deep model. If a model advertises a huge memory for one document, that is the one to use for the hundred-page contract.

Part 5: Spend the deep model on things you keep

The best use of an expensive model is not doing one task well. It is building something a cheaper model can run for months. Three moves, in this order.

Move 1: Upgrade the instructions you already use.

You probably have saved instructions somewhere. Project instructions in your AI app, saved Skills, or instruction files in a project folder. These are usually vague, because you wrote them in a hurry. A big model rewrites them into something specific, and the upgraded version keeps working after you drop back down to a cheaper model.

Do one at a time. Copy your instructions out, run the upgrade prompt below, and paste the improved version back in.

If your instructions live in files on your computer, there is a version of that prompt that edits them directly. Before you run it, copy the whole folder somewhere safe. That prompt rewrites every instruction file in one pass and does not stop to ask. Review each change afterward and keep the good ones.

Move 2: Unstick the problem you have been stuck on for weeks.

The trick is not a smarter question. It is packaging the whole mess and moving it somewhere clean.

1. In the chat where you hit the wall, run the capture prompt below. It turns the entire thread into one complete handoff: the goal, what exists now, every attempt, every error, every dead end, the rules you cannot break, and the exact place you are stuck.

2. Open a brand new chat. Pick your deep model. Set effort high.

3. Paste the handoff and run the solve prompt.

You get a root cause instead of another surface fix.

Move 3: Build a plan a cheaper model can run.

Have the deep model build the whole plan once. Phases, tasks broken small, and a clear description of what "done well" looks like for each task. Then hand one step at a time to a cheaper model. You keep the good thinking without paying for the thinker every day.

The part that makes this work is the acceptance criteria. If a task is vague, a cheap model will guess. If a task says exactly what finished looks like, it will not.

Part 6: Make it show doubt

Newer models are better at saying they are unsure instead of making something up, but you still have to ask. Add one of the uncertainty lines below to any answer you plan to rely on. Use them for research, legal questions, health questions, and anything with a number, date, quote, or name in it that you might repeat later.

If you get answers that open by agreeing with you before they have thought about it, add the pushback line too.

Part 7: When the answer feels thin, fix your words first

Newer models tend to read your prompt more literally and match the answer length to how big the ask sounds. That is usually why an answer comes back shorter and flatter than you expected. Five fixes:

If the answer is / Do this

• Too literal, it only did the one thing you said: Spell out every step you want

• Too short: Ask for "a full breakdown, not a summary" and say how long

• Overblown, it went strange on you: Take out the all-caps pressure words like CRITICAL and MUST, and write a normal direct instruction

• Stopping at the bare minimum on creative work: Add "go beyond the basics" and name the sections you expect

• Not thinking hard enough: Turn on the thinking mode, or add the think-carefully line

The before-and-after pairs are in the prompts section. Try the same task twice, once each way, and you will see the gap.

Part 8: The part that survives every model change

A new model ships every few weeks. Chasing every one of them is tiring, and it is not where the real gain lives. What makes your AI good is what it knows about you. Your work, your projects, your goals, the way you write, your rules, what you are doing right now.

If all of that lives inside one app, you start from zero every time that app is down, worse at something, or replaced by something better.

So keep it in plain files on your own computer. Make a folder and put three plain text files in it:

File / What goes in it

• `about-me`: Who you are, your role, your goals

• `how-i-write`: Your tone, plus a few real samples of your actual writing

• `projects`: What you are working on right now

Paste those into any AI before a real request and watch the answer change. Module 5 goes much deeper on this. Ten minutes now is enough to feel the difference.

Reference: what the model names meant

Everything in this table is an example of how the naming worked at the time these guides were written. Every name and number here will drift. The sizes in Part 1 will not.

Name pattern / Size slot / What it was for

• Haiku

• Size slot: Fast

• What it was for: Classification, extracting fields, formatting, batch work

• Sonnet

• Size slot: Default

• What it was for: The daily driver. Most work, most days. Planning tasks and using a browser or terminal on its own

• Opus

• Size slot: Deep, and the long-haul worker

• What it was for: Long jobs that run for hours, deep research, production-quality work, natural long-form writing

• Fable

• Size slot: The ceiling

• What it was for: The hardest problems, the highest stakes, the decisions you cannot get wrong

A few details from those releases, kept only because they show you what to look for in the next one:

• Version numbers do not line up across sizes. One lineup ran Haiku 4.5 alongside Sonnet 5 and Opus 5, because there was no Haiku 5. Do not assume a matching number exists.

• Prices are quoted per million tokens, as input and output. A token is a chunk of text, very roughly three quarters of a word. Those are pay-as-you-go prices for people building on the API. If you use the app on a Free, Pro, or Max plan, a new model does not change your bill. It changes what you can open and how fast you hit your limit.

• Published prices disagreed across guides, because some quoted a launch discount and some quoted the standard rate. The rule: never trust a price in a guide, including this one. Open the vendor's own pricing page.

• Context windows got large. Around a million tokens became normal at the top, which is roughly a very long book in one go. That is what makes a big model good at the hundred-page contract.

• Benchmark scores get quoted a lot and they are tests, not your workload. A model that scores well on fixing real code may still be wrong about your spreadsheet. Use scores for a rough feel, then test on your own real task.

• Token counting changed between versions, so the same job could use around thirty percent more tokens on a newer model even at the same price per token. Compare the cost of the whole task, not the sticker price.

• Some models hand sensitive topics to a different model on a small share of chats, as a safety feature. If the tone shifts on a touchy subject, that may be why.

• New releases come with claims. One release was described as no longer making things up. No model stops making things up. Treat every launch claim as a direction, not a promise, and keep checking the output.

Copy and paste

Your model rule, one page

Keep this where you can pull it up mid-chat. Fill in your own current model names.

MODEL RULE, WHICH ONE TO OPEN

THE ONE RULE:
Don't pick the "best" model, pick the right one for the job.
Capability costs money, so match it to the stakes.

FAST MODEL [name: ____]
Reach for it when the task is simple and you want it instant and cheap:
quick lookups, sorting, labeling, tidying text, bulk repeats.
Never hand it a long document.

DEFAULT MODEL [name: ____]
Reach for it for everyday output: emails, drafts, reports, summaries,
research that needs some thought, normal coding, document work.
This is the tab that stays open. About 80-90% of the week lives here.

DEEP MODEL [name: ____]
Reach for it when the problem is genuinely hard, the reasoning is long,
or the decision is high-stakes and you'd want your smartest advisor in
the room. Strategy, tough analysis, difficult code, research you can't
afford to get wrong.

THE DEFAULT SETUP:
- Live in the default model for almost everything.
- Escalate the moment you're stuck, the stakes are high, or the first
  two answers weren't good enough.
- Drop down to the fast model for bulk, boring, or speed-critical work.
- Turn thinking on for decisions. Turn it off for emails.

HONEST NOTE:
Every model can both reason and produce work. The real choice is top
capability versus best value. When in doubt, start on the default and
only pay up when the problem earns it.

Ten habits that work on any model

These do not care which model you are in. Keep them next to the rule above.

1. ASK FOR 3 OPTIONS.
End any request with "give me three versions: safe, bold, and weird."
You go from one flat answer to a real range in one shot.

2. MAKE IT INTERVIEW YOU.
"Before you answer, ask me the 5 questions you most need answered to do
this well." It gathers context instead of guessing.

3. SHOW IT A GOLD STANDARD.
Paste an example of great work and say "match this quality and format."
AI copies excellence far better than it follows adjectives.

4. TELL IT WHO TO BE.
"You are a [specific expert] with 20 years of experience in [niche]."
A specific role pulls a sharper answer than a blank prompt.

5. SLOW IT DOWN.
"Think through this step by step before you answer." For anything with
logic or numbers, this alone lifts the quality.

6. RUN A RED TEAM.
"Try to prove me wrong. Where does this fall apart?" Turns a cheerleader
into a pressure-tester.

7. USE PROJECTS FOR ANYTHING YOU REPEAT.
Put your context, rules, and examples in a Project once, and every chat
inside it already knows your world.

8. BUILD A SKILL FOR ANY REPEATED TASK.
If you explain the same process twice, turn it into a Skill so you never
explain it again.

9. LET IT SEE.
Paste a screenshot and ask "what's off here?" AI reads design,
dashboards, and error messages straight from an image.

10. KEEP A MEMORY FILE.
Keep a running note of how you work and your preferences, and paste it at
the top of big tasks so the output sounds like you, not like a stranger.

WHICH MODEL FOR THESE:
Run the everyday ones on your default model. Escalate the hard ones.

Deep research report

Switch to your deep model first. Use this when you want wide research turned into something you can decide from.

Create a deep research report on [topic].

Use the context I provide and any research tools available to you.

Organize the report so I can make a decision from it.

Include:
- The main findings.
- The strongest evidence.
- Open questions.
- Risks or caveats.
- A clear recommendation based only on the evidence.

Make sense of a giant document

For a long contract, report, transcript, or PDF. Deep model.

Read this full document: [attach or paste document].

Do not skim.

Give me:
- The plain-English summary.
- The most important details.
- Anything risky or confusing.
- The parts I should read myself.
- The questions I should ask next.

Plan and build a real project

For a big multi-step build where the pieces have to stay connected.

Help me plan and build [project].

Here is the full context: [paste goals, constraints, files, users, deadline, and requirements].

First, make the plan.

Then break it into steps.

Then work through the steps without losing the full goal.

Stop and ask me only when you need a decision I have not given you.

Turn a pile of notes into a plan

For a year of notes, voice transcripts, or brain dumps you never processed.

Read all of these notes: [paste or attach notes].

Find the patterns I am too close to see.

Turn them into:
- The main themes.
- The decisions I keep avoiding.
- The ideas worth acting on.
- The ideas to drop.
- A simple plan for the next phase.

Untangle a mess

For the process, spreadsheet, or system nobody understands anymore.

Untangle this system: [describe or attach the messy process, spreadsheet, migration, or workflow].

First, map what is happening now.

Then find the failure points.

Then give me the cleanest path to fix it.

Include:
- What to do first.
- What not to touch yet.
- What could break.
- How to verify each step worked.

Upgrade one saved instruction (copy and paste version)

Move 1, one at a time, in any AI app. Safe for beginners.

Here is one of my skills, or my Project's custom instructions. Make it dramatically better.

1. Tell me in one line what it is for.
2. Audit it hard. Where is it vague, incomplete, or likely to give inconsistent results? Look for missing context, undefined terms, no examples, no defined output format, and no rules for what it should never do.
3. Rewrite it into a sharper version with specific instructions, a clear step-by-step process, the exact output format I want, one or two concrete examples of a great result, and explicit rules for what it should never do. Keep my voice and intent.
4. In a few bullets, tell me what you changed and why it will perform better.

Give me the upgraded version as one clean, copy-paste-ready block so I can drop it straight back in.

Here it is:

[paste one skill, or your Project's custom instructions, here]

Upgrade every instruction file at once (files version)

Move 1, for when your instructions live in files an AI can edit. Copy the whole folder to a safe place before you run this. It rewrites files directly and it is told not to stop and ask.

Go through every skill, workflow, and instruction file in this project: my CLAUDE.md, everything in .claude/skills, and my subagents in .claude/agents. Read each one, then for each:

1. Tell me in one line what it is for.
2. Audit it hard. Where is it vague, incomplete, or likely to give inconsistent results? Look for missing context, undefined terms, no examples, no defined output format, and no rules for what it should never do.
3. Rewrite the file in place to fix all of that: specific instructions, a clear step-by-step process, the exact output format it should produce, one or two concrete examples of a great result, and explicit do-nots. Keep my intent and my voice.

When you are done, give me a short summary of what you changed in each file, tell me which skills overlap or should be merged, and name the one skill or workflow I am clearly missing for the kind of work I do. Don't ask me anything until you have done all of this.

Capture a stuck chat

Move 2, step 1. Run this inside the conversation where you got stuck.

Turn this entire conversation into a single, complete markdown handoff so I can paste it somewhere fresh and a new AI can pick up with zero prior context and full understanding.

Use these headers:
- Goal: what I am actually trying to accomplish, and what outcome it affects.
- Current state: what exists right now, what is working, what is not.
- What we have tried: every approach we attempted and exactly what happened with each one, including the precise errors, dead ends, and anything we ruled out and why.
- Constraints and rules: anything that must stay true, the tools I am using, things I cannot change.
- Key details: the specific files, names, numbers, links, or snippets that matter.
- Where I am stuck: the exact roadblock, described precisely.

Be thorough and specific. Do not summarize away any detail a fresh AI would need to actually solve this. Output only the markdown block, nothing else.

Solve it in a fresh chat

Move 2, step 3. New chat, deep model, effort high, then paste the handoff into this.

Below is the full context of a problem I have been stuck on. I moved it into a fresh chat on your most capable settings so you have room to actually solve it.

Do this:
1. Diagnose the REAL root cause of the roadblock, not just the surface symptom. If my own assumptions about what is wrong look mistaken, say so directly.
2. Lay out the 2 to 3 best ways to solve it with the tradeoffs of each, then pick the one you would actually do, and why.
3. Then either fully solve it and give me the finished result, or, if it genuinely cannot be finished in one pass, complete as much as you can and give me the next three concrete steps in order.

Rules:
- Only stop to involve me if a payment, a login, or a decision that is truly mine to make is required. Otherwise keep going and solve it. Do not ask me questions you can reasonably answer yourself.
- Do not violate any of the constraints or rules listed in the context.
- Lead with the solution, then show your reasoning briefly so I can follow it.

Here is the full context:

[paste the markdown from the capture prompt here]

Build a plan a cheaper model can run

Move 3, step 2. Deep model, effort high.

You are my strategist. Build me a detailed, durable [content plan / project roadmap / launch plan] for [describe the goal, the timeframe, and any context] that I can execute against for the next few months.

Important: I am going to build this once with you now, then have a cheaper, less capable AI actually run it day to day. So it has to be clear and specific enough that a simpler model could follow it without you.

Include:
1. Strategy: the overall approach and the single most important outcome we are driving toward.
2. Phases: the major phases in order, each with one clear goal.
3. Tasks: inside each phase, the specific tasks or deliverables, broken down small enough to hand to a simpler AI one at a time.
4. Acceptance criteria: for every task, spell out what "done well" actually looks like, so it is obvious when a piece is truly finished and not just technically done.
5. Inputs and risks: what I will need, including tools, assets, and info, plus the bottlenecks or risks to watch for.
6. Rhythm: a simple weekly or daily cadence so I always know exactly what to work on next.

Make it specific to my situation, not generic filler. Before you build it, ask me up to five sharp questions that would make the plan meaningfully better, then deliver the full plan.

Here is my context:

[paste your niche, audience, offer, current numbers, timeline, and any constraints]

Hand one step to a cheaper model

Move 3, step 3. Use this every time you want the next piece done.

Here is a plan my most capable model built. Execute the next step exactly as written, meet its acceptance criteria, and flag anything that does not line up with the plan. Do not redesign the strategy, just run it well and give me the finished deliverable.

Here is the plan, and the step I want done:

[paste the plan, and name the step]

Ask for uncertainty

Add to anything you are going to rely on.

Give me the answer and flag anything you're not sure about with a confidence score.

Ask it not to guess

Use on factual questions.

If you don't know, say you don't know. Do not guess.

Ask it to mark risky facts

Use before you copy anything into a document you will send.

Mark any specific fact (number, date, quote, name) that you'd want me to verify before using.

Stop the flattery

Use when the answers keep opening by agreeing with you.

Do not flatter me or agree by default. Push back if I'm wrong.

Prompt fixes, before and after

Five pairs. The first line in each pair is the version that produces a thin answer.

Too vague:

Clean up this draft.

Better:

Fix any typos, tighten any sentence over 25 words, and remove anything unnecessary. Keep the structure.

Too small an ask:

Summarize this.

Better:

Give me a full breakdown of this in 3-4 paragraphs, not a summary. Cover the main argument, the evidence, and the counterargument.

Too shouty:

CRITICAL: You MUST use the calculator tool for ANY math.

Better:

Use the calculator for math problems.

Too thin for creative work:

Design a landing page for my app.

Better:

Design a landing page for my app. Go beyond the basics. Include hero, social proof, feature breakdown, pricing, FAQ, and a final CTA.

Not thinking hard enough:

Think carefully through the problem before responding.

Terminal commands

If you use the terminal version of Claude, these two switch the model and the effort.

/model

/effort high

Watch out

You open the deep model by default

It feels responsible. It is expensive, slower, and on routine work it can be worse, because it over-explains and adds complexity you did not ask for. Default first.

You never open the deep model because it costs more

The usage cost is real, and so is the quality gap on genuinely hard problems. If you are making a real decision, ten minutes of a deep model with thinking on is worth the spend.

You switch models instead of fixing the brief

Two thin answers in a row usually means a thin ask. Add context, say who it is for, say what good looks like, and try the same model again before escalating.

You hand a long document to the fast model

It will not warn you. It will summarize confidently and drop the middle. Long files go up a size.

You leave effort on max

It is a switch you flip for a hard job, then flip back. Left on, it burns your limit on emails.

You let an AI rewrite all your files in one pass

The bulk upgrade prompt in this lesson edits every instruction file in a project and is told not to stop and ask. Copy the folder somewhere safe first. Review every change before you keep it. There is no undo.

You trust a confident answer on facts

Confidence is not accuracy. Add an uncertainty line to anything with a number, date, quote, or name in it, and check the ones it flags.

You build a plan only your best model can run

If the steps are vague, the cheap model will guess and you will not find out until the work comes back wrong. Force acceptance criteria on every task. That is what makes a plan portable.

You read a model name as a permanent rule

Every name in the reference table will be replaced. Check the vendor's current model page and pricing page before you make a decision based on either.

Check yourself

• [ ] I can name the three questions and answer them without notes.

• [ ] I know which model my app opens by default, and I chose it on purpose.

• [ ] I found the effort or thinking setting in my app, or confirmed my model handles it by itself.

• [ ] I escalated one task only after two thin answers, not on a hunch.

• [ ] I ran the capture prompt on a stuck problem and solved it in a fresh chat.

• [ ] I upgraded at least one saved instruction with my deep model.

• [ ] I copied any folder to a safe place before letting an AI edit it.

• [ ] I added an uncertainty line to one factual request.

• [ ] I made the three context files and pasted them into one real request.

Do this now

Take the problem you have been stuck on for weeks. Open that old chat, run the capture prompt, then paste the result into a brand new chat on your deep model with effort turned up, and run the solve prompt.

Keep going

Next in AI for Entrepreneurs:

• Part 9: ChatGPT and Claude, compared fairly

• Part 10: Your starter stack, and what to ignore

• Part 11: Day one: the five things to do first

Every part of this course sits in reading order in the free AI for Entrepreneurs course outline.

This part is free, start to finish. Subscribing keeps the rest of the course coming and opens the paid parts.

Tell me in the comments the task you have been running on the wrong setting, and I will point you at the right next part.

One account holds four separate tools. Three of them do things chat cannot: read a folder of your own files, build software that actually runs, and make the visual instead of describing it. Skip those and you use a quarter of what you bought.

What you will be able to do

• Name the four Claude tools and say in one word what each is for.

• Open each one, and know which plan you need first.

• Point Claude at a folder of your own files and get real work back.

• Ask Claude to build you a small tool without writing any code.

• Run Claude inside a free code editor so your files and your AI sit in one window.

This is for anyone with a Claude account, free or paid. Some of the four need a paid plan, and the lesson says which before you get attached to one you cannot open.

The proof behind this

Liz Elliott built four software products with AI help: PRISM, AutoViral, UPRISE Social Studio, and Profounder, and she never studied computer science. That happened because she stopped treating AI as a chat box.

What this lesson is

Lesson 2.2 of the free AI for Entrepreneurs course, in Module 2: Choosing Your AI.

Most people meet Claude as a chat box, use it like a smarter search engine, and stop there. Chat is one door. There are four.

Picture a gym membership. One card, four rooms. The weights room, the pool, the studio, the sauna. If you only ever walk into the weights room, the card is not broken. You are just using a quarter of it.

Claude works the same way. The four rooms are Chat, Cowork, Code, and Design. They live in different places and they are good at different things.

Chat is for thinking. You talk, it talks back, nothing gets saved or sent anywhere.

Cowork is for doing work. It reads your files, runs jobs with many steps, and hands you a finished document instead of instructions for making one.

Code is for building. It writes, tests, and fixes software on its own. You do not need to know how to code. You describe what you want.

Design is for creating. You describe a deck, a page, or a set of graphics, and it builds the actual visual while you watch.

One warning before you go looking. Not all four are on every plan. Free accounts get Chat. The others need a paid plan, and the heavier ones need a higher paid tier. Check what your plan includes before you get attached to a tool you cannot open yet. Nothing in this lesson is worth upgrading for until you have tried the thing you already have.

New here? Use the free AI for Entrepreneurs course outline to read this series in the best order.

Do it

Part 1: Chat, to think

Where to find it: claude.ai, or the desktop or mobile app.

This is the plain chat box. Three things worth doing with it this week:

1. Explain something confusing. Paste in the thing you have been avoiding. A lease. A contract. A tax form. Bloodwork. A confusing email. Then use the plain-English prompt below. This is usually faster and clearer than searching for it.

2. Test a big choice before you make it. Pick something you are weighing up: leaving a job, signing a lease, launching an offer. Run the pre-mortem prompt. It surfaces the problems while you can still avoid them.

3. Talk through what is in your head. Use voice and just dump it out. Then ask Claude to reflect it back, find the patterns, and help you name what you are dealing with. This is not care and it is not treatment. If you are struggling, talk to a real person. Use this for the ordinary noisy days, not the hard ones.

Part 2: Cowork, to do work

Where to find it: the Claude desktop app, then Cowork in the sidebar. It also runs on web and mobile, so a job you start on your laptop can keep going while you are out, and it can ask for your approval from your phone.

Cowork does not answer questions. It finishes jobs. Three to try:

1. Build a Monday morning briefing. Connect Gmail, Calendar, and Slack, then run the briefing prompt below. You read one page with your coffee instead of opening four apps.

2. Build a dashboard that stays current. Point Cowork at your sales tracker or a spreadsheet and ask for a live dashboard showing your top numbers, the change week over week, and the trend. A live dashboard pulls fresh data each time you open it, instead of freezing the numbers from the day you made it.

3. Turn meeting notes into follow-ups. After a meeting, hand it your notes. It can pull out the action items, draft a follow-up email for each person with their own tasks in it, and write the agenda for next time.

Cowork can read and write real files on your computer. That is the point of it, and it is also the risk. Start it on a folder of copies, not your only copy of anything. Read what it says it is about to do before you approve.

Part 3: Code, to build

Where to find it: claude.com/product/claude-code. It also runs on web and mobile, so you can start a job from your phone and review it when you get back to your desk.

Claude Code writes, tests, and fixes software on its own. You describe what you want in plain English. Three builds worth trying:

1. A tracker you keep looking for and never find. A meal planner. A workout log. A content idea inbox. A price comparison sheet. A habit tracker. It can build the whole thing, including where the data lives and how you log in.

2. A weekly job you keep doing by hand. Two examples are in the prompts below. These are the kind of small automations people normally pay a developer to set up.

3. A small internal tool. Look at the little software subscriptions you pay for every month, the ones costing roughly thirty to a hundred dollars. A project tracker, a time logger, a client intake form, an invoice generator. Some of those can be built once and run cheaply. Not all of them. Hosting, updates, and backups are real work, so treat this as an experiment before you cancel anything.

From Liz's playbook: Liz Elliott is not an engineer and never studied computer science. She built four software products with AI help: PRISM, AutoViral, UPRISE Social Studio, and Profounder. "You do not need to know how to code" is not a slogan here. It is the normal path now.

Part 4: Design, to create

Where to find it: claude.ai/design.

Design works like a creative director. You describe what you need and it builds the visual, with a live preview while it works. Save your brand colors and fonts once, and everything after that can match them.

Three things to make:

1. A pitch deck. Use the deck prompt below, then export to PowerPoint or Canva to finish it.

2. A one-pager you can email. An investor one-pager, a service offer, a media kit, a course outline. One clear sentence in, one clean PDF out, instead of writing another long email.

3. A reusable set of social graphics. Use the templates prompt, then export. Later you swap the text instead of rebuilding the design.

Part 5: The fifth place, if you want more control

The desktop app is one window with limited ways to arrange things around your real files. If you want your files, your AI, and your other tools in one place, run Claude Code inside VS Code.

VS Code is Visual Studio Code, a free app from Microsoft. No trial, no credit card, Mac and Windows. It was built for developers, but you do not need to be one to use it this way. You still type to Claude in plain English.

You only need three places in it: the Extensions panel, the File menu, and the Claude side panel.

Set it up once:

1. Go to code.visualstudio.com. Click download and install it like any other app.

2. Open VS Code. On the left edge, click the icon that looks like four small squares. That is the Extensions panel. The shortcut is Cmd+Shift+X on Mac, Ctrl+Shift+X on Windows.

3. In the search box, type Claude Code. Check that the publisher is Anthropic before you install, so you do not pick up a lookalike.

4. Click the blue Install button and wait a few seconds.

5. Sign in with the Claude account you already use. Claude opens in a side panel next to your files.

You do this once. After that you just open VS Code, open your folder, and use the side panel.

Use it day to day:

1. Top menu, File > Open Folder. Pick the folder holding what you are working on. Now Claude can see those files.

2. Click into the Claude side panel and type what you want in plain English.

3. It works on the real files in that folder and shows you what it did.

Claude in this setup can create, change, and delete real files on your machine. Before you point it at anything you care about, make a copy of the folder. Read what it proposes before you approve it.

Why bother: you are not stuck with one AI. Other AI tools have their own VS Code extensions, and each gets its own side panel in the same window. The best tool for your work changes over time, and this way you can run two side by side and compare on the same files.

Five add-ons worth having. Install any of them the same way: open Extensions, search the name, click Install.

Extension / What it does for you

• Markdown All In One: Claude often writes in Markdown, the simple format that uses symbols for headings and lists. This gives you a clean preview, so drafts and summaries are easier to read.

• Code Spell Checker: A free spell checker. It underlines typos as you type, the same way email does.

• Live Server: If Claude builds a simple web page, click Go Live at the bottom of the window and watch it in your browser. It refreshes when the file changes.

• Rainbow CSV: CSV files are spreadsheets saved as plain text. This colors each column so the rows line up and you can actually read them.

• PDF Viewer: Open and read PDFs in the same window, instead of switching apps.

Install the ones that solve a real problem for you. Skip the rest.

The shortest test: make a new empty folder on your desktop, open it in VS Code, drop in three documents, and ask Claude to do something simple with them. That one test tells you whether this setup suits you.

Copy and paste

Chat: explain a confusing document

Paste the document first, then this.

ELI5. What does this actually mean and what should I be worried about?

Chat: pre-mortem a big decision

Describe the decision, then run this before you commit to anything.

Pretend it's 6 months from now and this decision blew up. Walk me through exactly why.

Cowork: Monday morning briefing

Run this after you have connected your email, calendar, and chat app.

Pull what happened over the weekend, today's meetings, and the 5 emails I actually need to reply to.

Code: a weekly automation

Use either of these as the shape for the repeat job you want handled.

Every Monday, pull my calendar, format it as a one-pager, and email it to my team.

Whenever a new lead comes in via my form, draft a personalized response.

Design: a pitch deck

Fill in the brackets and run it in Design.

Build me a 10-slide pitch deck for [your business], in [your brand colors], with [your specific sections].

Design: a social template pack

Use this once, then reuse the templates by swapping the text.

Build me 5 Instagram templates in my brand: quote post, carousel cover, carousel content, promo, testimonial.

VS Code: two starter asks

Type either of these into the Claude side panel after you open a folder.

Read these three notes and write me a one-page summary.

Rename these files so the dates come first.

Watch out

You only ever open Chat

Chat is useful and it is one quarter of what you are paying for. The tools that do work, build things, and make designs are behind different doors. Open one of the other three this week.

You assume everything is on your plan

Cowork, Code, and Design are gated by plan tier, and the tiers change. Check what your account includes before you plan a workflow around a tool you cannot open.

You use the wrong Claude for the job

Do not ask Chat to behave like Cowork. Do not ask Code to behave like Design. Chat to think. Cowork to do. Code to build. Design to create.

You ask tiny questions

These tools get better the more real material you give them. Hand over the document, the notes, the spreadsheet, and the goal, then ask for the finished thing you actually want.

You skip opening a folder

In VS Code and in Cowork, Claude only sees what you point it at. If you never open a folder, it is guessing from whatever you paste.

You point it at your only copy

Cowork and Claude Code change real files. Copy the folder first. Review what they propose before approving. There is no undo button on somebody else's machine.

You try to learn all of VS Code on day one

You need three places: the Extensions panel, the File menu, and the Claude side panel. Everything else can wait forever.

Check yourself

• [ ] I can name all four Claude tools and what each is for.

• [ ] I checked which of them my plan actually includes.

• [ ] I used Chat on one confusing document.

• [ ] I opened Cowork, or I know why my plan cannot.

• [ ] I asked Claude Code for one small thing and saw what came back.

• [ ] I made one thing in Design.

• [ ] If I set up VS Code, I opened a folder and got real work back from the side panel.

• [ ] I made a copy of any folder before letting an AI edit it.

Do this now

Pick the confusing document you have been avoiding. Open Chat, paste it in, and run the ELI5 prompt.

Keep going

Next in AI for Entrepreneurs:

• Part 8: Fast or deep: picking a model per task

• Part 9: ChatGPT and Claude, compared fairly

• Part 10: Your starter stack, and what to ignore

Every part of this course sits in reading order in the free AI for Entrepreneurs course outline.

This part is free, start to finish. Subscribing keeps the rest of the course coming and opens the paid parts.

Tell me in the comments which of the four you had never opened, and I will point you at the right next part.

So you overpay on easy work and get thin answers on hard work, and you never notice either one, because it is the same box every time. You also let the thing that built your work be the thing that checks it, which is the one job it is worst at.

What you will be able to do

• Name the three model sizes and say what each one is for.

• Switch models inside your AI app in under a minute.

• Turn a thinking mode on and off, and know when it is worth it.

• Split one real job into think, build, and check, using a different model for the check.

• Read a guide with old model names in it and still use the advice.

This is for anyone who has seen a model dropdown and never touched it. No technical knowledge needed.

The proof behind this

Liz Elliott built four software products with AI help: PRISM, AutoViral, UPRISE Social Studio, and Profounder, and she is not an engineer. Running work across different models, instead of one favorite, is how a non-technical person ships software.

What this lesson is

Lesson 2.1 of the free AI for Entrepreneurs course, in Module 2: Choosing Your AI.

A model is one brain, not one brand. Anthropic makes Claude. OpenAI makes ChatGPT. Google makes Gemini. Inside each of those brands sit several models. They are not just older and newer versions of one thing. They are different sizes, built for different jobs.

Think about a kitchen. You own a small pan and a big pot. The big pot is not better. It is bigger. You do not boil one egg in the big pot, and you do not make soup for twelve people in the small pan. You pick by the job.

Every AI company sells about the same three sizes. A small fast one for simple, high volume work. A middle one for most of your day. A big slow one for hard thinking. The names change all the time. The three sizes stay.

That is the part worth learning. Model names turn over every few months. Prices change. New ones ship. If you memorize the names, your notes go stale. If you learn the sizes, you can drop any new model into the right slot the week it comes out.

The trap is picking a favorite. When you have one favorite, you use it for everything. So you overpay on easy work and you get thin answers on hard work. People who are good at this do not have a favorite. They have a rule for which one they open.

There is one more habit that matters as much as the sizes. When a model builds something, it is bad at checking its own work. It made the same guesses on the way in and on the way out. Ask it to review itself and it will often say the work looks great. A different model reads it cold, with no attachment to how it got made. Same reason a friend catches the typo you read past ten times.

New here? Use the free AI for Entrepreneurs course outline to read this series in the best order.

Do it

Part 1: Learn the three sizes

Every big AI company sells these three. Learn this table and you can ignore the names.

Size / Plain meaning / Use it for

• Fast

• Plain meaning: The small quick one

• Use it for: Lookups, definitions, short summaries, labeling, sorting, tidying text, bulk jobs you repeat

• Default

• Plain meaning: The everyday one

• Use it for: Most of your week. Emails, reports, drafts, documents, summaries, normal coding, research that needs some thought

• Deep

• Plain meaning: The big slow one

• Use it for: Hard problems, big decisions, long research, complex plans, work where being wrong costs you real money

The names attached to those sizes when this course was written are in the reference table at the end of lesson 2.3. Treat them as examples. They will change.

One rule beats all the others: when you are not sure, use the default one. It is strong enough for thinking work and fast enough for daily use. Most people should run about eighty percent of their week there.

Part 2: Find the model picker

You switch models inside a chat. It takes a few seconds.

1. Open any chat in your AI app.

2. Look at the top of the chat. You will see the name of the model you are using right now.

3. Click that name. A list opens showing every model your plan can use.

4. Pick one. The chat uses it starting with your next message.

If the picker is not at the top of the chat in your version, look in the app's settings. App menus move. The picker itself has stayed in every version so far.

What you see in that list depends on what you pay for. Free plans usually get the fast one and the default one. Paid plans add the deep one, with tighter limits on how much of it you get. Higher paid tiers give you more of the deep one. If a model you read about is missing from your list, that is why.

Part 3: Turn thinking on when the job needs it

Most AI apps also have a thinking mode. It is a mode, not a model. You switch it on top of whatever model you picked.

With it on, the model works through the problem step by step before it answers. You can usually watch that thinking happen. The answer takes longer and it catches things a fast answer misses.

To turn it on:

1. Pick the default or the deep model first.

2. Look at the lower left of the box you type in.

3. Click Search and tools.

4. Turn on Extended thinking.

Turning it on may start a new chat, so finish what you are doing first. If your app words it differently, look for anything called thinking, reasoning, or think longer. If you cannot find a toggle at all, you can ask for the same behavior in words. Add this to your prompt: Think through this carefully before responding.

Use it when you have gone back and forth on a decision, when a problem has several competing pieces, or when you want to see the reasoning and not just the answer. Skip it for emails, summaries, and quick questions.

Turn it off when you are done. Leaving it on burns your usage limit on simple work for no gain.

Part 4: Run the three-model workflow

This is the habit that separates strong AI users from everyone else. One job, three steps, three different models.

1. Think. Open your deep model. Ask it to plan the approach, break the problem into parts, and decide how the work should happen. Do not let it start building. You want the shape of the job first. Speed does not matter here.

2. Execute. Switch to a fast strong model. The hard thinking is done. This step is writing, building, formatting, turning the plan into a real draft. You do not need your deepest thinker to follow a plan that already exists.

3. Review. Copy the finished work into a chat with a different model. Not the one that built it. Paste the goal, paste the work, and run the review prompt below.

Moving work between models is plain copy and paste. Copy the plan out of chat one, paste it into chat two. Nothing clever.

Step 3 is the one most people skip, and it is the one that saves you. A fresh model has different blind spots. It notices the gaps the first one talked itself past.

Part 5: Stop wasting your limit

These five habits work on every model and every plan.

1. Edit your last message instead of sending "no, try again." Hover over your last message, click the edit pencil, and change the prompt. Editing forks the chat, so the bad answer does not stay in the conversation as extra baggage.

2. Batch tasks into one message. Three tasks in one message costs less than three separate messages. Use a numbered list.

3. End simple prompts with "No commentary. Just the output." It cuts the intro, the summary, and the offer to help further. On short tasks that can cut the answer length by a third or more.

4. Start a new chat for a new topic. A long chat carries all the old context forward, and the app reprocesses it every message. New topic, new chat.

5. Do not upload a long PDF when the text will do. PDFs carry layout data that costs more to read. For long or dense files, paste the plain text into the chat instead.

Part 6: Write your own rule

Run the model audit prompt below in your default model. It maps your actual work to the sizes and hands you back a short personal rule. Save that rule somewhere you can see it. Three or four sentences is enough.

Copy and paste

The review prompt

Use this after one model builds something. Paste it into a chat with a different model.

Review this work cold. Look for mistakes, missing pieces, weak assumptions, confusing parts, and places where it does not match the goal. Be direct and specific. Do not rewrite the whole thing yet. First, tell me what is broken and what should change.

Goal:
[PASTE THE GOAL]

Work to review:
[PASTE THE WORK]

The model audit prompt

Use this once, to build your own rule for which model you open.

I want to figure out how to use AI models more strategically for my specific work.
What I do:
[your job title and 2-3 sentences about what my day-to-day work actually involves]
How I currently use AI:
[what I use it for most, and which model I default to]
My biggest bottleneck:
[what's taking the most time or where the output has felt shallow]
Based on this, I want you to:
1. Audit my current model usage.
Given how I described my work, am I using the right model for the right tasks? Where am I likely wasting the deep model on things the default model handles just as well? Where am I using the default model on things that would genuinely benefit from deeper reasoning?
2. Map my specific tasks to the right model.
Take the types of work I described and tell me, for each major task category, whether that's a fast-model task, a default-model task, a deep-model task, or somewhere I should turn on extended thinking. Be specific to my role, not generic.
3. Tell me the 3 tasks in my role where extended thinking would make the biggest difference.
Not general advice. Specific to what I described. These should be tasks where seeing the reasoning step by step would genuinely change the quality of the output or help me think something through more clearly.
4. Give me a simple personal model policy.
Write me a 3-4 sentence rule of thumb I can actually remember, tailored to my specific role and work style, for when to use the fast, default, and deep models, and when to turn on extended thinking.
Be specific to my situation. The point of this exercise is to stop using models on autopilot.

The thinking line

Add this to a prompt when you cannot find a thinking toggle, or when the task involves logic or numbers.

Think through this carefully before responding.

The output trimmer

Stick this at the end of any simple request to cut the padding.

No commentary. Just the output.

The batching pattern

Use this shape when you have several small asks about the same thing.

1. Rewrite this email.
2. Make it shorter.
3. Give me 3 subject lines.

The shape to remember

Save this line. It survives every model launch.

smartest to think, fastest to build, a different one to review, cheapest for the boring stuff

Watch out

You use the biggest model for everything

The deep model on a short email works fine. It is just an expensive way to write a short email, and it eats the part of your limit you wanted for the hard job later in the week. Default for most things. Deep when the work earns it.

You let the builder check its own work

It made the same assumptions twice, so it will tell you the work looks great. Hand the output to a different model. This matters most on anything you are going to send, publish, or make a decision from.

You memorize model names

Guides go stale within months. The version numbers you read this year will be wrong next year. Learn fast, default, and deep. When a new model ships, ask which of the three slots it fits and carry on.

You hand a long document to the fast little model

Small fast models lose track of detail across a long file. They will not tell you. They will quietly miss things in the middle. Long documents go to the default or the deep model.

You blame the model when your brief was thin

A lot of weak output comes from a weak ask. Before you switch models, give the one you have more context. Say what you want, who it is for, and what a good version looks like. Then try again.

You leave thinking mode on

It is a switch, not a setting you turn on once. Left on, it spends extra effort on simple work. Turn it off when the hard task is done.

Check yourself

• [ ] I can name the three model sizes without looking.

• [ ] I found the model picker in my app and switched models once.

• [ ] I turned extended thinking on, used it, and turned it off.

• [ ] I ran one real job through think, build, and review with three chats.

• [ ] The review used a different model than the build.

• [ ] I ran the model audit prompt and saved the rule it gave me.

• [ ] I know which models my plan actually includes.

Do this now

Take the last real piece of work an AI made for you. Open a chat with a different model, paste the review prompt, fill in the goal and the work, and read what it finds.

Keep going

Next in AI for Entrepreneurs:

• Part 7: The four Claudes, and which one you actually want

• Part 8: Fast or deep: picking a model per task

• Part 9: ChatGPT and Claude, compared fairly

Every part of this course sits in reading order in the free AI for Entrepreneurs course outline.

This part is free, start to finish. Subscribing keeps the rest of the course coming and opens the paid parts.

Tell me in the comments which model you have been using for everything, and I will point you at the right next part.

Four things stop people from using AI, and none of them are about how the tool works. Guilt about energy. A foggy feeling about the future. Attention you cannot spare. Predictions you cannot rate. Left alone, they make the decision for you, quietly, on bad information.

What you will be able to do

• Quote the actual numbers on data center energy use and on one AI question.

• Tell the difference between the cost of training a model and the cost of using one.

• Name what social media was doing for you, and which of those jobs AI now does.

• Repeat the main AI predictions accurately, including who made them and what they rest on.

• Run any scary AI headline through a two question filter before it decides your week.

This is for anyone who wants a straight answer before they commit. You will finish able to hold your own opinion with the real figures behind it.

The proof behind this

Liz Elliott teaches beginners and non-technical people, and she runs 33 active social brands on AI automations across several computers. She rebuilt a 403-guide archive into this 103-lesson course, so the numbers here are sourced and marked, not remembered.

What this lesson is

Lesson 1.5 of the free AI for Entrepreneurs course, in Module 1: What AI Actually Is.

Four things stop people from using AI, and none of them are about how the tool works.

The first is guilt about energy. You have seen the posts about data centers boiling the planet, and you do not want to be part of it.

The second is a foggy, tired feeling. You cannot see what your job looks like in five years, so every long plan feels shaky.

The third is attention. You already lose hours to a screen. Adding another thing to open sounds like a bad trade.

The fourth is the predictions. Smart people are saying enormous things about the next few years, and you cannot tell which ones to take seriously.

Every one of those is a fair thing to feel. They also all have real information behind them, and most of what circulates online is a mangled version of it.

Think about the last time you looked up whether some food was bad for you. Half the internet screamed yes, half screamed no, and the actual study said something narrower and duller than either side. The numbers existed the whole time. They just travelled slower than the noise.

Same story here. So this lesson gives you the numbers, names who said what, and marks clearly where a claim stops being data and starts being a guess about the future.

Nobody is going to tell you what to conclude. You are going to end this lesson able to hold your own opinion with the real figures behind it.

New here? Use the free AI for Entrepreneurs course outline to read this series in the best order.

Do it

Part 1: The energy question, with the actual numbers

The International Energy Agency published a report called Energy and AI in April 2025. These are its main figures.

Data point / Number

• World electricity used by data centers in 2024: 1.5 percent, or 415 TWh

• Projected data center share of world electricity by 2030: About 3 percent, or about 945 TWh

• Growth in AI-focused data centers in 2025: 50 percent

• Growth in data centers overall in 2025: 17 percent

So yes, data centers use real energy, and yes, that number is growing.

Now the other end of the scale. Epoch AI, an independent research group, found that a typical AI question uses about 0.3 watt-hours. A 1,000 watt microwave running for one second uses about 0.28 watt-hours. That is where the "one AI question is about a second of microwave" comparison comes from, and it is roughly right. Hannah Ritchie at Our World in Data helped make the comparison well known.

Here is what a heavy personal day looks like:

100 questions times 0.3 Wh is 30 Wh. An average US home draws about 1.2 kW. That works out to about 90 seconds of normal household electricity use.

Older viral graphics used 3 watt-hours per question, ten times the current credible estimate. The IEA report rejected those inflated numbers. If a chart you are shown uses 3 Wh, it is not using the current figure.

The last piece is where the growth actually comes from. The IEA points at GPU-heavy systems, called accelerated servers, which are used to train and run large models. Those account for about half of the net increase in data center energy through 2030. A company training a brand new frontier model for three months uses vastly more energy than one person asking questions over the same period.

That is the split worth keeping. Training a new model is a company-level decision about whether to build it. Asking questions of a model that already exists is a different question with a much smaller number attached. Both facts are true at once. What you do with them is yours.

Part 2: The foggy feeling has a name

Harvard Business Review published a piece by Toby E. Stuart, the Helzel Chair at UC Berkeley Haas, called The Future Is Shrouded in an AI Fog.

His point: this is reduced visibility, not darkness. You can still see. You just cannot see far. That makes long bets feel harder. A four year degree. A thirty year mortgage. One career path for decades. A clean five year plan.

The move is not to guess the future correctly. You will not know which jobs change first, which tools win, or what five years from now looks like. The move is to build more options, so more doors are open whichever way it goes.

Five things that hold their value no matter which way it breaks:

1. Put your context in one place. Set up one Project for the thing you do most. Add your context, your voice, and your reference files once. Then stop retyping your setup into every chat.

2. Build trust. A small group of people who trust you is a strong asset. Show up often, talk like a person, help before you ask for anything, be clear and honest. Trust is slow, and it is hard for AI to replace. Two hundred people who open your emails beat two hundred thousand who scroll past.

3. Learn on purpose. The real flexibility is how fast you can learn a new area when your old one changes. Pick one new thing, learn a little each week, use it in public, and teach it back in plain words. Teaching is what makes it stick.

4. Make one public thing. Something small with a real address on the internet. A page, a tool, a guide. Being the person who makes real things holds up regardless of which tool wins.

5. Start one small income stream that uses AI. A weekend product, a freelance service, a small digital product. The money is not the point at first. Having a second bet running changes how the first one feels.

Pick one. Not five. Small work done often beats a big plan you never start.

Part 3: Where your attention actually goes

There is real data here, and it is more interesting than the usual screen-time lecture.

An analysis by GWI, covering more than 250,000 adults across over 50 countries, was shared by John Burn-Murdoch of the Financial Times. It found social media use peaked in 2022 and has fallen every year since. In developed markets, average daily social time is down almost 10 percent, the first steady drop in ten years.

Around the same period, Appfigures reported that ChatGPT became the most-downloaded app in the world, beating Instagram and TikTok for the first time. Reported user numbers went from about 100 million monthly users in early 2023 to around 900 million weekly users by early 2026.

The strategist Matt Paige put an explanation on top of that data. He argues people used social media for five jobs, and AI now does four of them better. Treat the data as data and his explanation as an argument, because a decline in one thing does not prove another thing caused it.

The five jobs, and where they now go:

What you opened social media for / What that looks like now

• Information: A short question beats 30 minutes of scrolling for "what happened this week"

• Entertainment: AI does not replace funny videos, but it does replace the endless "show me something" scroll

• Distraction: Ask a curious question and follow the answer for ten minutes

• A quick hit of novelty: You can ask for something surprising instead of waiting for a feed to surprise you

• Feeling connected: Still social media. If your nephew takes his first steps, you see it where his mom posts it

Try this yourself. Notice which of the five you are reaching for next time you open an app out of habit. Four of them have a swap. One of them does not, and that one is worth protecting.

If you want to put the reclaimed time somewhere, these are the ideas people build first, with rough build times attached:

• A personal newsstand that pulls from sources you choose and writes one weekly digest. About 2 hours.

• A habit tracker that nudges you and writes a weekly review. About 3 hours.

• An inbox triage setup that finds the five emails that matter and drafts replies. About 1 hour.

• An idea validator that researches a market and competitors before you commit. About 2 hours.

• A weekly planner that reads your calendar and your last seven days and tells you what to focus on. About 2 hours.

Those are ideas, not instructions. The build steps come later in this course.

Part 4: The predictions, marked as predictions

Everything in this part is a forecast. None of it has happened.

The 2028 forecast. Jack Clark, an Anthropic co-founder and head of Policy at the company that makes Claude, wrote in his Import AI newsletter, issue 455, that he puts a 60 percent or higher chance on AI doing AI research and development with no human involved by the end of 2028. He gives a similar 60 percent chance to recursive self-improvement by the same date, which in plain words means AI helping build its own successor.

What that number rests on is worth knowing. Researchers track how long an AI agent can work on its own before something breaks, which they call time-horizon length. An independent evaluation lab, METR, has measured this over time and found it roughly doubling every four months. If that doubling slows, the date moves further out. If it speeds up, it moves closer. So the forecast is a bet on a trend line continuing, not a schedule.

His suggested response is steady, not dramatic: learn to build one small agent, learn to judge AI output rather than just produce it, learn one set of connected tools deeply, ship one small thing other people can click, and understand the rules side, including model evaluations and laws like the EU AI Act.

Three softer predictions. These circulate widely and have no data behind them at all. Hold them loosely.

• Talking will take over from typing, and dictating will become the normal way people work with AI.

• More than half of professionals will have at least one agent working for them daily.

• Companies past about 20 people will hire someone whose job is running their AI systems.

One more, from the hardware side. Jensen Huang, CEO of NVIDIA, gave a Stanford lecture called The Compute Behind Intelligence. He said computing got about a million times faster over the last decade, which is part of why AI felt like it arrived everywhere at once. He also said the energy computing will need is far beyond today's supply, on the order of a thousand times more. Both figures come from the head of a company that sells AI chips, so weigh them accordingly.

Part 5: A filter for scary claims

Huang pushed back on AI doom claims in that lecture, saying some of them "are all being made up." His filter is a useful one, and it cuts in every direction, including at him.

Ask whether the claim could be proven wrong. If no evidence could ever prove it false, it is a fear story, not a checked fact. That test applies just as much to a chip executive saying the risks are invented as it does to a headline saying everything is about to end.

Then do the thing that actually resolves it. You learn more from using the tool on one real task than from reading one more panic post, and more from checking one real number than from arguing about a screenshot.

Copy and paste

The energy math

Keep this so you can check any energy claim you get sent.

100 questions x 0.3 Wh = 30 Wh total
Average US home draw: about 1.2 kW
30 Wh / 1.2 kW = about 90 seconds of normal household use

The fear filter

Run any alarming AI headline through these four questions before it changes what you do.

Can this claim be checked?
What evidence would prove it wrong?
Is this a fact, or is it a fear story?
What is one small task I can try with AI today?

The direction questions

Ask these when you are trying to decide what to learn next.

Where is this field clearly heading?
What skill will people need in two years?
What tool will everyone wish they understood later?
What problem is still small now but growing fast?

The control questions

Ask these when a headline leaves you feeling stuck.

What can I control here?
What can I make clearer?
What skill can I build?
What proof can I create?
What would make me hard to ignore?

Watch out

Comparing your questions to the cost of training a model

These are two different numbers and they get mixed together constantly. Training a new frontier model is a company decision measured in months of heavy compute. Your questions are measured in fractions of a watt-hour.

Passing on a graphic without checking the figure

Old charts using 3 watt-hours per question are still circulating. The current credible estimate is about 0.3. Check which number a chart is built on before you share it.

Treating a correlation as a cause

Social media time is falling, and AI use is rising. That is two facts. The explanation that one caused the other is an argument someone made, and it may be right, but the data does not prove it.

Repeating a prediction as if it happened

The 2028 forecast is a probability someone assigned to a trend continuing. The three softer predictions have no data behind them at all. Say "he predicts" out loud when you repeat them.

Taking a number from someone who sells the thing

Chip company figures on how fast computing got and how much energy it will need are real numbers with a commercial interest attached. So is a model company's forecast about models. Note the interest, keep the number, and go find a second source.

Trying to predict your way out of the fog

You will not guess which jobs change first. Building options works whether or not you guessed right. Predicting only works if you guessed right.

Building five things at once because five sounded doable

Every list in this lesson has five items. Pick one from one list. Small work done often is what actually moves.

Swapping the one job AI does not do

Four of the five reasons people open social media have a decent AI swap. Feeling connected to actual people is not one of them. Keep that one where it lives.

Check yourself

• [ ] I can say what share of world electricity data centers used in 2024, and the 2030 projection.

• [ ] I know roughly what one AI question costs in watt-hours, and what it compares to.

• [ ] I can explain the difference between training a model and using one.

• [ ] I can name which of the five social media jobs AI does not replace.

• [ ] I can state the 2028 forecast, who made it, and what trend it rests on.

• [ ] I marked the three softer predictions as unsourced guesses in my own head.

• [ ] I picked one of the five career options to build, and only one.

• [ ] I ran one claim through the fear filter.

Do this now

Write down the one AI claim that has been sitting in the back of your head, the one that makes you hesitate. Run it through the four fear filter questions and see whether any evidence could prove it wrong.

Keep going

Next in AI for Entrepreneurs:

• Part 6: Models explained, and why one favorite is a trap

• Part 7: The four Claudes, and which one you actually want

• Part 8: Fast or deep: picking a model per task

Every part of this course sits in reading order in the free AI for Entrepreneurs course outline.

This part is free, start to finish. Subscribing keeps the rest of the course coming and opens the paid parts.

Tell me in the comments which AI headline made you nervous this month, and I will point you at the right next part.

Three things went wrong that day and none of them were the tool. No context, the wrong first task, and an old product. People run that same bad first week once and treat it as a verdict for years.

What you will be able to do

• Explain why a short request gives you a plain answer, and fix it in one line.

• Set up a free account and run a first task that is not "write me an email".

• Write a prompt with all five parts: role, task, format, examples, and limits.

• Build a Project that holds your role, your voice, your priorities, and your rules.

• Follow a seven day plan that adds one thing a day and ends with a saved setup.

This is for anyone starting from scratch, or restarting after a bad first try. Free accounts are enough for the whole week.

The proof behind this

Liz Elliott built four software products with AI help: PRISM, AutoViral, UPRISE Social Studio, and Profounder. She is not an engineer and never studied computer science. Everything she runs started the way this week starts.

What this lesson is

Lesson 1.4 of the free AI for Entrepreneurs course, in Module 1: What AI Actually Is.

Plenty of people try AI once, get a bland answer, and quietly decide it is overhyped. Almost always, three things went wrong, and none of them were the tool.

The first is context. If you type "write me an email," the AI has no idea who you are, how you write, who the email is for, or what you want to say. It fills the gaps with the most average version of everything. Asking a stranger to write your email would go the same way.

The second is the first task. Most people start by asking AI to write something from nothing. That is one of the harder jobs to do well. It is much stronger at analyzing data, summarizing long documents, comparing options, finding patterns, organizing a mess, brainstorming, and research.

The third is the tool. People judge everything they hear about AI by one old experience with one old product.

Here is the frame that fixes all three. Treat AI like a very smart colleague who started at your company this morning. Sharp, fast, and completely ignorant of your job, your style, your customers, and your goals. You would not hand that person a one-line instruction and expect a finished piece of work. You would tell them the situation first.

More context in, better work out. Less context in, generic work out. That is the whole trick, and it does not change as the tools change.

Stanford's AI Index report found that generative AI reached 53 percent of the world's population within three years. So you are not late. Most people are still learning to hold this thing.

From Liz's playbook: Liz Elliott is not an engineer and did not study computer science. She built four software products with AI help: PRISM, AutoViral, UPRISE Social Studio, and Profounder.

New here? Use the free AI for Entrepreneurs course outline to read this series in the best order.

Do it

Part 1: Pick your starter tools, and add them one at a time

Four tools cover almost everything a beginner needs. Do not learn them all at once.

Tool / Use it for / Where to go

• Claude

• Use it for: Thinking, writing, planning, analysis, brainstorming, organizing

• Where to go: claude.ai

• Perplexity

• Use it for: Research and current web answers with sources

• Where to go: perplexity.ai

• NotebookLM

• Use it for: Asking questions about documents you upload

• Where to go: notebooklm.google.com

• Wispr Flow

• Use it for: Talking instead of typing

• Where to go: Search for Wispr Flow and use the official site

The order matters. Use Claude on its own for a week. Add Perplexity when you next need to look something up. Add NotebookLM when you have a long document you have been avoiding. Add Wispr Flow when you get tired of typing.

Claude has a free plan that covers a lot. The paid plan has been around $20 a month and adds more usage plus extra features. Check the current price and what your plan includes before you pay for anything.

Part 2: Learn the five parts of a prompt

Every strong prompt has five parts. Miss one and you can usually feel it in the answer.

Part / What it means / Example

• Role

• What it means: Tell it who to be

• Example: "You're a bookkeeper with 15 years of experience."

• Task

• What it means: Give one clear job

• Example: "Draft 5 subject lines for a late invoice reminder."

• Format

• What it means: Say how to lay it out

• Example: "Use a table."

• Examples

• What it means: Show what good looks like

• Example: Paste 2 or 3 samples.

• Constraints

• What it means: Say what not to do

• Example: "No buzzwords. Under 50 words."

Then there are two habits that matter as much as the five parts.

Paste in real content. It is strongest when it has something real to work with. Meeting transcripts for action items. Job postings for resume fit. Contracts for a plain English explanation. Sales numbers for patterns. A rough draft for a rewrite. Working on real material beats generating from nothing.

Have a conversation. Your first prompt does not need to be perfect. Ask, read, then adjust. "That's too formal, make it sound like how I actually talk." "Good, but add specific numbers." "Cut this in half." Going back and forth three to five times is normal, not a sign you did it wrong.

And one use most people miss. Use it to think, not only to write. Ask it to poke holes in your plan, argue the other side, list ten things that could go wrong, or compare two options side by side.

Part 3: Your first week, one thing a day

Thirty to sixty minutes a day. One focus per day. If a day is taking 90 minutes, you are going too deep, so finish the action and move on.

Day 1. Sign up and ask something real. Go to claude.ai and make a free account. There are desktop apps for Mac and Windows and a phone app too. Then ask about something you genuinely want to understand or decide. Do not start with "write me an email."

Day 2. Use it on work you already had to do. Not a demo. A real thing on your list. Prepping for a meeting, summarizing a document, planning your week, drafting a reply. Give it the context first.

Day 3. Rewrite one prompt properly. Take a prompt you used on day 1 or 2. Rewrite it with all five parts. Run both versions. Save the before and after so you can see what the difference bought you.

Day 4. Build your first Project. A Project is a workspace that keeps your context across conversations. Steps are in Part 4.

Day 5. Connect one or two tools. Start with your email and wherever your documents live. Steps are in Part 4. Then run one real prompt against each one so you know it works.

Day 6. Try a research tool and a document tool. Next time you would normally search the web, use Perplexity instead and ask in full sentences. Then take a long PDF or report you have been putting off, upload it to NotebookLM, ask for a summary, and ask it three specific questions.

Day 7. Save your profile and write your routine. Put your details into settings so every future chat starts with them, then write down your daily routine and block it on your calendar.

Here is the daily routine to aim for once the week is done:

Routine / Time / What to do

• Morning brief

• Time: 5 minutes

• What to do: Ask for your brief across mail, calendar, and messages.

• Hand off one task

• Time: 15 to 30 minutes

• What to do: Give it one task you would normally do yourself.

• End of day debrief

• Time: 5 minutes

• What to do: Say what shipped, what is open, ask what to start tomorrow.

• Weekly review

• Time: 15 minutes on Friday

• What to do: Pull wins, spot patterns, draft your update.

Part 4: The three layers, in order

Do not build all three in one afternoon. Build one, use it, then add the next.

Layer 1. Projects. This is the one that stops you re-explaining yourself.

1. Go to claude.ai.

2. Click Projects in the left sidebar.

3. Click New Project.

4. Name it for how you think about your work. My Work, Client Work, Content, Operations.

5. Write 5 to 10 sentences of Project instructions. Cover your role, your communication style, what the Project is for, and your standing rules.

6. Upload real material: past work you are proud of, templates, brand notes, anything it needs for context.

7. Add 2 or 3 emails or documents you actually wrote, so it can match your voice.

Every conversation inside that Project picks up the instructions and files automatically.

Layer 2. Connectors. Without them, it only knows what you paste in. With them, it can read from tools you already use, like your mail, your files, your calendar, and your notes.

1. Go to claude.ai.

2. Open Settings and find the section for connected apps.

3. Connect the tools you use most, starting with email and wherever your documents live.

4. Sign in and grant permission.

5. Run one real prompt against each one.

There are hundreds of connectors available. You need two.

Layer 3. Cowork. Chat is a conversation. Cowork is a handoff. In chat it replies and waits. In Cowork it makes a plan, works with real files in a folder, and can run tasks on a schedule.

1. Download or update the Claude desktop app from claude.com/download. It runs on Mac and on Windows.

2. Open the app and click the Cowork tab.

3. Click Work in a folder.

4. Give it one focused project folder. Not your whole drive.

5. Describe the outcome: what file should exist when it is done, what goes in it, what format.

6. Review the plan before it runs.

7. Approve the plan.

8. For work that repeats, save it as a scheduled task.

Two things to know before you count on this. Cowork needs a paid plan. And scheduled tasks only run while the desktop app is open, so a task set for Monday 8am does nothing if your laptop is shut.

Part 5: When the menus do not match

Two things in this lesson will not match your screen exactly, and that is normal.

Menu names move. One version says Settings then Profile. Another says Settings then Memory. Another says Settings then Integrations, or Settings then Connectors. They are all the same few screens with different labels. The rule: open Settings and look for the section that matches the job. Connected apps for connectors. Instructions or profile or memory for telling it about you. Do not assume a feature is missing because a word changed.

Free and paid keeps shifting. Sources disagree about which features are free. Chat is free. Projects and connectors are usually available on the free plan, though limits vary. Cowork and scheduled tasks need a paid plan. The rule: check your own account screen before you believe any article, including this one.

Part 6: The question to ask before every request

Before you hit enter, ask yourself one thing. Am I describing a task, or am I describing an outcome?

If your prompt does not say what finished looks like, it has to guess. So say the file type, the length, where it should go, the format, and what goes in it. "Create a 1-page summary as a Word doc in /outputs" is a finished description. "Summarize this" is not.

Copy and paste

Role line

Put a line like this at the top of a prompt to give it a point of view.

You are a bookkeeper with 15 years of experience helping small shops keep clean books.

Weak prompt versus strong prompt

Run both. The gap is the whole lesson.

Write me an email.

I'm a project manager at a tech startup.
I need to email my team about a deadline change.
The original deadline was Friday.
The new deadline is next Wednesday.
The team is stressed and I want the tone to be calm and reassuring.
I usually write in a direct but friendly way.
Write the email.

Constraints block

Paste this at the end of any writing request to kill the corporate voice.

Don't use corporate buzzwords.
Don't start with "I hope this email finds you well."
Don't use bullet points.
Keep it under 150 words.

Day 1 starter prompts

Use either one for your very first conversation.

Explain how [topic] works.

Help me think through [decision].

Learning coach instructions

Paste this into a Project called AI Fluency on day 1, so the whole week builds on itself.

You're my AI fluency coach for the next 7 days. Default to plain English. Reference what we've covered so far when you teach me new things.

Project instructions, short version

Paste this into a new Project and fill in the brackets. Good for day 4.

You're my daily co-pilot. I'm a [role] at [company/business]. My main goals this quarter are [goals]. My voice is direct, no buzzwords. When you draft anything for me, mirror my voice from the writing samples I've uploaded.

Project instructions, full version

A worked example of the 5 to 10 sentences. Swap the details for yours.

I run a small bookkeeping business with two employees.
My communication style is direct, with no filler and no corporate speak.
This Project is for client-facing work.
Always use formal language in client communications.
When writing emails, lead with the key point.
Never include more than 3 action items in a single message.
When you're unsure about tone, ask before drafting.

Work rules to add to any Project

Small lines that change every answer. Add your own.

I prefer bullet points over paragraphs.
Never use the word "synergy."
Keep things under 200 words unless I ask for more.

Connector test prompts

Run one of these the moment a connector is switched on, to prove it works.

Pull my last 3 emails from Sarah.
What's on my calendar tomorrow?
Search Notion for the Q3 strategy doc.

Connected briefing

Use this once two or three tools are connected and you want a real status check.

Check my Gmail, read the Notion board for Project Phoenix, and give me a 5-bullet briefing on where things stand, what needs a response today, and whether we're on track for the Friday deadline.

Morning brief

Your five minute start to the day.

Run my morning brief. Pull from Gmail and Slack and Calendar. Surface the 3 most important things, draft replies for the routine ones, and prep me for today's meetings.

End of day debrief

Your five minute close.

Here's what I shipped today: (list). Here's what's open: (list). What should I tackle first tomorrow?

The question before you prompt

Ask yourself this before you hit enter.

Am I describing a task, or am I describing an outcome?

Defining done

Add a line like this whenever you want a file, not a chat reply.

Create a 1-page summary as a Word doc in /outputs.

Turning a repeat task into a Skill

Paste your written-out task instructions, then this line.

Turn this into a Skill I can trigger with one sentence.

A scheduled task

Save something like this once the desktop app and Cowork are set up.

Every Monday at 8am:
Read the project files in /active-projects and my Notion boards.
Check Gmail for anything from clients over the weekend.
Create a weekly priorities doc in /briefings with the top 5 things I need to move this week.
Flag any client messages that need a reply today at the top.

Five thinking prompts

Use these when you want help thinking, not writing. Read the caution in Watch out before pasting contracts or offers.

Here's what I'm thinking about doing. Give me the 5 strongest reasons to do it and the 5 strongest reasons not to.

Based on this, what are they doing better than us and where are they vulnerable?

Here are 6 months of sales data. What trends do you notice that I should be paying attention to?

I need to tell my team that we're cutting a project they care about. Help me think through the best way to frame this, anticipate their objections, and prepare responses.

Tell me what I should be paying attention to that I might be missing.

Four research questions for day 6

Ask these in full sentences, the way you would ask a person.

What is [Company X]'s pricing model and how does it compare to [Company Y]?

Notion vs. Monday.com vs. Asana - which is best for a 10-person marketing team?

Explain the new DOL overtime rule changes for 2026 and how they affect salaried employees

Is it better to max out my 401k or pay off my student loans first? I make $85k and owe $40k at 5.5%

Five things to build in one chat

Ask for any of these and you get a working tool you can click, not a description of one.

Build me a monthly budget tracker where I input my income and expenses by category and it shows me where my money is going with a visual breakdown.

Build me a project status dashboard where I can list my active projects, their deadlines, their status, and see which ones are at risk.

Build me a weekly meal planner where I can drag meals between days, add grocery items, and generate a shopping list.

Build me a client pipeline tracker with stages (lead, proposal, negotiation, closed) where I can move clients between stages and see my total pipeline value.

Build me a savings goal calculator where I enter my target amount, my current savings, and my monthly contribution - and it shows me exactly when I'll hit my goal with a countdown.

Fixing what it built

If the first version is wrong, say so in plain words. It updates live.

Change the colors.

Add a column for notes.

The full setup prompt

The big one. Paste this when you are ready to have your whole setup written for you in one go. It is long on purpose.

I want to set up Claude properly for vibe working, delegating outcomes, not managing tasks.
Help me build the full setup for my role.

My role:
[job title and 2-3 sentences on what you actually do day to day]

My industry:
[e.g. marketing, finance, operations, healthcare, legal]

Tools I use most:
[e.g. Gmail, Notion, Slack, Google Drive, Salesforce]

What eats most of my time right now:
[2-3 things that take too long or happen too often]

Step 1: Write my Project instructions.
Based on my role and context, write a complete Project instruction set I can paste into a new Claude Project right now.
It should cover: who I am and what I do, my communication style and tone, standing rules Claude should always follow in this Project, what "good output" looks like for my work, and anything Claude should always ask before doing.

Step 2: Tell me which connectors to set up first.
Given the tools I listed, which 2-3 connectors should I set up this week for maximum impact?
For each, give me one specific prompt I could run the moment it's connected that would immediately save me time.

Step 3: Write me my first Cowork task.
Based on what's eating most of my time, write a complete Cowork task brief I can paste in right now.
Include the specific outcome, file format, what to do with edge cases, and a request to show me the plan before executing.

Step 4: Give me my first outcome-delegation prompt.
Write an example of what a well-structured outcome-delegation prompt looks like for one of the tasks I described.
It should give Claude full context, a clear outcome, and everything it needs to complete the work without coming back to ask me questions.

Make everything specific to my role.
The goal is to walk away from this with a setup I can use today.

Watch out

Starting with "write me an email"

It is the worst possible first task. No context, and generating from nothing is the hardest thing to do well. Start with something to think about or something to analyze.

Judging AI by one bad experience

Tools move fast and models change every few months. If you tried something once and it was flat, the fix is usually more context, not a better product.

Trying to build all three layers in one afternoon

Projects, connectors, and Cowork in the same sitting is how people burn out and use none of them. One Project this week. Use it. Then add the next layer.

Setting Project instructions but uploading nothing

Instructions help. Real examples help more. Upload three to five pieces of work you are proud of, like templates, past drafts, or client briefs.

Stopping after one answer you did not like

Sending one prompt, disliking the first reply, and giving up is the most common way people decide AI is useless. Push back, adjust, and go around three to five times.

Connecting your inbox with no thought about privacy

Once your mail, calendar, and files are connected, real client details, contracts, and personal messages become readable. Check your employer's policy first. Connect the account you are willing to expose, and know that you can disconnect it again in the same settings screen.

Giving Cowork your whole drive

Cowork reads and writes real files on your computer. Give it one focused project folder, and keep a backup of anything you cannot afford to lose.

Assuming a scheduled task ran

Scheduled tasks only run while the desktop app is open. Check the output before you rely on it.

Uploading contracts, offers, and financial details without thinking

Some of the prompts here invite you to paste in a contract, a job offer, or your loan numbers. That is real personal information going to a company's servers. Decide case by case, strip out names and account numbers when you can, and remember that AI is not a lawyer or a licensed financial advisor. Use the answer to prepare your questions, not to make the decision.

Skipping the profile step

The last step of the week is the one that makes every future chat better. Do not stop at day 6.

Check yourself

• [ ] I created a free account and asked something real, not "write me an email".

• [ ] I used AI on one task I actually had to do this week.

• [ ] I rewrote one prompt with all five parts and ran both versions.

• [ ] I built one Project with my role, my voice samples, my priorities, and my rules.

• [ ] I connected at least one tool and tested it with a real prompt.

• [ ] I used a research tool instead of a search engine at least once.

• [ ] I uploaded one long document and asked it three questions.

• [ ] I saved my details in settings so new chats start with context.

• [ ] I wrote my daily routine down and blocked it on my calendar.

Do this now

Go to claude.ai, make a free account, and ask it to help you think through one real decision you are sitting on. Give it the situation before you ask for the answer.

Keep going

Next in AI for Entrepreneurs:

• Part 5: The honest picture: energy, tiredness, attention, and where this goes

• Part 6: Models explained, and why one favorite is a trap

• Part 7: The four Claudes, and which one you actually want

Every part of this course sits in reading order in the free AI for Entrepreneurs course outline.

This part is free, start to finish. Subscribing keeps the rest of the course coming and opens the paid parts.

Tell me in the comments which day of the seven you are on, and I will point you at the right next part.

That is level one, and it works. That is the problem. It works well enough that people stay there for years while others, with the same tools and the same hours, get far more out of it. The gap is not talent and it is not smarter sentences. It is setup.

What you will be able to do

• Place yourself on a five level ladder without guessing high.

• Name the single next move for your level, instead of trying to do everything.

• Spot the five habits that keep people stuck on level one.

• Sort your last 20 AI sessions into three buckets and see what your mix is missing.

• Run five ready-made prompts that each fix one stuck habit.

This is for anyone who uses AI most days and suspects they are only scratching at it. No tech background needed. You place yourself, then you take one step.

The proof behind this

Liz Elliott runs 33 active social brands and runs her whole operation on AI automations across several computers. That is the top of this ladder, and it was built one process at a time by someone who never studied computer science.

What this lesson is

Lesson 1.3 of the free AI for Entrepreneurs course, in Module 1: What AI Actually Is.

Most people who use AI use it the same way. Open a blank box, paste something in, ask a question, close the window. Next time, start from zero again.

That works. It is also the first rung of a ladder, and most people never find out there are four more.

Think about someone learning to cook. Week one you follow one recipe at a time and check the page after every step. Later you have five meals you can make without looking. Later still you keep a stocked kitchen, so dinner starts with what is already in the fridge instead of a trip to the shop. Same person, same hours, very different results. Setup did that, not extra talent.

AI works like that. The people getting the most out of it are not typing smarter sentences. They built a setup that already knows their job, their voice, their team, and their tools. So a short request like "write the Monday update" produces a finished report, because everything it needs is already loaded.

This lesson does three things. First, it shows you the five levels so you can find your real one. Read it honestly, because most people rank themselves a rung too high. Second, it gives you the one move that lifts you to the next level, and only that move. Third, it gives you a way to check the balance of your AI time, so you neither grind the same three prompts forever nor chase shiny new ideas that never turn into anything.

You do not need to be on level five. You need to know your rung and take one step.

From Liz's playbook: Liz Elliott runs 33 active social brands and runs her whole operation on a set of AI automations spread across several computers. That is the top of this ladder, and it was built one process at a time.

New here? Use the free AI for Entrepreneurs course outline to read this series in the best order.

Do it

Part 1: Find your real level

Read all five. Pick the closest match, not the one you would like to be.

Level 1. One-off asking. You use AI for single tasks. Draft this email. Summarize this PDF. Give me five ideas. It feels impressive, but your actual work has not changed. You still open AI the way you open a search engine. You are probably here if you use AI when you happen to remember, and you do not save prompts anywhere.

Level 2. Saved prompts. You have prompts and Skills you reuse every week. You are faster on your own, but nothing around you has changed. You are probably here if you reuse a few prompts and would notice if AI stopped working for a day.

Level 3. Whole tasks. You have Projects and workflows that handle entire jobs, like drafting briefs, processing customer feedback, or building reports. You are probably here if there are two or three tasks you no longer do by hand, and you use Projects instead of blank chats.

Level 4. Connected and running. Your real tools are connected. Work happens in the background on a schedule instead of only when you open a chat. You are probably here if you review AI output more often than you create from scratch.

Level 5. Rebuilding. You redesign your systems as the tools change, and you spot new uses before they are obvious. You are probably here if you rebuild something monthly and other people ask you to help set theirs up.

Level / You are probably here if

• 1: You use AI when you remember. You do not save prompts.

• 2: You reuse a few prompts. You would notice if AI broke for a day.

• 3: You have two or three tasks you no longer do by hand, and you work inside Projects.

• 4: Work runs in the background. You review output more than you create it.

• 5: You rebuild systems monthly. Other people ask you for setup help.

Part 2: Do the one move for your level

One move. Not five.

• On level 1: save five prompts you reused this month. Put them anywhere you will find them again, in a notes app or a Project.

• On level 2: pick one recurring task that eats more than an hour a week. Build a full Project or workflow for it. A weekly content brief, a Monday morning brief, a client follow-up.

• On level 3: connect your tools. Email, files, calendar, your spreadsheets. Then set up one scheduled task. The shift you are after is work that runs without you opening a chat.

• On level 4: block 30 minutes every Friday and ask one question. What did I do this week that could have been handed off? Rebuild one process a month.

• On level 5: teach it. Explain your setup to five friends, post what you build, or run a small workshop. Explaining your thinking is what improves it.

Part 3: Fix the five habits that keep people on level one

These five habits show up over and over. Each has a prompt in the next section that fixes it.

Habit 1. Starting from a blank chat every time. The AI does not know your job, your day, or what you talked about last time, so every chat starts slow. Fix it with a Role Brief you paste into a Project. Use the Role Brief Builder.

Habit 2. Typing "write me an email" and nothing else. Thin prompts give plain answers, and then you spend your time fixing them. Give your role, who the work is for, the format, the tone, what good looks like, and any limits. Use the Prompt Upgrader.

Habit 3. Copying and pasting out of your own apps. If AI can connect to your email, files, calendar, and chat tools, then copying text out of them by hand is work you do not need to do. Use the Email Summarizer once you have connected your mail.

Habit 4. Using AI for one small thing. If it only rewrites emails for you, you are using a sliver of it. It can also review spreadsheets, plan projects, prep meetings, build visuals, and organize messy information. Use the Spreadsheet Analyzer on one real file.

Habit 5. Never teaching it your voice or preferences. Skip this and the writing stays generic forever. Give it examples of your best writing and save the result. Use the Voice Calibrator.

Part 4: Check your mix with 60/30/10

Levels tell you where you are. This tells you how you spend the time.

The split is 60 percent repetition, 30 percent iteration, 10 percent experimentation. Dharmesh Shah, co-founder of HubSpot, shared this framework in a keynote talk.

• Repetition, 60 percent. Tasks you already trust. Summarizing an article into five bullets. Drafting replies you will edit before sending. Pulling key points out of a meeting transcript. Reformatting a doc or a table. This is the biggest bucket because these tasks pay you back again and again. If AI saves you 20 minutes a day here, that is roughly 80 hours over a year of workdays.

• Iteration, 30 percent. Taking prompts that already work and making them better. Same task, better ask. Add a quality bar to a prompt. Test three opening lines. Move a prompt into a Project so it carries context every time.

• Experimentation, 10 percent. Brand new ideas where you do not know if they will work. Most will fail. Some teach you a lot. The ones that work graduate into your 60 percent.

Now audit yourself. It takes about ten minutes.

1. Open your AI history and list the last 20 things you used it for.

2. Tag each one R, I, or E. Be honest. Something is not experimentation just because it felt fun.

3. Compare your mix to 60/30/10. Almost all R means you are using AI but never improving it. Almost all E means you are chasing ideas and never banking the time savings.

4. Block 30 minutes a week for whichever bucket is empty.

Copy and paste

Role Brief Builder

Use this when every chat starts from zero. It fixes habit 1.

I need you to help me build my Role Brief, which is a set of custom instructions I can paste into a Claude Project so you always know who I am and what I do.

Ask me these questions one at a time:

1. What is your job title, and what do you actually do day to day?
2. Who do you report to, and who reports to you?
3. What are your top 3 priorities right now?
4. What kind of work do you usually ask AI to help with?
5. How do you like things formatted? For example: bullets vs. paragraphs, formal vs. casual, short vs. detailed.
6. Any tools, platforms, or systems you use daily?

After I answer all 6, write me a polished Role Brief I can paste directly into my Claude Project custom instructions.

Make it concise and clear.

Prompt Upgrader

Use this when your prompts are too short. It fixes habit 2.

I am going to give you a basic prompt I normally use.

Your job is to upgrade it into a stronger prompt that gets much better output.

For every prompt I give you, rewrite it to include:

- My role and context. Ask me if you need this.
- Who the output is for.
- The specific format and structure I want.
- The tone and style.
- What "good" looks like. Ask me for an example if needed.
- Any constraints, like word count, things to avoid, or required details.

Show me the before and after side by side so I can see the difference and learn the pattern.

Here is my first prompt to upgrade:

[PASTE YOUR BASIC PROMPT HERE]

Email Summarizer

Use this after you have connected your mail, and read the warning in Watch out first. It fixes habit 3.

Check my Gmail for emails from today.

For each email that actually needs my attention, give me:

- Who it is from
- One-line summary
- Whether I need to reply, just read it, or can skip it
- If I need to reply, draft a short response for me

Skip newsletters, notifications, and marketing emails completely.

Only show me what matters.

Start with the most urgent and work down.

Spreadsheet Analyzer

Use this when you want AI to do more than write emails. It fixes habit 4.

[Upload your spreadsheet]

Analyze this data and give me:

1. The 3 most important trends you see.
2. Anything that looks unusual or concerning.
3. What is improving and what is declining.
4. 3 specific recommendations based on the data.
5. A 2-sentence summary I could paste into a Slack message to my team.

Be specific.

Use actual numbers from the data.

Do not give me generic advice.

Give me insights tied to what is actually in this spreadsheet.

Voice Calibrator

Use this when the writing does not sound like you yet. It fixes habit 5. You need three writing samples.

I am going to paste 3 examples of my best writing below.

Analyze my voice and writing style.

Then create a "Voice Profile" I can save and reuse.

Your Voice Profile should capture:

- My typical sentence length and structure
- Words and phrases I use often
- My level of formality
- How I open and close communications
- My default tone, such as direct, warm, casual, or authoritative
- Things I never do, like jargon I avoid or patterns I do not use

After the profile, write me a short paragraph about any topic in my exact voice so I can check if it sounds like me.

Here are my 3 writing samples:

[PASTE SAMPLE 1]

[PASTE SAMPLE 2]

[PASTE SAMPLE 3]

Two lines that upgrade a prompt you already use

Add either one to a prompt in your 30 percent iteration bucket.

Before answering, find one weakness in this.

Give me a full breakdown, not a summary.

Four experiment starters

Pick one for your 10 percent bucket. They are meant to be a stretch.

What if I had Claude write a script for a podcast episode I haven't recorded yet?
What if Claude analyzed my last 12 months of decisions and told me what patterns I keep falling into?
What if I built a Skill that interviews customers via DM?
What if I had Claude design the brand identity for an unannounced product?

Context in a single line

The smallest possible upgrade. Compare these two on the same piece of text.

Rewrite this.

Rewrite this for my team in my normal voice.

Watch out

Ranking yourself too high

Almost everyone does. If you cannot point at a saved prompt, you are on level one, and that is a fine place to be. The ladder only helps if you stand on the right rung.

Building workflows before you have saved a single prompt

Skipping rungs wastes time. Save the prompts that already work first. Those are what the workflow is made of.

Connecting tools before you know the task

A connected inbox does not help if you have not decided what you want done with it. Get the workflow clear, then connect.

Handing your inbox over without thinking about privacy

Connecting mail, files, and calendar means the AI can read real messages, real client details, and real attachments. Before you turn that on, check your employer's rules, and know that what the AI reads is going to the AI company's servers. Connect the account you are comfortable exposing, not every account you own.

Letting drafted replies go out unread

The Email Summarizer writes replies for you. It does not send them, and you should never let a drafted reply leave without reading every line. A wrong tone in a real client thread costs more than the ten seconds you saved.

Trusting an auto-pulled report without checking it

When a weekly update is built from your real mail and calendar, it can quietly miss something or misread a thread. Skim the source before you send the summary on.

Treating the setup as finished

The tools change. Once you are running things in the background, rebuild one process a month or the setup slowly rots.

Guessing your mix instead of checking it

Most people believe they do all three buckets. Most do not. Look at the actual history before you decide.

Check yourself

• [ ] I picked my real level, not the one I want.

• [ ] I can say the one move for my level out loud.

• [ ] I saved at least five prompts I reuse.

• [ ] I ran the Role Brief Builder and saved the result somewhere reusable.

• [ ] I upgraded one thin prompt with the Prompt Upgrader.

• [ ] I used AI for one task bigger than an email.

• [ ] I listed my last 20 AI sessions and tagged each R, I, or E.

• [ ] I booked 30 minutes this week for whichever bucket was empty.

Do this now

Open your AI history, look at the last 20 things you used it for, and tag each one R, I, or E. Your gap will be obvious in under ten minutes.

Keep going

Next in AI for Entrepreneurs:

• Part 4: Starting the right way, and the first week that works

• Part 5: The honest picture: energy, tiredness, attention, and where this goes

• Part 6: Models explained, and why one favorite is a trap

Every part of this course sits in reading order in the free AI for Entrepreneurs course outline.

This part is free, start to finish. Subscribing keeps the rest of the course coming and opens the paid parts.

Tell me in the comments which level you landed on, and what your one move is, and I will point you at the right next part.

That is ten minutes a day spent reintroducing yourself to a machine that forgot you. Over a year it adds up to days. Worse, you get tired of typing the background, so you start skipping it, and the answers go back to average.

What you will be able to do

• Write a heading in Markdown using nothing but hash marks.

• Build a one-page How I Work file that covers your role, your goals, and your rules.

• Paste that file into any AI chat so you stop repeating yourself every time.

• Name the things that must never go in the file.

• Test in one question whether the AI is actually using it.

This is for anyone who saw a file ending in .md and assumed it was a coder thing. It is not. It is a note with a few hash marks in it.

The proof behind this

Liz Elliott runs her whole operation on AI automations spread across several computers, and plain text files like this one hold it together. She is not an engineer. She rebuilt a 403-guide archive into this 103-lesson course, and every lesson in it is one of these files.

What this lesson is

Lesson 1.2 of the free AI for Entrepreneurs course, in Module 1: What AI Actually Is.

An .md file is a Markdown file. That sounds technical. It is not.

It is plain text with a few simple marks that make it easy to read. That is the whole thing. You are not coding an app. You are writing a clean note.

AI reads these files very well, which is why they show up everywhere in AI work. Skills, instructions, memory files, agent notes. Under all of it is plain text with a few hash marks.

Think about a recipe card box in a kitchen. Each card is short. Each one has a title at the top so you can find it. Nobody needs training to read a recipe card. You flip to the one you need and follow it.

That is what a Markdown file is for AI. A note it can find and follow. Give an assistant a folder of these notes and it reads them like a table of contents, picking the right one when it needs it. Your work style is one card. Your business facts are another. Your rules are another.

The best first card is a How I Work file. One page that says who you are, what you do, how you like to be talked to, and what the AI should never do. Write it once. Paste it in, or save it in settings, and stop explaining yourself in every new chat.

That is why this course keeps coming back to files. A good chat helps you today. A good file helps you every day after that.

New here? Use the free AI for Entrepreneurs course outline to read this series in the best order.

Do it

Part 1: Learn the four marks

You do not need to learn a whole language. These are the marks you will use most.

One hash makes a big heading. Two hashes make a smaller one. Three hashes make a smaller one still. A dash makes a bullet.

That is enough to start. Everything else can wait.

Part 2: Write your How I Work file

The file should tell the AI six things:

1. Who you are.

2. What you do.

3. How you like to work.

4. How you want it to talk to you.

5. What you need help with.

6. What it should never do.

There are three ways to build it. Pick whichever one feels easiest. The finished file is yours either way.

• By hand. Open a plain text file, name it how-i-work.md, and fill in the template below.

• By interview. Ask Claude or ChatGPT to interview you, then turn your answers into a Markdown file.

• By talking. Turn on voice-to-text, talk out loud for five minutes about your work, then say "turn this into a clean markdown file."

Part 3: Keep secrets out

This file carries context, not secrets.

Safe to include: your role, your goals, your voice, your business facts, your work style.

Leave out: passwords, card numbers, bank numbers, ID numbers, customer personal information, your home address.

The test is simple. If you would not paste it on a public website, it does not go in the file.

Part 4: Put it to work

The easy way works in Claude, ChatGPT, Gemini, and almost every AI tool. Start a new chat. Paste the priming line, then paste your whole file, then ask your question. Do this first, before you set up anything fancier.

To save it in Claude, open Settings, find the instructions section for Claude, and paste your file there. Now every chat starts with it.

To save it in ChatGPT, open Settings, go to Personalization, then Custom Instructions. You get two boxes. The first is for who you are and what your goals are. The second is for how you want it to talk to you and what it should never do. Each box has a size limit, so put a shorter version there and keep the full file for pasting.

Menu wording changes between versions. If you do not see those exact words, look through Settings for the section about how the AI should respond to you. It will be there under some name.

Part 5: Keep it alive

The file gets better the more you use it.

Every time you catch yourself explaining the same thing again, add that line to the file. Update the date at the top when you change it. Once a month, skim it and refresh your goals.

Then run the test question below. If the AI repeats your rules back to you, your setup is working.

Copy and paste

The Markdown marks

The four marks that cover most of what you will write.

# Title
## Section
### Small section
- Bullet point

The How I Work file template

Fill in every bracket, delete what does not apply, and save it as how-i-work.md.

# How I Work
Last updated: [DATE]

## Who I am
[Your name, your role, your business or job, where you are based if it matters.]

## What I do
[The work you actually do day to day. Two or three sentences.]

## What I need help with
[The tasks you want AI help with most. List three to five.]

## How I like to work
[Bullets or paragraphs. Short or detailed. Fast drafts or careful ones.]

## How I want you to talk to me
[Tone, formality, length. Say it plainly, for example: casual, direct, no long intros.]

## What you should never do
[Your hard rules. For example: never invent a statistic, never send anything on my behalf, never use words I have banned.]

The priming line

Paste this above your file at the top of any new chat.

This is my How I Work file. Use it for everything in this chat. Read it, confirm you have it, then wait for my question.

The test question

Run this after you paste or save the file, to check it landed.

Based on my file, how should you talk to me and what should you never do?

The voice-to-text instruction

Say this out loud after you have talked through your work.

turn this into a clean markdown file

Watch out

Adding secrets instead of context

Passwords, card numbers, bank details, ID numbers, customer information, and your home address do not belong in a file you will paste around. Keep it to context you would be comfortable sharing.

Making it so private it says nothing

The opposite mistake. A file with no real detail in it does no work. It should carry enough about your role, goals, and rules to change the answers you get.

Writing it once and never touching it again

The file goes stale. Add a line every time you explain something twice. Skim it monthly.

Keeping it in a note app and never pasting it

A file that lives in your notes and never reaches the AI does nothing. Paste it into the chat or save it in settings.

Making it so long the rules get buried

Keep a short version for settings boxes with size limits. Keep the full version for pasting and for Projects.

Check yourself

• [ ] I can write a heading and a bullet in Markdown without looking it up.

• [ ] I have a file saved as how-i-work.md.

• [ ] It covers who I am, what I do, how I work, how to talk to me, and what to never do.

• [ ] It has no passwords, card numbers, IDs, addresses, or customer details in it.

• [ ] It has a date at the top.

• [ ] I pasted it into a chat with the priming line.

• [ ] I ran the test question and the AI repeated my rules back.

Do this now

Open a blank file, name it how-i-work.md, and fill in the template above. Ten minutes is enough for a first version.

Keep going

Next in AI for Entrepreneurs:

• Part 3: The five levels of AI use, and which one you are on

• Part 4: Starting the right way, and the first week that works

• Part 5: The honest picture: energy, tiredness, attention, and where this goes

Every part of this course sits in reading order in the free AI for Entrepreneurs course outline.

This part is free, start to finish. Subscribing keeps the rest of the course coming and opens the paid parts.

Tell me in the comments what the first line of your How I Work file says, and I will point you at the right next part.

Skipping one word is cheap. What happens next is not. The following paragraph uses that word to explain the next idea, so now two things are fuzzy. Three pages later you decide AI is not for you. The tool was fine. One skipped word did all of that.

What you will be able to do

• Say what a prompt, a token, a context window, and a hallucination are, in your own words.

• Tell a connector apart from an MCP, and a skill apart from a plugin.

• Guess roughly how many tokens a page of text will use.

• Look up any AI word you meet later in about ten seconds.

• Stop nodding along in a meeting when someone drops a term you do not know.

This is for anyone who has felt behind in an AI conversation. You do not need a tech background. Nothing here assumes you have used AI before.

The proof behind this

Liz Elliott is not an engineer and never studied computer science. She runs 33 active social brands and built four software products with AI help: PRISM, AutoViral, UPRISE Social Studio, and Profounder. She learned every one of these words as a beginner, which is why this course starts with them instead of assuming them.

What this lesson is

Lesson 1.1 of the free AI for Entrepreneurs course, in Module 1: What AI Actually Is.

AI has a pile of words that people use like everyone already knows them. Tokens. Context window. Agent. MCP. Nobody hands you the list. You are just supposed to have it.

You do not have to fake it. Every one of these words has a plain meaning. Most of them are simpler than they sound.

Think about walking into a coffee shop the first time. Grande, macchiato, flat white, single origin. It sounds like a code. Once someone spends two minutes explaining the words, you order without thinking about it ever again. The coffee was never hard. The menu was just unfamiliar.

AI words work the same way. There are about ten you will hit in your first week. There are another twenty you will meet over the next few months. That is the whole list. It is not growing as fast as it feels.

Here is the thing to hold on to. If you skip a word you do not know, the confusion piles up. The next paragraph uses that word to explain the next idea, and now two things are fuzzy. Three pages later you decide AI is not for you. The tool was fine. One skipped word did all that.

So this lesson is one dictionary, not four. The ten that matter come first. The rest sits below in alphabetical order. You are not meant to memorize any of it. You are meant to read the ten, skim the rest, and come back here whenever a word trips you.

From Liz's playbook: Liz Elliott is not an engineer and did not study computer science. She teaches beginners and non-technical people, so nothing in this course assumes you already know the words.

New here? Use the free AI for Entrepreneurs course outline to read this series in the best order.

Do it

Part 1: The ten that matter

Read these ten now. They cover almost everything you will meet in your first month.

Prompt. What you type into the AI. It is the instruction you give it.

Model. The AI engine doing the work. One tool can offer several models, and they are better at different jobs.

LLM. Short for large language model. It is the brain behind a chatbot. It read a huge amount of text and learned to predict the next word, over and over, which is enough to write drafts, plans, summaries, and code.

Token. A chunk of text the AI reads. It does not always see whole words. The word basketball might split into basket and ball.

Context window. How much the AI can hold in its head during one chat. Go past it and the AI starts forgetting the beginning of the conversation.

Hallucination. When the AI makes something up and says it like it is true. It does not lie on purpose. It guesses, and sounds sure while it guesses.

Memory. What the AI keeps about you between chats. Helpful when it is right. Annoying when it is stale.

Project. A saved workspace that holds files, instructions, and chat history for one topic or job. Context stays put between conversations instead of vanishing.

Connector. A ready-made, one-click link between the AI and an app you already use, like your email, your calendar, or your files. With one on, the AI can go get the information itself instead of waiting for you to paste it.

Agent. An AI that takes several steps on its own to finish a whole task. You ask for the outcome instead of each step. Without an agent you say "search flights, compare prices, fill in my details, click pay." With an agent you say "book me a flight."

Part 2: The rest, A to Z

Skim this once. Do not memorize it. Come back when a word shows up.

Term / Plain meaning

• Agentic: A word for AI that plans and acts across several steps instead of just replying.

• AGI: Artificial general intelligence, the idea of AI that could do any task a person can, including tasks it was never trained on. It does not exist yet.

• Artifact: A finished thing the AI builds beside the chat, like a document, a slide, a web page, or a small working tool you can click.

• Bias: When answers lean one way because the training text leaned that way.

• Chatbot: An AI you talk to that answers and then waits for you.

• Computer Use: A mode where the AI can see a screen, click, and type its way through apps, for jobs like filling in forms or moving data between tools.

• Cowork: A shared workspace where a team can work with Claude on the same projects and skills.

• Deep Research: A slower mode where the AI checks many sources and writes you a report. It can take 10 to 30 minutes.

• Extended Thinking: A mode where the AI reasons for longer before it answers.

• Fine-tuning: Extra training that reshapes a model using your own data. It is a developer job and it is expensive. Almost nobody reading this needs it.

• Generative AI: The umbrella name for AI that makes new things: text, images, video, music, or code. If it creates instead of only sorting, it is generative.

• Knowledge cutoff: The date the model's training text stops. Anything newer than that needs a web search.

• Machine learning: Teaching a computer from examples instead of typing out every rule by hand. Show it thousands of emails marked spam or not spam and it learns the signals itself.

• MCP: Model Context Protocol. An open standard, like an agreed plug shape, that lets tools without a ready-made connector hook into AI anyway.

• Multimodal: An AI that handles more than text, like images, audio, and video.

• Open weight model: A model anyone can download and run for free. People often call these open source, but that phrase means more than just downloadable, so open weight is the safer word.

• Plugin: A bundle that installs several skills and connectors at once, so you get a whole setup instead of building it piece by piece.

• RAG: Retrieval-augmented generation. A setup that lets AI answer from your own documents instead of only from what it was trained on.

• Scheduled task: A job you set to run at a chosen time, like every Monday at 8am.

• Skill: A saved set of instructions for one task, so you stop explaining your method every single time.

• Sycophancy: When the AI agrees with you too much instead of pushing back.

• System prompt: Standing instructions that sit underneath a chat, a Project, or a Skill and shape every answer by default.

• Vibe coding: Building software by describing what you want in plain words and letting the AI write the code.

Part 3: The two mix-ups that trip almost everyone

Connector versus MCP. These get used as if they mean the same thing. They do not. A connector is a finished, one-click integration the AI company built for a popular app. MCP is the open standard that other tools use to plug in when no ready-made connector exists. As a user they feel similar. Under the surface they are different, and you will meet MCP again.

Skill versus plugin versus Project. A connector is the cable. A skill is the manual. A plugin is the package that holds several cables and manuals together. A Project is the workspace where you keep the job.

Part 4: The token math, and what to do when sources disagree

Tokens matter because limits and prices are counted in tokens, not words.

Text amount / Tokens

• 1 token: About three quarters of a word

• 100 tokens: About 75 words

• 1 word: About 1.3 tokens

• A page, around 375 words: About 500 tokens

Sources disagree here. Some say a token is about half a word. Some say it is about one word. Some say it is about three quarters.

The rule for picking: use three quarters of a word. It is the version that matches the math table above, and it lines up with the common estimate that a 5,000 word document is around 6,500 tokens. Exact counts vary by model, so treat every token number as a rough guide, never a promise.

Context windows are big now. The newest top models can hold roughly a million tokens at once, which is about 750,000 words. That number moves with the model and sometimes with your plan, so check what your own tool says instead of trusting a number from an article.

Part 5: When a word is not on this list

Do not guess and do not scroll past. Paste it into the AI and ask.

Copy and paste

The term explainer

Use this any time a new AI word shows up and you do not want to fake it.

Explain this AI term in plain English: [TERM]

Tell me three things:
1. What it means in one simple sentence.
2. When I would actually use it.
3. Whether a non-technical person needs to care about this, or can safely ignore it for now.

The prompt shape

Use this to turn a vague request into one that works. Fill in every bracket.

Write a [length, for example 3-sentence] [format, for example Instagram caption]
for [who it is for, for example a life coach]
targeting [the exact audience, for example burnt-out 9-to-5 workers],
[tone, for example casual],
end with [the ending you want, for example a question].

Vague versus specific

Keep this pair where you can see it. It is the fastest lesson in prompting.

VAGUE: Write something about marketing

SPECIFIC: Write a 3-sentence Instagram caption for a life coach targeting burnt-out 9-to-5 workers, casual tone, end with a question

Three one-line fixes for three quality problems

Add these lines when the AI makes things up, agrees too easily, or leans one way.

If you do not know something, say "I don't know" instead of guessing.

Argue against me. Do not just support what I said.

Give me the strongest case for the opposite view.

Watch out

Nodding along instead of asking

You read past a word you do not know and the confusion compounds. Three paragraphs later nothing makes sense. Look it up, then keep moving. Ten seconds now saves the whole page.

Blaming the model for a vague prompt

"Write something about marketing" was never going to give you anything usable. Before you decide the AI is weak, check whether you named the length, the audience, the tone, and the format.

Trusting a confident answer

A made-up fact sounds exactly as sure as a correct one. That is what makes it risky. Check names, dates, prices, numbers, and anything legal or medical. Treat AI output as a first draft, never as the final authority.

Staying in one long chat forever

Once you fill the context window, the AI starts dropping the beginning of the conversation. If a long chat starts feeling forgetful, that is your sign. Start a fresh chat and paste in the few things it still needs.

Counting words when the bill counts tokens

Your word count and your token count are not the same number. Use 100 tokens for about 75 words, and about 500 tokens for a page.

Confusing today's chatbot with AGI

AGI is a debate about the future. It is not a description of the tool in front of you. Arguments about AI taking over are usually arguments about AGI.

Reaching for fine-tuning

Fine-tuning sounds like the serious option. It is technical, it costs money, and you almost certainly do not need it. A good Skill or Project gets you most of the same benefit.

Trying to learn every term at once

You do not need the whole list on day one. Learn the ten. Pick up the others as they come up in real work.

Check yourself

• [ ] I can explain a prompt, a token, a context window, and a hallucination without looking.

• [ ] I said three of the ten terms out loud in my own words.

• [ ] I know a connector is a ready-made link and MCP is the open standard for everything else.

• [ ] I know a skill is a saved method and a plugin is a bundle of them.

• [ ] I can estimate that a page of text is roughly 500 tokens.

• [ ] I know to start a fresh chat when a long one goes foggy.

• [ ] I have the term explainer prompt saved somewhere I can find it.

• [ ] I know AGI is not what I am using today.

Do this now

Pick the three terms on this page you could not have explained ten minutes ago. Say each one out loud, in your own words, without looking back at the definition.

Keep going

Next in AI for Entrepreneurs:

• Part 2: What an .md file is, and why this course lives in them

• Part 3: The five levels of AI use, and which one you are on

• Part 4: Starting the right way, and the first week that works

Every part of this course sits in reading order in the free AI for Entrepreneurs course outline.

This part is free, start to finish. Subscribing keeps the rest of the course coming and opens the paid parts.

Tell me in the comments which AI word tripped you up first, and I will point you at the right next part.

Published lessons are links. Everything else says Coming, with the date it goes out.

Start at the top if you are new. If you already use AI every day, skip to Part 3.

New lessons land every other day.

Part 1 Foundations

What AI is, which tool to pick, how to set it up, how to talk to it

22 lessons

1. What AI Actually Is

The words and the file

• The words you need before anything else — Coming Aug 18, 2026

• What an .md file is, and why this course lives in them — Coming Aug 20, 2026

Where you are starting from

• The five levels of AI use, and which one you are on — Coming Aug 22, 2026

• Starting the right way, and the first week that works — Coming Aug 24, 2026

The honest picture

• The honest picture: energy, tiredness, attention, and where this goes — Coming Aug 26, 2026

2. Choosing Your AI

How models actually differ

• Models explained, and why one favorite is a trap — Coming Aug 28, 2026

• The four Claudes, and which one you actually want — Coming Aug 30, 2026

• Fast or deep: picking a model per task — Coming Sep 1, 2026

Picking your tools

• ChatGPT and Claude, compared fairly — Coming Sep 3, 2026

• Your starter stack, and what to ignore — Coming Sep 5, 2026

3. Setting It Up

Your first three days

• Day one: the five things to do first — Coming Sep 7, 2026

• Day two and three: the next ten things — Coming Sep 9, 2026

Pick a setup path

• The 60 minute setup — Coming Sep 11, 2026

• The full one day setup — Coming Sep 13, 2026

Where to work, and what to claim

• Desktop, phone, or browser: where to actually work — Coming Sep 15, 2026

• The settings nobody checks, and the free stuff nobody claims — Coming Sep 17, 2026

4. Talking To AI

How a prompt is built

• The anatomy of a prompt — Coming Sep 19, 2026

• Say what you want, not how to do it — Coming Sep 21, 2026

• The steering wheel: seven levels of control — Coming Sep 23, 2026

Getting sharper answers

• Words that change the answer — Coming Sep 25, 2026

• Prompt packs worth saving — Coming Sep 27, 2026

Past the single prompt

• Stop prompting, start looping — Coming Sep 29, 2026

Part 2 Make It Yours

Give it memory, keep it honest, plug it into your real tools

19 lessons

5. Memory And Context

Giving AI a memory

• Why AI forgets you, and the fix — Coming Oct 1, 2026

• The most important file you will ever write — Coming Oct 3, 2026

• The about-me file — Coming Oct 5, 2026

• Global instructions that apply everywhere — Coming Oct 7, 2026

Feeding it what you know

• Knowledge files and your second brain — Coming Oct 9, 2026

• Context hygiene, tokens, and never hitting a limit — Coming Oct 11, 2026

Your voice, and getting better

• Teaching AI your voice and your taste — Coming Oct 13, 2026

• Making it get better every week — Coming Oct 15, 2026

6. Trust And Safety

Catching what it gets wrong

• The three ways AI lies to you — Coming Oct 17, 2026

• Check the work before you use it — Coming Oct 19, 2026

• Build a council, not a cheerleader — Coming Oct 21, 2026

Setting hard limits

• The safety playbook — Coming Oct 23, 2026

• Guardrails, permissions, and what AI may never touch — Coming Oct 25, 2026

• Do not let AI embarrass you — Coming Oct 27, 2026

7. Connecting Your Tools

Making the connection

• Connectors: what they are and how to add them — Coming Oct 29, 2026

• MCP, explained without jargon — Coming Oct 31, 2026

Add-ons and marketplaces

• Plugins, and how they differ from skills — Coming Nov 2, 2026

• The free skills marketplace — Coming Nov 4, 2026

Putting it to work

• AI as an assistant across every app you own — Coming Nov 6, 2026

Part 3 Build

Skills, agents, apps, and design work you make yourself

26 lessons

8. Building Skills

What a skill is

• Skill, workflow, or agent: which do you need — Coming Nov 8, 2026

• Skills in ten minutes — Coming Nov 10, 2026

Building one that works

• Three ways to build one — Coming Nov 12, 2026

• Building skills that actually work — Coming Nov 14, 2026

• Turn any guide into a skill — Coming Nov 16, 2026

Ready-made libraries

• A skill library for your life — Coming Nov 18, 2026

• A skill library for your job — Coming Nov 20, 2026

9. Agents And Automation

Agent basics

• What an agent is, in plain words — Coming Nov 22, 2026

• Build your first agent — Coming Nov 24, 2026

More than one at a time

• Agent teams and org charts — Coming Nov 26, 2026

• Running several at once — Coming Nov 28, 2026

• Stop babysitting it — Coming Nov 30, 2026

Other platforms

• Managed agents and other platforms — Coming Dec 2, 2026

Choosing what to hand over

• What to hand off, and what to keep — Coming Dec 4, 2026

• The weekend automation list — Coming Dec 6, 2026

10. Building Things

Apps you make by asking

• Artifacts: apps you make by asking — Coming Dec 8, 2026

• AI can build you any tool — Coming Dec 10, 2026

Coding without coding

• Vibe coding for people who do not code — Coming Dec 12, 2026

• Claude Code, all seven levels — Coming Dec 14, 2026

Real projects

• Real builds: apps, stores, landing pages — Coming Dec 16, 2026

• Home and business builds worth copying — Coming Dec 18, 2026

11. Design And Visuals

The design tool

• Claude Design, start to finish — Coming Dec 20, 2026

• Why your AI designs look like AI — Coming Dec 22, 2026

Borrowing and replacing

• Borrowing a layout the right way — Coming Dec 24, 2026

• Replacing a designer — Coming Dec 26, 2026

Decks, images, and sight

• Decks, images, and AI that can see — Coming Dec 28, 2026

Part 4 Use It

Work, career, learning, money, life, business, marketing

36 lessons

12. Getting Work Done

The working day

• Ten ways to use AI at work today — Coming Dec 30, 2026

• Meetings, prep, and follow-through — Coming Jan 1, 2027

• Making better decisions — Coming Jan 3, 2027

Getting ahead, and teams

• Reviews, raises, and promotion — Coming Jan 5, 2027

• Cowork: AI for a whole team — Coming Jan 7, 2027

13. Your Career

Where you stand

• Which parts of your job AI should already do — Coming Jan 9, 2027

• Do not get replaced, become the one who runs it — Coming Jan 11, 2027

Getting hired

• Resume and LinkedIn — Coming Jan 13, 2027

• Interview prep — Coming Jan 15, 2027

• A 30 day job search system — Coming Jan 17, 2027

Reading the market

• The new AI jobs, and how to qualify — Coming Jan 19, 2027

• Certifications worth putting on a resume — Coming Jan 21, 2027

14. Learning And Thinking

Learning AI itself

• How to learn AI for free — Coming Jan 23, 2027

• Turning AI into a tutor — Coming Jan 25, 2027

Research and language

• Real research instead of guessing — Coming Jan 27, 2027

• Languages — Coming Jan 29, 2027

Capturing, and deciding

• Notes, voice, and audio — Coming Jan 31, 2027

• Finding your gaps — Coming Feb 2, 2027

• Designing the life you want — Coming Feb 4, 2027

15. Money And Life

Money

• Five ways AI should be saving you money — Coming Feb 6, 2027

• Personal finance tasks to hand over — Coming Feb 8, 2027

• Investing, with the warnings intact — Coming Feb 10, 2027

• Legal help and spotting scams — Coming Feb 12, 2027

Home and travel

• Home, family, and the Sunday reset — Coming Feb 14, 2027

• Travel and big events — Coming Feb 16, 2027

Your body

• Health, food, and daily habits — Coming Feb 18, 2027

16. Business And Marketing

Starting the thing

• Setting up AI for a small business — Coming Feb 20, 2027

• Pressure testing an idea — Coming Feb 22, 2027

• Building a company with AI — Coming Feb 24, 2027

• Moats, stacks, and what lasts — Coming Feb 26, 2027

Getting customers

• Email marketing — Coming Feb 28, 2027

• Ads on Google, Meta, and Reddit — Coming Mar 2, 2027

• Getting found: the new SEO — Coming Mar 4, 2027

Content and selling

• A content system that runs itself — Coming Mar 6, 2027

• Selling with AI — Coming Mar 8, 2027

Earning from it

• Making money with AI — Coming Mar 10, 2027

How the course is built

103 lessons. 16 modules. 4 parts. Written for people who have never opened an AI tool.

Every lesson has the same shape, so you always know where you are: what you will be able to do, the idea, do it, the prompts to copy, what to watch out for, a checklist, and one thing to do now.

No code. Third grade reading level on purpose. If a lesson needs a technical word, it explains it first.

Until this morning you needed 100,000 followers to claim a Search profile. Now you need 35,000. On TikTok it dropped from 300,000 to 100,000.

That one change moved a huge group of creators from locked out to eligible overnight.

Here is the part people are going to miss. This is not another social profile. This is a page on Google itself, and it changes what Google shows the next time somebody types your name.

What a Search profile actually is

Google launched Search profiles in June 2026. It is a free page that pulls your whole online life into one place:

• Your social accounts

• Your website

• Your latest posts and videos

• Any links you want to add

• A short bio you write yourself

You can pin your best posts to the top so those are the first thing people see.

Your content feeds in on its own through the platforms you connect. You do not upload anything by hand. New posts usually show up within a day.

Then there is the bigger piece. When somebody Googles your name, Google may show a Knowledge Panel. That is the box with your photo and your info off to the side of the results. If you do not have one, claiming your profile may get you one. If you already have one, claiming your profile makes it better. Your photo, your links, your pinned posts, all of it starts coming from you instead of from whatever Google scraped together.

Google also writes a short AI summary of you inside that panel, and it builds that summary from the bio you write.

Read that last line again.

Five reasons this is worth your ten minutes

I know how this reads. Another platform, another profile to fill out, another thing on the list.

Here is why I stopped what I was doing and set mine up.

Google is where people check you out. Somebody sees your Reel, likes it, and wants to know if you are real. They do not open Instagram again. They Google your name. Whatever comes up is your first impression, and until now you had almost no say in it.

You get to write what AI says about you. The summary in your Knowledge Panel comes from your bio. Every time somebody asks an AI who you are, that panel is part of what it reads. This is the first time creators get to hand Google the words instead of hoping it guesses right.

It feeds Google Discover. Discover is the feed on your phone in the Google app. It sends real traffic. Your profile helps Google connect your content to you as a person, and people can find their way to your profile from your content there.

Your name is a search term. People already search it. Right now those searches land on whatever Google decided. A profile turns that into a page you control, with links pointing where you want them to point.

It costs nothing and takes ten minutes. The math on that is easy.

Free stuff from a big platform never stays this easy to get. Early is the whole advantage here.

Who qualifies now

Here are Google's own numbers, straight off its help page:

• YouTube: 35,000

• Instagram: 35,000

• X: 35,000

• TikTok: 100,000

You only need to clear the bar on one of them. Not all four. If you have 41,000 followers on Instagram and 900 on everything else, you are in.

The rest of the rules:

• You must be 18 or older.

• Your account has to be public.

• Your content has to follow Google's policies.

• You have to be in the US for now.

Google says it plans to open this up to more countries later in the year. If you are outside the US, this is a wait, not a no.

Check before you build anything

Do this first. Google may have already made a profile for you.

If you already have a Knowledge Panel and you meet the follower rules, Google may have generated a profile in the background. You would claim that one instead of starting fresh.

Search your own name on Google and look for a panel on the side of the results. If there is one, look for a link to claim it.

Either way, both roads start at the same place.

The 10-minute setup

1. Go to the claim page.

Open profile.google.com/claim and sign in with the Google account you want tied to this. Think about which one. This is the account that owns your public presence on Google now, so pick the one you actually keep.

2. Connect the platform that qualifies you.

Sign in to at least one platform that clears the follower bar. Google checks the count on its own. You do not send screenshots or fill out a form.

3. Add your other accounts.

Link the rest of your real accounts. One warning from Google's own docs: only link accounts that are actually yours and actually related to you. Random unrelated links can get your profile denied.

4. Write the bio like it is your only shot.

This is the step everybody is going to rush, and it is the one that matters most. Google builds your AI summary from this. So write it like you are teaching Google how to describe you.

Say the plain version of what you do, who you help, and what you are known for. Skip the clever tagline. If your bio says "storyteller and dreamer," that is what Google has to work with. If it says "Liz Elliott teaches beginners how to use AI to make money online," Google knows exactly what to say about you.

5. Add your website and your money link.

Your profile holds links. Put the one that actually earns first. For me that is my Substack. For you it might be a shop, a booking page, or an email list.

6. Pin your best work.

You choose what sits at the top. Do not let the newest post win by default. Pin the post that makes a stranger trust you in five seconds.

7. Save it and check your name.

Search your name in Google. Give it some time. If nothing shows up, that is often a name collision. Common names are harder for Google to sort out, and that is Google's own explanation, not a guess.

You can get back in any time at profile.google.com/account, or by tapping your photo in the Google app and going to "Your Search profile."

One more thing to know before you edit later. Changes to your name, your bio, or your social links go into a "Pending" state while Google reviews them. So get it right the first time instead of planning to fix it Thursday.

What this does not do

I am not going to oversell this.

A Search profile does not make your content rank higher. Google says that outright. Claiming it is not an SEO trick and anybody telling you it is has not read the docs.

It does not get you paid. There is no money attached to it.

It does not replace your website. It is a front door, not a house.

What it does is control your first impression on the biggest search engine on earth, for free, in ten minutes. That is enough.

If you are under 35,000

Then this is your target, and it is a real one now.

35,000 followers on one platform. Not 100,000. Not four platforms. One.

That number is reachable inside a year for somebody posting with a plan. I went from 0 to 30,000 followers in 90 days once, and I have written up exactly how that worked. The bar moved toward you today. Go take it.

And watch this pattern, because it is the actual lesson. Platforms open the door quietly, at the bottom, and the people who walk through early get treated better than the people who show up after everybody else. That has been true for every platform I have been early on.

The proof behind this

I made $100K in my first 27 days online with zero experience. I built a no-code SaaS that made $60K. My first digital product made $30K. I have 241K followers on Instagram and 100K+ readers on Substack.

I am not selling you a course on this one. I read Google's docs, set up my own profile, screenshot the parts that matter, and wrote down what I did.

New here? Use the free One Person Empire course outline to read this series in the best order.

Keep reading

• Make Your First $20K on Instagram Before You Hit 1,000 Followers

• 0 to 30K Followers in 90 Days: How to Hack the Instagram Algorithm

• Instagram Leaked Its Viral Formulas

• How to Grow an Audience Before You Have a Product

• The Marketing Funnel Died in 2026: The 4-Part Framework That Made Me $100K in 27 Days

• 30 Social Media Content Prompts to Revive a Stale Account in 30 Days

Tell me in the comments what your follower count is on your best platform, and I will tell you how far you are from the door.

Source note

Every number here comes from Google's own pages. The follower rules and the US-only note are from Google's Search Help page "Create a new Search profile" and the eligibility answer on creators.google. The change from 100,000 to 35,000 was posted by Robby Stein, VP of Product for Google Search, on August 13, 2026. The screenshots are mine, taken from those pages today. I have no traffic numbers from my own profile yet because I set it up this morning. When I have them, you get them.

What's next

Paid subscribers get the full playbook behind these posts: the exact steps, the prompts, the real numbers, and the mistakes I paid for so you do not have to. Next up is the Pinterest lesson on finding the exact words your buyer types without paying for a keyword tool, plus the 8-board setup for a second brand.

Google just moved the bar toward you. Go stand on the right side of it.

This prompt keeps your real face, moves you into a bright blue sky, adds butter-yellow shades, and builds the curved fisheye text style right into the photo. I made the full prompt free so you can copy it and try it today.

What you need

Start with one clear photo of your face. Use a photo with good light. Your face should not be covered. The photo is only there to help ChatGPT keep you looking like you.

Open a new ChatGPT chat and attach your selfie. Then paste the full prompt below. Do not shorten it. The long list is what keeps the face, hand, sky, light, hair, text, and safe zones working toward the same result.

The full butter fisheye prompt

Copy everything inside this box:

CREATE ONE EXTREMELY REALISTIC VERTICAL 9:16 OUTDOOR SELFIE OF THE PERSON IN THE UPLOADED PHOTO.

The uploaded selfie is a STRICT IDENTITY REFERENCE ONLY.

Use it to preserve the EXACT SAME REAL PERSON. Everything else must follow the specifications below.

==================================================
IDENTITY
==================================================

IDENTITY IS THE #1 PRIORITY.

The finished image must immediately and unmistakably look like the exact real person in the uploaded selfie.

Preserve their real:
• face shape
• forehead and hairline
• eyebrow shape and placement
• eye shape, size, spacing, and placement
• nose shape and proportions
• lips and mouth proportions
• cheeks and cheekbones
• jaw and chin
• skin tone and undertone
• natural facial asymmetry
• approximate age
• distinctive facial characteristics
• general natural hair coloring

Do NOT copy the reference selfie’s pose, expression, hairstyle, clothing, background, framing, camera angle, or lighting.

Do not beautify them into another person.
Do not create a generic AI face.
Do not reshape the nose, lips, eyes, jaw, cheeks, or forehead.
Do not de-age them.
Do not excessively smooth their skin.

==================================================
FORMAT + COMPOSITION
==================================================

Generate EXACTLY ONE finished image.

True vertical 9:16 composition.

Final resolution:
1080 × 1920 pixels.

Make this an EXTREMELY SKY-HEAVY low-angle selfie.

Approximately 86–90% of the composition should visually read as vivid blue sky and open negative space.

Confine the person to the VERY BOTTOM of the image.

Do NOT center the person vertically.

Do NOT enlarge the person to fill the tall canvas.

Show only:
• hair and bun
• face
• neck if at all
• very tops of shoulders if any
• at most a tiny hint of the tank top

The extra vertical space belongs ABOVE the subject as sky.

The image should look like the person noticed something interesting in the sky, held the camera low, tilted it dramatically upward, and quickly took a spontaneous selfie.

==================================================
INSTAGRAM + TIKTOK TEXT SAFE ZONE
==================================================

ALL ON-IMAGE TEXT MUST REMAIN COMPLETELY INSIDE A CONSERVATIVE SHARED INSTAGRAM REELS, INSTAGRAM STORIES, AND TIKTOK SAFE ZONE.

On the 1080 × 1920 canvas, keep the ENTIRE typography composition inside this exact region:

• no closer than 8% from the LEFT edge
• no closer than 22% from the RIGHT edge
• no closer than 15% from the TOP edge
• no closer than 36% from the BOTTOM edge

Approximate pixel boundaries:

• left boundary: x = 86 pixels
• right boundary: x = 842 pixels
• top boundary: y = 288 pixels
• bottom boundary: y = 1229 pixels

In proportional terms, every part of the typography must remain between:

• x = 8% and x = 78% of the canvas width
• y = 15% and y = 64% of the canvas height

THIS SAFE ZONE IS ABSOLUTE.

Every:
• letter
• number
• punctuation mark
• serif
• italic flourish
• ascender
• descender
• curved baseline
• enlarged central letter
• warped outer edge
• fisheye bulge

must remain completely inside this safe zone.

Do not allow the fisheye distortion to push any portion of any letter outside the safe zone.

Keep the entire rightmost 22% of the image free from text so TikTok and Instagram interface buttons cannot cover it.

Keep the entire bottom 36% free from text so captions, account information, call-to-action buttons, and platform controls cannot cover it.

Keep the entire top 15% free from text so profile information and top interface controls cannot cover it.

If the supplied quote is long:
• use additional lines
• adjust line breaks
• reduce the overall type size slightly
• tighten the editorial composition carefully
• increase the internal margins before applying the fisheye warp

Do NOT:
• crop text
• abbreviate text
• remove text
• push text outside the safe zone
• sacrifice readability for oversized typography

Complete text visibility is more important than maximum text size.

The safe-zone restriction applies to the typography. Preserve the extreme sky-heavy subject composition described elsewhere.

==================================================
CAMERA + VINTAGE PHOTO LOOK
==================================================

Use a low handheld selfie perspective from approximately upper-chest height, angled sharply upward toward the person and sky.

Use a slightly wide field of view without distorting the person’s facial identity.

Make the finished image feel like a vintage disposable-camera snapshot mixed with an imperfect older consumer-camera/selfie aesthetic.

It should NOT look like a professional DSLR portrait.

Add subtle:
• fine film grain
• slight color noise
• mild analog softness
• imperfect sharpness
• slightly warm film rendering
• subtle highlight blooming
• occasional crushed shadows
• clipped bright highlights
• slight lens haze
• imperfect exposure
• mild compression/scanned-photo texture
• tiny optical imperfections

Keep these effects realistic and subtle.

NO fake film border.
NO timestamp.
NO fake scratches.
NO heavy sepia.
NO obvious vintage filter.

==================================================
HAIR
==================================================

Give the person a very aesthetic, intentionally effortless:

HALF-UP, HALF-DOWN hairstyle.

This is NOT a full bun.

Approximately the upper half of the hair is gathered into a HIGH, LARGE, VERY MESSY, VERY LOOSE BUN near the crown/top of the head.

The remaining half stays DOWN.

The bun must be:
• large
• loose
• airy
• imperfect
• asymmetrical
• casually twisted
• slightly collapsing
• textured
• messy
• full of escaping pieces

NO slick bun.
NO ballerina bun.
NO neat top knot.
NO perfectly round bun.

Leave LOTS of hair down around and behind the head.

Most importantly, leave MANY loose face-framing pieces.

Include:
• highlighted pieces around both temples
• long pieces beside the sunglasses
• strands falling toward the cheeks
• loose pieces over or near the forehead
• tiny flyaways around the hairline
• escaped strands around the bun

Add a gentle natural breeze.

Several front pieces and flyaways should visibly blow sideways or across the face.

Keep the wind subtle and realistic.

Respect the person’s natural overall hair color while adding flattering dimensional sun-kissed highlights:
• beige
• honey
• warm dimensional pieces
• brighter face-framing highlights

Do not make the highlights stripey.

==================================================
SUNGLASSES
==================================================

Add chunky BUTTER-YELLOW sunglasses.

Use thick pale creamy-yellow acetate frames with dark lenses.

Shape:
• stylish
• vintage-inspired
• slightly oversized
• softly oval/rectangular
• substantial frame

The sunglasses completely cover the eyes.

Do NOT lower them down the nose.

Include realistic reflections and bright sunlight glare in the dark lenses.

The butter-yellow frames should become one of the signature visual elements.

==================================================
EXPRESSION
==================================================

Relaxed closed-mouth expression.

Soft neutral lips.

No teeth.
No exaggerated smile.
No duck face.
No dramatic modeling expression.

The person looks naturally toward the camera through the sunglasses.

==================================================
POINTING HAND
==================================================

The subject raises ONE hand beside their head and POINTS UP AT THE SKY with the index finger.

The pointing gesture should feel casual.

==================================================
ABSOLUTE LEFT-SIDE HAND PLACEMENT
==================================================

THE POINTING HAND MUST APPEAR ON THE LEFT SIDE OF THE FINAL IMAGE FROM THE VIEWER’S PERSPECTIVE.

IMAGE-LEFT IS REQUIRED.

This means:
• the hand appears to the LEFT of the person’s head in the finished image
• the raised index finger appears on the LEFT side of the canvas
• the arm enters from the lower-left area
• the hand remains beside the left side of the visible composition
• the hand must NEVER appear on the viewer’s right side

Use the subject’s RIGHT hand and render it on the VIEWER’S LEFT side of the final image.

Do not mirror the hand placement.

Viewer orientation controls placement.

Regardless of selfie-camera mirroring conventions, the finished image must visibly show the pointing hand on the LEFT side of the image.

Do NOT:
• place the pointing hand on image-right
• place the hand to the right of the person’s head
• mirror the final composition
• switch the hand to the opposite side
• center the hand directly above the person’s face

Leave a small amount of blue-sky breathing room between the hand and the extreme left canvas edge.

Do not crop the pointing finger or any visible portion of the hand.

==================================================
HAND HEIGHT + GESTURE
==================================================

Keep the hand LOW.

The knuckles/base of the hand should sit approximately beside the upper side of the head on the VIEWER’S LEFT.

The pointing fingertip should reach approximately to the height of the messy bun or only slightly above it.

Do NOT extend the hand far upward toward the heart cloud.

There must be lots of blue sky between the fingertip and the cloud.

The BACK OF THE HAND faces the camera.

The palm should NOT face the camera.

The index finger extends upward.

The other fingers naturally curl inward.

The pointing index fingernail must face the camera and be clearly visible.

Hand anatomy must be completely realistic.

Exactly five fingers.
No extra fingers.
No fused fingers.
No malformed joints.
No duplicated fingertips.
No disappearing fingers.
No twisted wrist.

==================================================
NAILS
==================================================

All visible nails are sophisticated ARMY-GREEN CHROME.

Think metallic muted olive/army green with a reflective chrome finish.

NOT emerald.
NOT lime.
NOT teal.
NOT black.

Shape:

SKINNY ALMOND-OVAL.

The nails should be:
• narrow
• elegant
• tapered
• softly pointed
• almond/oval hybrid
• moderately long

NOT square.
NOT coffin.
NOT bulky acrylics.
NOT extreme stilettos.

The pointing index nail is especially visible.

Harsh sunlight should create a metallic highlight across the army-green chrome polish.

==================================================
RINGS
==================================================

Add approximately 3–5 mixed warm GOLD rings to the pointing hand.

Combine:
• chunky gold bands
• thin stacking bands
• subtle textured rings
• small signet-inspired pieces
• delicate gold rings

The styling should feel casually collected and mismatched.

Mix chunky and dainty pieces.

Do not put a ring on every finger.

==================================================
CLOTHING + JEWELRY
==================================================

The person wears a simple BUTTER-YELLOW tank top.

Because of the extreme crop, show very little clothing.

Only a tiny strap/top-edge hint should be visible if necessary.

Do not extend the composition downward just to show clothing.

If necklaces are visible, use subtle layered delicate warm-gold chains.

==================================================
EXTREME SUNLIGHT
==================================================

LIGHTING IS CRITICAL.

Create BRUTALLY HARSH DIRECT LATE-AFTERNOON SUNLIGHT.

The person is essentially FACING INTO THE SUN.

The actual sun remains OUTSIDE THE FRAME.

DO NOT show the sun itself.

Instead, communicate its intensity through extremely strong direct illumination.

The sunlight blasts across:
• forehead
• nose
• cheeks
• lips
• sunglasses
• top of head
• messy bun
• loose highlighted hair
• flyaways
• pointing hand
• gold rings
• chrome nail

The forehead and crown should have partially CLIPPED highlights.

Individual highlighted hairs and flyaways should glow intensely.

The sunglasses should contain bright specular glare/reflections.

The rings should catch tiny gold highlights.

The chrome nail should catch a strong metallic reflection.

This should look almost slightly overexposed because the person is facing directly into intense sunlight.

NO studio lighting.
NO beauty lighting.
NO ring light.
NO soft cloudy illumination.
NO evenly lit AI portrait.

==================================================
SKIN
==================================================

Make the skin look flattering but unmistakably REAL.

Preserve:
• pores
• freckles if present
• natural pigmentation
• fine lines
• realistic texture
• natural tonal variation
• realistic under-eye structure
• subtle shine
• tiny imperfections

Direct sunlight can make the skin glow, but do NOT airbrush it.

No plastic skin.
No waxy skin.
No porcelain smoothing.
No beauty-filter appearance.

==================================================
SKY + CLOUDS
==================================================

Create a vivid, rich, believable BLUE SKY filling almost the entire image.

Include scattered natural white clouds with LOTS of open blue negative space.

Cloud placement should feel random and meteorologically plausible.

Use different sizes and densities.

Do not make the sky look like a digital backdrop.

==================================================
HEART CLOUD
==================================================

Include ONE naturally occurring cloud in the UPPER-MIDDLE portion of the sky that unmistakably RESEMBLES A HEART.

Place it high enough that it remains clearly separated from:
• the pointing fingertip
• the messy bun
• the main typography block

The heart cloud is a key visual element.

It should be recognizable immediately but still look like a REAL CLOUD.

Use:
• imperfect fluffy edges
• uneven heart lobes
• wispy areas
• irregular atmospheric texture
• slight asymmetry

It should look like someone genuinely looked up and thought:

“Wait, that cloud looks like a heart.”

NOT a perfect graphic heart.
NOT clip art.
NOT CGI.
NOT a symbol pasted into the sky.

The pointing finger on the LEFT side of the image visually directs attention toward the heart cloud while remaining far below it.

==================================================
REALISM + IMPERFECTIONS
==================================================

The image should feel spontaneous rather than staged.

Include subtle real-photo imperfections:
• asymmetrical messy bun
• random escaping hairs
• slight wind movement
• a few hairs crossing the face
• imperfect cloud placement
• slightly awkward low selfie angle
• film grain
• minor softness
• clipped sunlight
• mild lens haze
• imperfect exposure
• slightly inconsistent sharpness
• slight motion softness in individual hairs

Do NOT make the image clinically perfect.

It should feel like an imperfect real photograph that happened to turn out exceptionally aesthetic.

==================================================
ON-IMAGE TEXT
==================================================

Add the EXACT text supplied after “TEXT:” at the bottom of this prompt.

Reproduce it WORD-FOR-WORD.

Do not:
• rewrite it
• correct it
• paraphrase it
• add words
• remove words
• repeat words
• change punctuation

Bake the text permanently into the photograph.

==================================================
TEXT STYLE
==================================================

Use warm BUTTER-YELLOW typography that visually coordinates with the sunglasses.

Use an elegant HIGH-FASHION VINTAGE EDITORIAL SERIF.

The typography should have:
• dramatic thick/thin strokes
• narrow elegant letterforms
• oversized editorial proportions
• selective italic words or phrases
• irregular line widths
• varying line sizes
• tightly art-directed composition

==================================================
TRUE TWO-DIMENSIONAL FISHEYE TYPOGRAPHY
==================================================

Apply a TRUE TWO-DIMENSIONAL RADIAL FISHEYE-LENS DISTORTION to the ENTIRE typography composition.

This must look like the complete text block is being viewed through the rounded center of a real ultra-wide fisheye camera lens.

This is NOT ordinary curved text.

Do NOT create only:
• one curved baseline
• a simple upward arch
• a simple downward arch
• a rainbow-shaped layout
• a single horizontal bend
• a single vertical bend
• flat text with curved lines
• identical curves on every line
• text arranged along one arc

The distortion must occur simultaneously across BOTH AXES:

• horizontal curvature
• vertical curvature
• diagonal curvature
• radial center enlargement
• edge compression
• corner recession
• spherical convex projection
• multidirectional optical warping

Treat the complete typography composition as ONE unified rectangular text surface being projected outward through the center of a rounded glass fisheye lens.

The CENTER of the full typography field should appear:
• enlarged
• closer to the viewer
• pushed outward
• visibly magnified
• slightly wider
• slightly taller
• convex
• optically dominant

The OUTER EDGES should appear:
• smaller
• compressed
• pulled backward
• curved away from the viewer
• progressively narrower
• increasingly distorted toward the corners

Create genuine RADIAL distortion extending outward from one central optical point.

The left, right, top, and bottom edges of the full text composition must all bend and recede around the enlarged central area.

The text block must curve in multiple directions at the same time:

• upper lines curve upward and backward
• lower lines curve downward and backward
• left-side letters curve toward and around the left edge
• right-side letters curve toward and around the right edge
• central words bulge forward
• outer words become progressively compressed
• all four corners recede
• diagonal areas bend toward their nearest corners

The baselines must NOT all curve in the same direction.

Each line should follow a different portion of the spherical projection based on its position inside the overall typography field.

Lines above the optical center should bow differently from lines below it.

Letters to the left of the optical center should warp differently from letters to the right.

The upper-left, upper-right, lower-left, and lower-right portions must each visibly recede in their own direction.

Letters near the center may become:
• larger
• wider
• taller
• more open
• slightly stretched outward

Letters near the outer edges may become:
• smaller
• narrower
• compressed
• subtly angled away
• optically pulled toward the surrounding corners

Allow subtle deformation within the letterforms themselves so the typography feels projected through curved glass rather than merely arranged along curved paths.

The complete text composition should resemble:
• typography photographed through a true circular fisheye lens
• a convex glass bubble
• text wrapped over the front of a rounded globe
• a spherical vintage magazine-print distortion
• an ultra-wide lens enlarging the center while compressing every outer edge

The typography must visibly curve:
• left to right
• top to bottom
• diagonally toward all four corners
• radially outward from the central optical point

The distortion must have a clearly enlarged convex center and visibly receding edges on every side.

This is a FULL RADIAL FISHEYE BULGE.

It is NOT a one-direction curve.

It is NOT a simple arch.

It is NOT text placed on a circular path.

It is NOT several lines bending identically.

It is NOT a flat text block with one curved baseline.

The overall effect should feel intentional, editorial, strange, nostalgic, high-fashion, and optically real.

Keep every word readable.

Do not distort individual letters so aggressively that spelling becomes unclear.

The fisheye effect must remain immediately visible while preserving complete readability.

==================================================
FISHEYE SAFE-ZONE CONTROL
==================================================

The complete warped typography must remain entirely inside the defined Instagram and TikTok text safe zone.

This includes:
• every letter
• every serif
• every italic flourish
• every punctuation mark
• every ascender
• every descender
• every enlarged central character
• every compressed edge character
• all warped corners
• the full spherical bulge

If the radial fisheye warp risks crossing the safe-zone boundaries:

1. Start with a smaller unwarped text block.
2. Increase blank internal padding around all four sides.
3. Keep extra padding especially on the right and bottom.
4. Apply the full two-dimensional radial warp.
5. Scale the entire finished warped composition down as one unit until it fits.

Do NOT flatten the fisheye effect to make it fit.

Do NOT convert the warp into a simple curve.

Preserve the genuine two-axis spherical distortion while reducing the overall typography size enough to remain safe.

==================================================
TEXT PLACEMENT
==================================================

Place the typography entirely within the INSTAGRAM + TIKTOK TEXT SAFE ZONE defined above.

Position it primarily across the open middle-sky area.

The typography may occupy a substantial portion of the safe sky area, but it must remain completely inside:

• x = 18% to 68% of the canvas width
• y = 25% to 64% of the canvas height

Keep the typography away from the right-side interface-control area.

Keep it well above the bottom caption and call-to-action area.

Do NOT cover:
• face
• sunglasses
• pointing hand
• pointing finger
• army-green nail
• gold rings

Keep visible blue negative space around the typography.

No text box.
No outline.
No drop shadow.
No glow.
No gradient.
No background panel.

Do not add any text except what I provide.

The entire warped text composition must be fully readable before aesthetic distortion is considered successful.

If necessary:
• reduce the typography size
• use more lines
• tighten the line spacing carefully
• reposition the optical center
• increase the unwarped text margins

Never allow text to be:
• cropped
• cut off
• hidden behind platform controls
• pushed into the right-side UI area
• pushed into the bottom caption area
• pushed into the top profile area
• warped beyond the defined safe-zone boundaries

==================================================
ABSOLUTE DO-NOT LIST
==================================================

DO NOT:
• change the person’s identity
• create a generic attractive AI face
• overly smooth the skin
• use a full bun
• make the bun neat
• remove the hair-down portion
• remove the loose front pieces
• remove the windblown flyaways
• change the butter-yellow sunglasses
• expose the eyes above the sunglasses
• place the pointing hand on the RIGHT side of the image
• mirror the hand onto image-right
• center the pointing hand above the face
• make the hand too high
• extend the arm toward the heart cloud
• show the palm toward the camera
• hide the pointing fingernail
• change the army-green chrome nails
• use square or coffin nails
• remove the gold rings
• create malformed hands
• create extra fingers
• create a perfect graphic heart cloud
• show the actual sun
• use soft lighting
• center the subject vertically
• fill the frame with the person
• add unnecessary torso
• reduce the massive amount of sky
• make the image professionally polished
• make the typography flat and ordinary
• use a simple one-direction curve
• curve every line in the same direction
• create only an arched text layout
• create only a rainbow-shaped text layout
• apply distortion on only the horizontal axis
• apply distortion on only the vertical axis
• keep the center and outer edges the same size
• fake fisheye typography using only curved baselines
• place text on one circular path
• use identical curvature across every line
• omit the radial center enlargement
• omit the edge compression
• omit the four-corner recession
• allow the radial distortion to push letters outside the safe zone
• place any text outside the specified safe zone
• place text in the rightmost 22% of the image
• place text in the bottom 36% of the image
• place text in the top 15% of the image
• crop any letters or punctuation
• add UI elements
• add Instagram or Edits interfaces
• add TikTok interfaces
• add buttons
• add captions
• add usernames
• add logos
• add watermarks
• add borders

==================================================
FINAL RESULT
==================================================

ONE clean vertical 9:16 photograph at 1080 × 1920 pixels.

Exact person from uploaded selfie.
Massive vivid blue sky.
Person confined to bottom.
Half-up half-down highlighted hair.
Huge high loose messy bun.
Lots of front pieces and flyaways.
Gentle wind.
Butter-yellow sunglasses.
Low pointing hand clearly positioned on the VIEWER’S LEFT side.
Hand located to the left of the person’s head.
Back of hand facing camera.
Army-green chrome skinny almond-oval nails.
Mixed chunky and dainty gold rings.
Naturally occurring-looking Heart-shaped cloud high in sky.
Extremely harsh direct sunlight.
Vintage disposable-camera grain and imperfection.
True two-dimensional radial fisheye typography.
Spherical convex center bulge.
Simultaneous horizontal, vertical, and diagonal curvature.
Magnified center.
Compressed outer edges.
Receding corners.
Different curvature across different lines.
All text fully centered and contained inside the shared Instagram and TikTok safe zone.
No text behind platform interface areas.
No cropped letters.
No interface or watermark.

TEXT:
unfortunately for my haters, I will keep posting with the unfazed audacity of a below average man who just caught a mediocre fish

If the first image is close, but not right

Do not start over right away. Ask ChatGPT to fix one thing at a time.

If the face changed, say:

Fix only the identity. Match the uploaded selfie more closely. Keep the pose, sky, hand, hair, sunglasses, light, quote, and layout unchanged.

If the hand lands on the wrong side, say:

Move the pointing hand to the viewer's left side. Keep it low beside the head. Change nothing else.

If the quote is clipped or sits under the app buttons, say:

Scale the complete warped text block down. Keep every letter inside the safe zone. Keep the right 22%, top 15%, and bottom 36% free from text. Change nothing else.

If you see black lens corners, say:

Remove the black fisheye rim and every dark lens corner. Extend the blue sky naturally to all four corners. Change nothing else.

Small fixes work better because they tell ChatGPT what to keep.

If you try this prompt, tag @liz.on.the.web so I can see your version.

Want more full prompts like this? Subscribe to Liz on the Web. I share simple AI ideas you can use right away.

Result? 1M+ views and $6K+ in 7 days.

Random post ideas are too weak for this.

I made Claude study what was already working first.

Then I used it to turn those patterns into scripts, hooks, and a repeatable weekly plan.

If you are trying to grow on Instagram in 2026, you do not need more blank-page panic.

You need a simple prompt stack you can run again and again.

Use this in order.

Step 1: find what is already going viral

Before you ask Claude to write, ask it to study.

You want it to look for patterns.

That means the topics, hooks, formats, and feelings that keep showing up.

Copy this prompt:

Look at the top Reels, TikToks, and Reddit posts in [niche] from the last 30 days. Find the hooks, topics, feelings, and formats that keep showing up. Give me 5 angles to try.

Replace [niche] with your niche.

Use something simple like:

• beginner AI tools

• online business

• social media growth

• fitness for busy moms

• easy meal prep

• real estate tips

Do not make the niche too broad.

Claude needs a clear lane.

Step 2: turn one angle into a short script

Pick the best angle from step 1.

Then ask Claude for one short Reel script.

Short is the point.

Most people write too much.

Copy this prompt:

Write a 30-second Reel script about [topic]. Start with a strong first line, keep each line short, build curiosity, and end with one clear takeaway and a soft CTA.

Replace [topic] with the angle you picked.

If Claude gives you a script that sounds stiff, tell it:

Make this sound more like a smart friend talking. Use shorter lines. Remove any words that sound like marketing copy.

Now you have the first draft.

Do not post it yet.

Step 3: make 10 stronger hooks

The first line does the heavy lifting.

If the first line is boring, the rest of the script does not get a chance.

So make Claude write more hook options before you film or post.

Copy this prompt:

Take this script and write 10 hook options. Use proof, speed, fear, surprise, or a strong opinion. Make each hook clear enough to stop someone from scrolling.

Then pick the clearest hook.

Not the cutest one.

Not the one that sounds clever.

Pick the one a tired person can understand fast.

Step 4: make the whole workflow repeat

One good post is nice.

A weekly system is better.

Ask Claude to turn the idea into a repeatable plan.

Copy this prompt:

Build a simple weekly content flow for [niche]. Give me daily topics, scripts, visual ideas, captions, and hashtags. Keep the steps easy enough to repeat every week.

Now you have a full content flow.

You can use it for Reels, TikToks, carousels, Threads, emails, and Substack notes.

This is how you stop starting from zero.

The simple order

Run it like this:

1. Study what is already working.

2. Pick one strong angle.

3. Write a short script.

4. Make 10 hook options.

5. Build the weekly plan.

6. Repeat next week.

That is the full loop.

The result comes from better inputs.

Claude is much more useful when you make it research first, write second, and plan third.

FREE! Comment PROMPT26

Comment PROMPT26 to steal my full 90k followers in 90 days strategy, PDF guides, step-by-step plans, AI prompts, and exclusive secrets.

Save this prompt stack before your next content batch.

Then run it before you ask AI for one more random post idea.

That is not a small thing.

It is how strangers find where you live. It is how spam gets worse. It is how old data follows you around.

The annoying part is not that removal is impossible.

The annoying part is that it is boring.

One site has a form. One wants an email. One hides the opt-out link. One sends a verify link to your inbox. One removes you, then quietly lists you again later.

Doing this by hand across dozens or hundreds of sites can kill an afternoon.

This is the kind of work AI should do.

Not because AI is a lawyer. It is not.

Not because AI should send private forms without you looking. It should not.

Because this job is mostly search, copy, paste, track, check, and repeat.

So the workflow is simple:

1. Claude finds the people-search and data-broker sites.

2. Claude finds the opt-out page for each one.

3. Claude writes the removal request.

4. Claude opens each page and fills it out.

5. You approve before anything gets sent.

6. You click any email verify links yourself.

7. Claude checks again every week.

This is a free Genius B*tchwork playbook. Take it slow. Use a spare email. Do not paste your Social Security number, bank account numbers, or full ID into any AI chat.

The quick safety note

This is not legal advice.

Privacy rights depend on where you live and which site has your data.

In California, the CCPA gives people the right to ask many covered businesses to delete personal information, with some exceptions. California also has a data-broker system called DROP. It lets Californians send one delete request to registered data brokers, and data brokers must process those requests under the Delete Act schedule.

In the EU, the GDPR has a right to erasure in some cases.

In the US, the FTC says people-search sites usually have opt-out steps, and that after you opt out, the site should stop selling the existing info it has about you. The FTC also warns that your info can come back if it gets pulled again from public records.

So here is the plain version:

Use the law where it applies. Use the site's opt-out path where that is the faster route. Keep proof. Recheck later.

What you need before you start

You need Claude Desktop with Cowork.

You need a folder Claude can use as the workspace.

You need a spare email address. Some sites send a verify link. Do not make your main inbox the place where all these sites start sending mail.

You need the basic details the sites may ask for:

• Your full name.

• City and state.

• The spare email.

• The old address or phone number only when the site already shows it and needs it to match the record.

You also need one rule:

Claude fills. You approve. You click email links yourself.

Step 1: Find every site selling or listing you

Open Claude Desktop.

Open Cowork.

Open a folder you can use for this project.

Then paste this.

Search the internet for everything you can find about me. Find every data-broker and people-search site selling, listing, or sharing my personal information. Put it in a table: site name, exactly what of mine it shows (address, phone, relatives), the direct opt-out link, and the next step required.

Name: [your full name] · City: [city, state] · Email: [your email]

Replace the brackets first.

Claude should return a list. You want site name, what it shows, the opt-out link, and what to do next.

Do not panic if the list is ugly.

That is the point.

The mess is now in a table instead of in your head.

Step 2: Make Claude write the removal requests

Now send the second prompt in the same chat.

For every site in that table, write the exact removal request that site requires - filled-out form, email text, or letter. Cite the relevant privacy law (CCPA, GDPR, or state law). Keep my details consistent across all of them, and request written confirmation within 30 days.

Claude should now turn the table into a removal packet.

For a form, it should list what goes in each field.

For an email, it should write the email.

For a letter, it should write the letter.

If a site does not clearly fall under a law where you live, Claude should still use the site's own opt-out policy and ask for removal.

Step 3: Have Claude fill the forms, with approval

Now send Prompt 3.

Turn on Cowork first. Then: go to every site in the table, open the opt-out page, fill in my details, and ask for my approval before submitting each one. If a site needs email verification, tell me which inbox to check. Mark each row done, skipped, or needs-me.

This is where you slow down.

Do not let AI submit forms blind.

Have it fill the page. Read the page. Approve the send.

If a site asks you to click a verify link by email, that part is yours.

If a site asks for too much private info, pause.

Some sites ask for ID. Some ask for a current address. Some ask for a phone number. Make Claude explain why the site asks before you give anything extra.

Step 4: Make the cleanup repeat every week

The bad part about data brokers is that removal is not always forever.

Your record can come back.

So the weekly check matters.

Send Prompt 4.

Set up a scheduled task that runs every week. Recheck every site, resubmit any request that was ignored, and re-remove any site that re-listed me - until I no longer appear anywhere. Escalate any site that ignores 2+ requests with a complaint drafted to the FTC or my state AG. Send me a weekly summary of what's left.

If Claude calls it a routine instead of a scheduled task, that is fine. The point is the same.

You want the task to recheck the same list, update the table, and tell you what is left.

What to put in the tracking table

Have Claude keep a file in the workspace called:

data-removal-tracker.md

Tell it to track:

1. Site name.

2. Profile URL.

3. What data it showed.

4. Opt-out link.

5. Request sent date.

6. Email verify needed.

7. Status.

8. Next check date.

9. Notes.

Status should be simple:

• Found.

• Request sent.

• Needs email click.

• Removed.

• Re-listed.

• Escalate.

• Skip.

This keeps you from doing the same site twice.

It also gives you proof if a site ignores you.

What AI cannot do for you

AI cannot make a legal right appear if your state or country does not give you one.

AI cannot promise every site will remove every record.

AI cannot erase public records at the source.

AI should not click email verify links for you.

AI should not upload your ID without you reading the request first.

AI should not make legal threats it cannot back up.

AI can do the boring part:

Find the site. Find the form. Write the request. Fill the fields. Track the reply. Recheck next week.

That is enough to make this worth doing.

The exact prompts

Run these in order, in the same Claude Cowork chat.

Prompt 1 - Find every site

Search the internet for everything you can find about me. Find every data-broker and people-search site selling, listing, or sharing my personal information. Put it in a table: site name, exactly what of mine it shows (address, phone, relatives), the direct opt-out link, and the next step required.

Name: [your full name] · City: [city, state] · Email: [your email]

Prompt 2 - Write the requests

For every site in that table, write the exact removal request that site requires - filled-out form, email text, or letter. Cite the relevant privacy law (CCPA, GDPR, or state law). Keep my details consistent across all of them, and request written confirmation within 30 days.

Prompt 3 - Submit them for me

Turn on Cowork first. Then: go to every site in the table, open the opt-out page, fill in my details, and ask for my approval before submitting each one. If a site needs email verification, tell me which inbox to check. Mark each row done, skipped, or needs-me.

Prompt 4 - Keep me gone

Set up a scheduled task that runs every week. Recheck every site, resubmit any request that was ignored, and re-remove any site that re-listed me - until I no longer appear anywhere. Escalate any site that ignores 2+ requests with a complaint drafted to the FTC or my state AG. Send me a weekly summary of what's left.

Run Prompt 1 tonight.

Worst case, you find nothing.

Best case, you find the sites selling your data and make them deal with you.

Sources

• FTC guide to people-search sites

• California CCPA rights

• California data broker DROP rules

• California privacy complaint form

• Report bad business practices to the FTC

• EDPB guide to GDPR rights

• Claude Desktop install guide

• Claude routines docs

They add Gmail.

They add Google Drive.

They add Slack, Notion, GitHub, Calendar, or whatever tool they use all day.

Then they start asking Claude to work inside real accounts with real files, real messages, and real people on the other end.

That can be very useful.

It can also get messy if you never check the permission setting first.

The setting is connector permissions.

If you use Claude with connected apps, check it before you hand Claude a real work task.

What connector permissions do

Connectors let Claude use outside tools.

That means Claude can pull context from the places your work already lives.

Examples:

• Read files in Google Drive

• Search emails

• Draft a message

• Check calendar events

• Look up Slack or team context

• Work with issues, docs, records, or other tools

This is what makes Claude feel less like a blank chat box.

But a connected tool is not always just a reading tool.

Some tools can also change things.

Depending on the connector and your account, Claude may be able to draft, create, update, move, share, or delete items. Anthropic says connector access depends on the service and your source account permissions, and that Claude cannot get more access than you already have in the original tool.

That last part matters.

Claude is not getting a magic admin key.

But it may be using your own access.

That means your permission choices matter.

The three settings to look for

Anthropic's current connector docs use three permission choices:

Your account may show this at the workspace, connector, or tool level depending on your plan and setup.

If you do not see the full permission screen, use the same idea when Claude asks for approval in chat.

Do not click "allow always" on a risky action just to save one second.

My rule

Here is the simple rule I would use:

Reading can be fast.

Changing things should be slow.

Deleting things should usually be blocked.

That means:

• Search inbox: Always allow

• Read Drive files: Always allow

• Draft an email: Always allow or Needs approval

• Send a message: Needs approval

• Share a doc: Needs approval

• Move files: Needs approval

• Delete files: Blocked

• Spend money: Blocked or Needs approval

This keeps Claude useful.

It also keeps you in charge of the part that is hard to undo.

The easy mistake

The mistake is treating every connector the same.

Gmail is not the same as Drive.

Drive is not the same as Calendar.

Slack is not the same as GitHub.

Some connectors are mostly about reading. Some can create or change records. Some can touch team spaces. Some have different limits based on your plan.

For example, Anthropic's Google Workspace help page says Claude can search and read Gmail, and create drafts with your approval, but cannot send emails for you through the current Gmail connector. That is safer than a full send permission.

But Google Calendar can create, update, and delete events.

Other connectors can take different actions.

So the real habit is this:

Check the action list for each connector.

Do not assume.

The 60 second setup

Open Claude.

Go to your connector settings.

Pick one connected tool.

Look for tool permissions.

Then sort the actions into three buckets.

Bucket 1: safe enough to run fast

Use Always allow for low-risk actions.

Good examples:

• Search files

• Read docs

• Read public or shared context you chose

• Find a calendar event

• Pull up a message thread

These actions help Claude stop guessing.

They do not send anything or delete anything.

Bucket 2: useful, but needs your click

Use Needs approval for actions that change the outside world.

Good examples:

• Send a Slack message

• Create or update a calendar event

• Share a file

• Move a document

• Update a project record

• Create a task in a team tool

The approval step is not there to slow you down for no reason.

It gives you one clean moment to check what Claude is about to do.

Bucket 3: off limits

Use Blocked for anything you do not want AI touching.

Good examples:

• Delete files

• Delete emails

• Remove records

• Spend money

• Buy anything

• Change billing

• Change account access

If the action makes your stomach drop, block it.

You can always loosen the rule later if you have a real reason.

The permission map I would use

Copy this.

Claude connector permission map

READ
Search inbox: Always allow
Read docs: Always allow
Read calendar: Always allow
Read project records: Always allow

DRAFT
Draft email: Always allow or Needs approval
Draft Slack message: Always allow or Needs approval
Draft doc: Always allow
Create a private task draft: Always allow or Needs approval

ACT
Send email: Needs approval
Post Slack message: Needs approval
Share file: Needs approval
Move file: Needs approval
Update record: Needs approval
Create calendar event: Needs approval

BLOCK
Delete file: Blocked
Delete email: Blocked
Remove records: Blocked
Make purchase: Blocked
Change billing: Blocked
Change account access: Blocked

This is not about fear.

It is about making AI useful in the right lane.

How I would use this at work

Here is a safe first workflow.

Connect Google Drive.

Allow Claude to read the folder you actually need.

Ask:

Look at the files in this project folder.
Give me the 5 things I need to finish this week.
Tell me which file each item came from.
Do not edit, move, share, or delete anything.

That is a good use of a connector.

Claude gets context.

You get a clearer plan.

Nothing gets changed.

Then, when you are ready to let Claude draft something, ask:

Draft the reply I should send.
Do not send it.
Show me the final text first.
Tell me if anything in the draft came from a connected file or message.

That is the right order.

Read first.

Draft second.

Approve before action.

Quick checklist

Before you use a new Claude connector for real work, check this:

• Read actions are set on purpose.

• Search actions are set on purpose.

• Draft actions are set on purpose.

• Sending is set to Needs approval.

• Sharing is set to Needs approval.

• Moving is set to Needs approval.

• Deleting is set to Blocked unless you have a very clear reason.

• Buying or billing actions are set to Blocked.

• You review the permissions again when the connector adds new tools.

One minute here can save you from a lot of cleanup later.

Read next

If you are setting up Claude for real work, read these next:

• Claude AI Beginner Guide 2026: 5 Settings to Fix Before You Start Prompting

• ChatGPT Work vs Claude Cowork: Which AI Agent Should You Use for Business in 2026?

• 10 Free AI Certifications for Your Resume in 2026

• Use Fable 5 Before July 12 Like a Paid Consultant

• 6 AI Content Skills That Can Plan Your Week in 30 Minutes

I am also building more beginner-friendly Claude guides for setup, Projects, connectors, and work agents.

Comment with the Claude setting you want me to cover next.

Want my Claude setup prompts?

This guide is free.

Paid subscribers get the deeper prompt packs, workflows, checklists, and AI work systems I use to turn these settings into real output.

Subscribe here: https://lizontheweb.substack.com/subscribe

Sources

• Use connectors to extend Claude's capabilities

• Use Google Workspace connectors

• Get started with custom connectors using remote MCP

That is the wrong test.

Claude gets much better when you set it up like a work tool instead of treating it like a magic search box.

If you are new to Claude, do these five things first.

1. Install the desktop app

Start with the desktop app.

Claude works in the browser, but the desktop app is the better home base if you want Claude to feel like part of your actual workday.

Anthropic's download page lists Claude apps for Mac, Windows, iOS, and Android, plus extensions for tools like Chrome, Excel, PowerPoint, and Slack.

Get it here:

https://claude.com/download

Why this matters:

• It is easier to return to your work.

• You can keep Claude close while using other apps.

• You can pair desktop and mobile workflows.

• You are less likely to treat Claude like a one-off tab you forget.

Simple setup move:

Pin Claude to your dock or taskbar.

If you want to actually use AI more, reduce the friction.

2. Fill out your profile instructions

Claude has personalization features, including profile instructions, project instructions, and styles.

That matters because the best AI setup is not a giant prompt you paste every time.

It is a set of standing preferences Claude can remember.

Add simple instructions like this:

I prefer clear, direct answers.
Use plain English.
Ask before making risky assumptions.
For work tasks, give me the next action first.
For writing tasks, keep the tone practical and human.
When comparing options, give me a short recommendation and why.

Do not make this too fancy.

The goal is not to impress Claude.

The goal is to remove the same correction you keep typing over and over.

Good profile instructions tell Claude:

• how you like answers structured

• what tone you want

• what your work is

• what you hate wasting time on

• when Claude should ask before continuing

If Claude keeps doing something annoying, add one line to your instructions.

That is how you train your setup.

3. Connect the tools you actually use

Connectors are where Claude starts feeling less like a chatbot and more like a work assistant.

Anthropic says Claude supports connectors for tools like Gmail, Google Drive, Google Calendar, GitHub, Microsoft 365, and Slack depending on your plan and workspace settings.

Anthropic's connector help docs also say custom connectors are available on Claude, Cowork, and Claude Desktop, with free users limited to one custom connector.

Start small.

Do not connect every tool because it sounds powerful.

Connect the one place where your work already lives.

Good first connector choices:

• Google Drive if your docs live there

• Gmail if your work starts in email

• Google Calendar if your day is schedule-heavy

• Notion if your knowledge base lives there

• Slack if your team context lives there

Beginner rule:

Connect one tool, then give Claude one real job.

Example:

Look at my recent project notes and help me make a clean Monday priority list. Show me what you used, then give me the list.

The point is not to show off integrations.

The point is to give Claude enough context to stop guessing.

4. Learn the model picker

Do not ignore the model picker.

Anthropic's current model docs list Claude Fable 5, Claude Opus 4.8, Claude Sonnet 5, and Claude Haiku 4.5.

In plain English:

• Use the strongest model available to you for strategy, hard writing, research, deep analysis, and messy work.

• Use the fast model for quick cleanup, summaries, formatting, and low-risk drafts.

• Use the balanced model when you need solid thinking without waiting too long.

Model names change.

The habit matters more than memorizing the names.

Before you start a task, ask:

Is this task high-risk, hard, or messy?

If yes, use the stronger model.

If no, use the faster model.

That one habit will save you from using a lightweight model for a job that needed deeper thinking.

5. Create one Project

Projects are where Claude gets organized.

Anthropic says Projects are self-contained workspaces with their own chat histories and knowledge bases. The help center also says Projects are available to all users, including free users, with free users limited to five Projects.

That means you do not need to wait until you are advanced.

Make one Project now.

Good first Projects:

• Job Search

• My Business

• Content Ideas

• Client Work

• Home Admin

• Research

• Course Notes

Inside the Project, add:

• a short description of what the Project is for

• any key files Claude should know

• rules for how you want Claude to help

• a few examples of good output

Simple Project instruction:

This Project is for my job search.
Help me tailor resumes, write cover letters, prep for interviews, and track applications.
Keep everything honest. Do not invent experience. Make my real experience sound clearer and stronger.

That is already better than starting every chat from zero.

The first real prompt to run

After you do the five setup steps, run this:

I am new to Claude. Help me set up one repeatable workflow for [your goal].

Ask me up to five questions first.
Then give me:
1. the workflow
2. what I should put in a Project
3. which connected tool would help most
4. the first prompt I should use
5. the common mistake to avoid

Replace [your goal] with something real:

• planning content

• applying for jobs

• organizing client work

• managing school assignments

• turning notes into tasks

• cleaning up research

• writing weekly reports

Claude is much more useful when you give it a real job with real context.

The mistakes I would avoid

Do not start with a giant prompt library.

Start with one workflow.

Do not connect every app on day one.

Connect the app where your work actually lives.

Do not use the fastest model for high-stakes work.

Use the stronger model when quality matters.

Do not keep making new chats for the same topic.

Use Projects so Claude has context.

Do not accept the first answer blindly.

Ask Claude to check its own work, show assumptions, and tell you what it is unsure about.

My exact Claude setup checklist

Here is the short version:

1. Download Claude Desktop.

2. Add profile instructions.

3. Connect one real work tool.

4. Learn when to switch models.

5. Create one Project.

6. Run one real workflow.

7. Save the prompt that worked.

That is enough to stop dabbling and start using Claude like a work system.

Want my better Claude workflows?

This guide is free.

Paid subscribers get my AI work systems, Claude project templates, reusable prompts, content workflows, and the behind-the-scenes ways I use AI to turn scattered work into finished assets.

Subscribe here: https://www.lizontheweb.com/subscribe

Sources

• Download Claude

• Understanding Claude's personalization features

• Use connectors to extend Claude's capabilities

• Use Google Workspace connectors

• Claude models overview

• What are Projects?