Hi!

I have a quick question. Please take a second to answer, it’ll help me plan future content :)

Many of you know me from Medium.com. I started writing there in 2021, grew to 149k followers, and used that to build this Substack and two YouTube channels (around 50k subscribers each). So I have a lot of experience in writing and AI.

I’m thinking about sharing these two things with you:

1. My complete Claude writing system: All the files to make Claude write like you in posts, emails, reports, presentations, etc. No guides or tutorials. You just download it and start using it right away.

2. Guides on using AI for writing: How to use my Claude system to write more without sacrificing quality, repurpose content into other formats, etc. Plus everything I’ve learned in these 5 years as a writer to build an audience and make money.

Reply to this email or leave a comment: are you interested in 1, 2, both, or none? I’ll read every single reply. Thanks in advance!

You can think of today’s post like a mix of 1 & 2. Hope you like it!

In 2023, many people were against AI writing.

In 2026, many writers are using AI.

Some writers I follow have admitted they use AI for writing. Others haven’t mentioned it publicly, but they’ve suspiciously increased the number of posts they publish per week over the past few years.

Heck, Substack’s top newsletter in education is called “Write with AI.“

I’m not against AI.

I’m against using AI for bad writing.

If you use AI to enhance your writing and produce high-quality work more frequently, your readers will be happy.

The problem is that most people use AI to produce bad writing.

The fix isn’t a better prompt. It’s a file.

One file that teaches AI exactly how you write, what you never say, and what makes your writing yours. You create it once. You reuse it every time.

In this guide, you will:

• Learn why most produce AI slop

• Get the exact prompt that captures your writing voice in a single file

• Learn why and how to use this file in Claude

• Download my personal writing-voice file as a working example

Subscribe now

Why most produce AI slop

Since ChatGPT’s release, we’ve been obsessed with the em-dash and a few words that make your writing sound AI.

Here’s my (ironic) response to this:

and this

My advice: Don’t mind AI words.

It’s fine if you have a few AI words in your vocabulary.

What’s not fine is to produce content under your name that doesn’t sound like you.

By default, AI doesn’t know your voice. It doesn’t know you hate the word “utilize.” It doesn’t know you always open with a personal story, or that you’d never use a semicolon in a million years.

So it defaults to the safest, most average version of whatever you asked for.

That’s not AI’s fault. You gave it a vague instruction in your prompts and expected a specific result.

The file that fixes this

The fix is an .md file (a simple text file with some formatting). The one we’ll create contains everything Claude needs to write like you.

Not “write well.” Write like you.

Here’s a summary of what goes in it:

• Words and phrases you’d never use

• Sentence patterns you default to

• How you open and close pieces

• Your formatting instincts (short paragraphs? lists?)

• What your writing sounds like at its best

• Writers you admire and what you’d steal from each one

• Positions you’d never take

• and more!

The insight that makes this work: most of a good voice profile is about what you reject. Not “I like direct writing,” but “I’d never use semicolons because they make my writing sound like a college essay.”

That’s the kind of specificity Claude can actually use.

How to create your voice file

This guide is for paid subscribers only. Every week I publish a free and paid guide like this one. To get full access, consider becoming a paid subscriber today 👇

Subscribe now

Read more

For a long time, the way I used Claude Cowork was simple.

Ask for something. Get a result. Move on.

That works great when the task is a one-off, like a quick email, a summary, a draft you need today and forget tomorrow.

But it starts to fall short when the work doesn’t end in a single session. When you need to come back tomorrow or next week, and the AI ha…

Read more

This one is mind-blowing!

It’s called “Computer Use.” It lets Claude control your entire computer. This means controlling your keyboard and mouse in any app that you allow.

It’s only available for the macOS Claude app, BUT after testing it since its release, I think Windows users can still benefit from a simplified version of it (keep reading and you’ll understand what I mean).

Claude is releasing updates like crazy lately! I spend a lot of time testing and digging into each one so you don't have to. If it's useful, it's in my guides. If it's not, I save you the time

Click here to read the best Claude guides

If you like this Claude series, support me by becoming a paid subscriber :) 👇

Subscribe now

Computer Use is powerful, but Claude tries to avoid it. Here’s why

After doing dozens of tests, I’ve noticed that Cowork doesn’t always use computer use’s full potential, and there’s a reason for that.

Even if you prompt Cowork “use computer use for this task,“ Claude follows this order to execute tasks:

1. Connectors: If a connector is available (Gmail, Google Drive, etc), Claude uses it. This is the fastest and most reliable path.

2. Claude in Chrome: When there isn’t a connector for the tool you need, Claude will navigate the Chrome browser to work on your task

3. Screen interaction: If options 1 and 2 won’t help, Claude uses computer use to interact directly with your screen: clicking, typing, and navigating your desktop apps.

Claude prioritizes the fastest method (1 and 2) to avoid longer sessions and mistakes.

In my use-cases, you’ll notice that Claude avoids as much as possible using screen interaction (aka computer use). Instead, it uses connectors or Claude in Chrome. Is that bad? Not at all, you’re wasting fewer tokens and getting the job done anyway.

That’s why I think Windows users without the computer use feature can still benefit from the use cases we’ll see later (Claude mostly uses Claude in Chrome).

First, let’s see how to enable Claude’s computer use.

How to enable Computer Use

First, download the Claude desktop app: claude.com/download

Open the app → Select Cowork → Open the sidebar:

1. Click on your name

2. Settings

3. General (inside “desktop app” section)

4. Toggle on “computer use“

How to know you’re actually using computer use

The best way to trigger computer use is to make Cowork work with local files or apps. Here are some ideas:

• Create an analysis using local files and then create a formatted report in Excel

• Navigate apps that don’t have connectors (tools your team uses, macOS apps, etc)

• Fill in a spreadsheet with data from multiple sources, format it, and save it

Here’s a test you can do to see what your computer looks like when it’s controlled using Claude’s computer use.

Go to Cowork and paste in this prompt:

Use computer use for this task:

Open the calculator app to calculate 30x200 and then create an apple note to store the result

This might seem silly, but, believe me, sometimes it seems that this new feature is being used, but it’s not! If you see the orange outline on the whole screen, Claude is controlling your entire computer with computer use. If you see the outlines only in the Chrome window, Claude in Chrome is being used.

These are only technicalities. At the end of the day, we care to get things done without wasting too many tokens. That’s what we’ll do in the following use cases.

Subscribe now

My Best Use Cases

Computer use helped me automate my busywork. Here are my best use cases that you can adapt to your workflow.

#1 Automate LinkedIn responses (“Comment X, Get Y”)

If you use LinkedIn or Instagram, you probably came across the “Comment X and I’ll send you Y.“ X is a keyword that triggers an automation. Instagram has third parties that make this easy. However, as far as I know, LinkedIn doesn’t have any.

With the prompt below, I could automate this:

Use “Computer Use” to control my computer and do this task:

Open the Excel file “LinkedIn DMs.” Find today’s row, copy the message, and open the LinkedIn post. Then open the comments section and send a DM to everyone who commented, using the keyword. After you successfully sent the DM reply “Sent!” to their comment

Why do content creators use this formula? Platforms like LinkedIn penalize posts with links, so this is the best way to share links without hurting views.

By the way, if you want to connect with me on LinkedIn, click here: https://www.linkedin.com/in/thepycoach/

I accept everyone :)

#2 Accept connections

Every day, I get new connection requests on LinkedIn. I manually accept each of them because there’s no native LinkedIn way to automatically accept requests.

Thanks to the prompt below, I can get my connection request inbox down to zero:

use “computer use” to accept all my connection requests on linkedin

Before this test, I was manually accepting a lot of requests. Luckily, a new one came in right when I needed it (lucky me!)

One less task in my daily routine!

#3 Create a customized short link after publishing a new guide

Every time I publish a new guide, I manually create a Bitly shortlink to share it on other platforms. I use the original link slug to create a customized shortlink.

This prompt automated this process:

Use “computer use” for this:

I have a Substack newsletter at artificialcorner.com where I publish AI guides every Wednesday and Sunday. At the end of each Wednesday/Sunday I want you to go to my bitly.com account and create a customized bitly link using the same URL slug used in the original guide (if it’s taken come up with something similar)

Note: This week I published on a Tuesday (it was an exception)

#4 Create an Excel table with web data

To quickly keep track of all the links I create with Bitly, I need an Excel table with these columns: Guide name, Original Link, Bitly Link.

The prompt below extracts the data I need from Bitly and puts it into Excel.

Use “Computer Use” for this:

Build a table in Excel with these columns: Guide name, Original Link, Bitly Link. All the info is in bitly. Only consider the links from my site artificialcorner.com

#5 Update an Excel table

If I need to update a table with fresh data, I ask Claude to visit a website, extract the information I need, and update the table for me.

Here’s the prompt I use:

Use “computer use” for this:

Go to bitly.com and add the latest link created to my dataset Bitly_Links_ArtificialCorner.xlsx located in Download -> Bitly. Open the Excel file and add the data in a new row. The data needed is Guide Name, Original Link and Bitly Link

Final tips to get the most out of computer use

• You can turn a chat into a scheduled task, so it runs on any day/hour you want. Learn how here: artificialcorner.com/p/claude-automates

• You can use Dispatch to trigger computer use from your phone. Check out my guide: artificialcorner.com/p/cowork-dispatch

• To prevent Claude from using some apps, do this: Settings → General → Denied apps → Add app

• Close anything you don’t want Claude to see. When Claude uses your computer, it takes screenshots to navigate your screen

• Subscribe to this newsletter so you never miss guides like this, delivered straight to your inbox 👇

Subscribe now

Click here to read my best Claude guides

The tool wasn’t broken. My workspace was.

Most people open Cowork and do this: open a folder → write a prompt → hope for the best

Sometimes it works. A lot of times, it doesn’t.

And that’s not Cowork’s fault. Cowork is stepping into your workspace with zero context, no rules, and no clear way of working.

So it starts improvising.

It doesn’t know what tone to use.
It doesn’t know what it should never assume.
It doesn’t know what files it must always read.

That’s why so many people try it and walk away thinking it feels like just another chat … just with file access.

In this guide, we’ll fix that. I’ll show you a simple system that gets the most out of Cowork.

By the end of this guide, you’ll have:

• A system that gives Cowork the context it needs to stop improvising

• How to use this system with 3 real-world examples

• How to prompt this system (spoiler: you’ll need short prompts — the system does the heavy lifting)

No coding. No complex setup. Just a structure that works.

Click here to read my best Claude guides

This is the best Cowork guide I’ve written so far. I almost made it paid, but chose to keep it free because it gives you everything you need to get the most out of Cowork.

From here on, I’ll be publishing one paid guide and one free guide each week. Become a paid subscriber today so you never miss the full series! 👇

Subscribe now

Cowork doesn’t need more prompts. It needs a system

If you come from the chat world, this makes sense: write a better prompt, get a better result.

In Claude Cowork, that is no longer the main piece. The workspace is.

Because Cowork does not just work with what you type into a text box. It works with what it can read inside the folder you give it access to:

- your project files
- your working rules
- your context
- your preferred way of communicating
- your prior drafts
- your examples of success

You need to turn the way you work into a small, reusable system.

Here’s an overview of the system we’ll build:

That jump looks small. It isn’t.

Because the moment your way of working lives in files, Cowork stops starting from zero every time. When that happens, the results stop feeling generic and start feeling like “this thing already knows how I work.”

Quick Cowork Recap: Installation

To use Cowork, you need to download Claude for desktop:

Go to claude.com/downloads → download Claude Desktop → install it → sign in

Once it's installed, select Cowork. That's it. You're in.

Quick setup: Add Global Instructions

Global instructions is where you set the default behavior Cowork should bring into every session. Do this:

Open profile menu → Settings → Cowork → Add your Global Instructions → Save

Copy these global instructions:

You are my working partner inside this workspace.

Start by reading:
1. context/HOW-I-WORK.md
2. system/SYSTEM-RULES.md
3. the relevant folder inside projects/

Do not execute immediately if the goal, audience, or output format is unclear.
Ask questions first when missing context could weaken the result.

Do not invent facts, decisions, dates, numbers, or approvals.
Make uncertainty visible.

Use the workspace files as the main source of truth.
Save final outputs in outputs/.

This gives Cowork a reliable starting behavior before the real context comes from the workspace.

The simple system I’d actually start with

I wanted to strip this down as much as possible.

I wouldn’t start with ten folders.
I wouldn’t start with an entire empty office.
I’d start with four. That’s it.

YOUR-WORKSPACE/
├── system/
├── context/
├── projects/
└── outputs/

On your computer, it looks like this:

That’s already enough to get started.

• system/ → stores the base rules (stays fairly stable)

• context/ → stores the stable context (stays fairly stable)

• projects/ → stores the active work (it changes by project)

• outputs/ → stores what Cowork produces

A small structure that can grow without turning into a mess.

But a folder by itself doesn’t tell Cowork much. What makes the workspace useful are the files that explain Cowork how to work.

The files that bring the structure to life

Folders give Cowork a place to work. Files tell it how to work.

That’s why I’d start with just two:

1. context/HOW-I-WORK.md

2. system/SYSTEM-RULES.md

HOW-I-WORK.md captures who you are inside this workspace: how you work, how you think, how you communicate, and what kind of work you consider good.

SYSTEM-RULES.md defines how Cowork should behave: when it should ask before acting, what it shouldn’t assume, how to handle ambiguity, and what a strong result looks like.

What should live inside HOW-I-WORK.md? A simple version of this file could include:

• what kind of work you do

• who you do it for

• how you think

• how you communicate

• what feels natural in your writing

• what you dislike

• which sources carry more weight when there’s a conflict

• real examples when useful

What should live inside SYSTEM-RULES.md? A simple version of this file could include:

• core operating principles

• when Cowork should ask before acting

• how it should behave during execution

• how to handle ambiguity

• what counts as a strong output

• clear boundaries

• a final review before delivery

The structure would look like this:

Up to this point, everything sounds simple.

Here’s the most important part: I wouldn’t write those files from scratch. Because the moment you do, you start guessing.

You describe the version of yourself you wish were true — not the version that actually shows up in day-to-day work.

That’s why I prefer a different approach: build both files through an interview.

A good interview helps extract things most people wouldn’t write well on their own:

• how they actually work

• how they actually write

• what they value in a strong deliverable

• what makes them distrust a weak one

• what Cowork should never assume

• what kind of help is genuinely useful to them

Then that interview generates the two files that support the entire system.

Copy the prompt below to generate HOW-I-WORK.md and SYSTEM-RULES.md

You’ll be asked about 50 questions, one at a time. Answer each one honestly and in detail for the best results.

You are my workspace interviewer.

Your job is to extract enough high-quality signal to help build 2 things for Claude Cowork:

1. context/HOW-I-WORK.md
2. system/SYSTEM-RULES.md

Your goal is not to interview me forever.
Your goal is to gather enough specific, real, usable information to generate those files well.

## Interview objective

You are helping define how Claude Cowork should work with me inside a reusable workspace.

That means you need to understand:

- what kind of work I do
- what I usually create
- who I create it for
- how I think
- how I write
- what I value in strong outputs
- what I dislike in weak outputs
- how I want Claude Cowork to behave
- what Claude should never assume
- when Claude should ask questions before executing

## Interview structure

You will conduct the interview in 3 blocks:

### Block A - Work context

Use this block to understand:

- what I do
- what kinds of deliverables I create
- who my audience is
- what those people value
- what sources matter most
- what this workspace should help me do

### Block B - Voice, style, and thinking

Use this block to understand:

- how I sound at my best
- how I sound when I am writing lazily or on autopilot
- the words, phrases, and structures that feel natural to me
- what I hate in writing
- how I organize ideas
- how I think through problems
- what I tend to notice, distrust, or emphasize
- examples of real writing that sound like me

### Block C - System behavior and rules

Use this block to understand:

- how I want Claude Cowork to behave
- when Claude should ask before acting
- what makes an output strong or weak for me
- what should never be assumed, invented, or finalized too quickly
- how uncertainty should be handled
- how conflicting information should be handled
- what kind of help is actually useful to me

## Interview rules

1. Ask one question at a time.  
Wait for my response before moving on.

2. Do not ask a fixed number of questions.  
Ask only as many as needed to build the files well, up to a maximum of 50 questions.

3. Push back on vague answers.  
If I say something generic, ask for a specific example.

4. Ask for real evidence.  
Whenever useful, ask for:
- a real paragraph I wrote
- a real sentence I like
- a real sentence I would never publish
- a real example of a good or bad output
- a real example of a decision, document, or deliverable

5. Call out contradictions.  
If I say something that conflicts with an earlier answer, point it out and ask me to resolve it.

6. Distinguish reality from aspiration.  
If I describe the person I wish I were instead of how I actually work, write, or decide, call it out and ask for proof.

7. Go deeper only when useful.  
If an answer contains strong signal, follow it.  
If not, move on.

8. Do not flatter me.  
Do not say “great point,” “interesting,” or similar filler.  
Just ask the next question or push deeper.

9. Optimize for usable specificity.  
Do not collect abstract opinions if they will not help build the files.

10. Stop when you have enough signal.
Do not drag the interview longer than needed. Stop earlier if you already have enough strong signal to generate both files well.

## What you are building toward

At the end of the interview, you will generate 2 outputs:

### Output 1

context/HOW-I-WORK.md

This file should include:

- what kind of work I do
- what I create
- who I create for
- what my audience values
- how I think
- how I communicate
- what feels like me
- what I dislike
- what sources carry more weight
- real examples when useful

### Output 2

system/SYSTEM-RULES.md

This file should include:

- core operating principles
- when Claude should ask before executing
- how Claude should behave during execution
- how to handle ambiguity
- how to handle conflicting information
- what a strong output looks like
- what weak outputs get wrong
- collaboration preferences
- hard boundaries
- a final pre-delivery check

## Output requirements after the interview ends

When you judge that the interview is complete, generate both outputs in this exact format:

[context/HOW-I-WORK.md]
...

[system/SYSTEM-RULES.md]
...

After that, add this section only if needed:

[Missing or weak signal]
...

Do not generate the files before the interview is complete.

## How to begin

Start with Block A.
Ask me the first question.

Why does this approach actually work?

You’re extracting real info to build two files that will be reused again and again.

That’s the shift that matters.

No more files just for the sake of having more files. No more long prompts to compensate for missing context. Just a minimal structure, two well-built files, and a smarter way to give Cowork the context it needs.

Subscribe now

How to apply this method: 3 real-world examples

The advantage of this structure is that it doesn’t depend on a single task.

It works for an article, report, presentation, research project, weekly brief, proposal, document synthesis, etc

The logic is always the same:

project folder + 2 core files + short prompt

That’s the pattern.

You don’t need to reinvent how you work for every task. You only need to change the content inside projects/ and give Cowork a short prompt to trigger it.

In other words:

• system/SYSTEM-RULES.md stays fairly stable

• context/HOW-I-WORK.md stays fairly stable

• what changes each time lives inside projects/

The projects folder is where the specific work lives — the brief, notes, references, drafts, and any file tied to that particular task.

Click here to download all the folders and files used in the examples below

Example 1: Turn scattered notes into a decision brief

Inside projects/, you create this (it may vary by project):

projects/
└── market-review/
 ├── notes.md
 └── template/
 ├── references/
 └── drafts/

The system doesn’t change. The only new thing is the content inside market-review/

Notes.md could look like this:

# Market Review Notes

## Context
We are reviewing competitor moves from the last 30 days.

## Raw observations
- Competitor A launched a lower-priced offer
- Competitor B simplified onboarding
- Competitor C is publishing more case studies

## Risks noticed
- We may overreact to pricing changes
- We do not know yet if these moves are converting better

## What I need
Turn this into a decision brief for internal review.

references/ → previous presentations, competitor reports, benchmarks, internal documents, anything that helps interpret the notes

drafts/ → first version of the brief, a condensed version, a more executive version if you need one

The prompt could be as simple as:

Read the project folder first.
Then ask me questions before you execute.
I want a decision brief, not a generic summary.
Save the result in outputs/briefs/.

The method didn't change. Only the content did.

Here are the results:

The brief grounds the decision, separates risks and options clearly, and ends with an actionable recommendation and concrete next steps.

Example 2: Build an executive report for a meeting

Now consider this case:

projects/
└── q3-review/
 ├── brief.md
 ├── notes.md
 └── template/
        ├── references/
        └── drafts/

You're not working with scattered notes anymore. Now you have material for an executive report.

brief.md could look like this:

# Q3 Review Brief

## Goal
Prepare an executive report for the leadership meeting.

## Audience
Senior leadership team.

## What matters most
- What changed this quarter
- What risks need attention
- What decisions may be needed next

## Constraints
- Keep it concise
- No speculation without labeling it
- Focus on implications, not just activity

And notes.md could look like this:

# Q3 Review Notes

## Highlights
- Revenue held steady
- Customer retention improved slightly
- Two initiatives are behind schedule

## Open questions
- Whether resources need to be rebalanced
- Whether delays are operational or strategic

references/ → metrics, previous quarterly reports, financial documents, team summaries, older slides

drafts/ → first executive report, a shorter version, a risk-focused version, a structure ready to turn into a presentation

The prompt:

Go through the project folder first.
Use my quality standard and workspace context.
Ask me questions before you write.
Then create an executive report for review and save it in outputs/reports/

Here are the results:

Cowork turned scattered context into a clear executive report with metrics, risks, and decisions ready for discussion.

Example 3: Draft a presentation without starting from zero

The folder could look like this:

projects/
└── strategy-update/
    ├── brief.md
    └── template/
        ├── references/
        └── drafts/

And brief.md could look like this:

# Strategy Update Brief

## Goal
Create a presentation outline for a strategy update meeting.

## Audience
Internal leadership and cross-functional leads.

## What the deck should do
- explain what changed
- show why it matters
- align people on next steps

## What I don't want
- slide titles with no message
- too much text
- generic consulting language

references/ → metrics, previous quarterly reports, financial documents, team summaries, older slides

drafts/ → first structure, a condensed version, a more risk-focused variant, an outline ready to turn into slides

The prompt:

Read the project folder first.
Then ask me questions before you execute.
I want a presentation outline with a strong flow, clear sections, and what each slide should do.
Save it in outputs/drafts/

Here are the results:

Cowork didn't just organize ideas. It turned a fuzzy strategic change into a presentation with a clear narrative.

How to prompt Cowork using this system

At this point, you already have everything you need to make Cowork work.

But there’s an important difference between a tidy workspace and one that actually saves you time: how you use it once it’s already up and running.

Here are my two favorite ways:

#1 Question-first flow

A question-first flow is just a way of using Cowork.

The logic:

• Cowork reads the folder

• then it asks you questions if context is missing

• only after that does it build the deliverable

This is activated by how you give the instruction.

For example:

Read the project folder first.
Then ask me questions before you execute.

You already saw this in the examples:

• In the first one, Cowork asked what the real decision was

• In the second one, it asked about the initiatives and the decision that should come out of the meeting

• In the third one, it asked about the workstreams, the next steps, and the format

That was question-first flow in action.

It makes sense to use it when:

• the goal isn’t fully clear yet

• the audience matters

• the result has real impact

• one bad assumption could ruin the deliverable

• there are several possible outputs and you want Cowork to pick the right one

In other words, use it when you don’t want Cowork filling in the gaps on its own.

#2 Short prompts

A short prompt is simply a prompt that works well because the workspace already has context.

That point matters.

In an empty chat, a short prompt usually falls short. But in Cowork, if you already have:

system/
context/
projects/
outputs/

...the prompt no longer has to carry all the weight.

The context lives in the files. The prompt just activates the system.

For example:

Read the project folder first.
Then ask me questions before you execute.
Create the deliverable and save it in outputs/.

or

Use my HOW-I-WORK and SYSTEM-RULES files.
Then use the project folder to create a first draft in outputs/.

Short. It’s not magic. Your workspace is already doing the real work.

If you need to write a very long prompt just for Cowork to understand the basics, the problem usually isn’t the prompt. The problem is that your workspace isn’t carrying enough context yet.

How to grow the system without turning it into a mess

Once people understand this concept, they usually make one of two mistakes:

• They stick with a structure that’s too thin

• or they try to build a huge system from day one

You don’t want either one.

Here’s the core idea: start small and let the work tell you what’s missing.

Grow only when a pattern repeats.

Don’t create files because they might be useful someday. Create them when you notice something is already repeating and deserves its own file.

When it makes sense to expand system/

• Cowork keeps making the same mistake over and over

• you have rules that apply almost all the time

• your quality standard is clearer now

• you want to lock in a way of working that shouldn’t change from project to project

For example: if you notice Cowork keeps confusing a summary with a recommendation, that already deserves a more explicit rule inside SYSTEM-RULES.md or even a small new section.

When it makes sense to expand context/

• your audience already deserves its own file

• your writing style needs more precision

• you want to separate references, tone, or source-of-truth rules

• your way of working no longer fits inside a single file

The goal is context that’s clear and actually usable.

When it makes sense to expand projects/

• a certain kind of work always uses the same files

• a project needs more internal order

• you want to better separate the brief, notes, references, and drafts

• a recurring task already has its own logic

This is where most of the change happens anyway.

How to tell if the system is growing well

You’ll notice it because each new task needs:

• less explanation

• fewer corrections

• fewer restarts

• and delivers something more useful on the first pass

If that’s happening, the system is growing well. If not, you’re just accumulating structure.

Subscribe now

But there was always one problem: You had to be sitting at your desk to use it.

If you wanted to start a task or monitor Cowork’s output while you were at the gym or grabbing coffee, you couldn’t do anything about it. There wasn’t mobile access.

That just changed.

Cowork just got a new feature called Dispatch. This feature lets you text Claude a task from your phone, and it completes the work on your computer while you’re away. When you come back home, the work is done.

In this guide, you’ll learn:

• How to set up Dispatch on your phone and desktop

• How to use Dispatch

• How to unlock Dispatch full potential with skills, tasks, and more

• A few things to keep in mind before you use this feature

I’ve been getting a lot of sponsorship offers for this newsletter lately.

I appreciate every single one, but out of respect for my paid subscribers, I’m keeping this space ad-free.

Click here to read my best Claude guides

If you like my Claude guides and want me to keep making more, consider becoming a paid subscriber👇

Subscribe now

How to set up Dispatch

Here’s what you need:

• The latest version of the Claude desktop app (Mac or Windows)

• The latest version of the Claude mobile app (iPhone or Android)

• A Claude paid plan

Here’s how to set it up:

On your phone:

1. Open the Claude app → open the sidebar → Dispatch → Pair with your desktop

On your computer:

1. Open the Claude desktop app → open the sidebar → Dispatch. You’ll see a screen that says “Coworking on the go.” Click “Get started.”

2. Click “I’m signed in on my phone.”

3. Toggle on two things: give Claude access to your files, and keep your computer awake while Dispatch is running (if your computer goes to sleep, Claude stops working)

The setup is finished!

You should see the screen below when you go to Dispatch in the desktop app (in the mobile app, you’ll see a simplified version)

You now have one continuous conversation with Claude Cowork that works from both your phone and your desktop. Text it a task from your phone, and Claude works on your computer using your local files, skills, connectors, and more.

Let’s see how to use Dispatch with a basic example.

Subscribe now

How to use Dispatch

Dispatch is designed for moments when your phone is the only device you have. Here’s how to fire it up on your phone.

Go to the Claude phone app → open the sidebar → Dispatch

You’ll see this default message on your phone.

Use a prompt to point it to a folder on your computer and tell it what to do.

You can type something like this:

“In my [folder name], can you give me a quick summary of what’s in the final report spreadsheet?”

In your prompt, be specific about the folder and the file name.

In a few minutes, the report summary will be generated on your phone. No laptop needed. Just a text message to Claude and a clean summary back on the phone.

But that’s only the most basic thing you can do with Dispatch!

Now that you've seen it in action on your phone, let me show you the computer setup that unlocks its full potential.

Unlock Dispatch's full potential with skills, tasks, and more

If you’ve used Cowork before, you probably know about skills and scheduled tasks.

Skills are reusable instructions you save inside a folder so Claude knows exactly how to handle specific tasks. Scheduled tasks are prompts that run automatically at a certain hour and day.

Both make Dispatch really powerful, because you can trigger them from your phone with a short message.

Step 1: Add skills to your list

To see the list of skills you can trigger from your phone, do this on the desktop app:

Open the sidebar → Go to Customize → Skills → List of skills available

To download skills and learn how to create your own, read this: artificialcorner.com/p/claude

Step 2: Add tasks to Scheduled

To see the list of tasks you can trigger from your phone, do this on the desktop app:

Open the sidebar → Go to Scheduled → List of tasks available

To learn how to create scheduled tasks, read this: artificialcorner.com/claude-automates

Step 3: Trigger skills or tasks from your phone

Now that you know the list of skills and tasks available, you can trigger them with a short message from your phone.

Here are the steps to follow:

1. Open the Claude phone app → open the sidebar → Dispatch

2. Type “Run the [name of skill/task]“

3. Allow permission

4. You get your output

If your task/skill generates a file, you might need to explicitly tell Claude to show you the file in the phone Dispatch chat.

Subscribe now

Here’s an example

Say I’m in the barber shop or sitting in the taxi. I can grab my phone, open Claude, and type: Run the task “export top paid newsletters to CSV.” That’s a task that scrapes Substack newsletter rankings from a site.

Pro tip: Use simple names for your skills/tasks so they're easy to trigger on your phone.

Done! The CSV file was generated. I can open it on both my phone and computer.

When I come back home and open Claude Cowork on my computer, the same chat is right there in the Dispatch tab.

But there’s one thing in the desktop app that you won’t find in the mobile app — in recent chats, you’ll see one chat tagged “Dispatch.“ That’s a chat that is created every time we run a task using Dispatch (but it’s only visible in the desktop app)

Inside the chat tagged Dispatch, there’s the hidden prompt and process used to carry out my task.

The complete chat is like the longer version of the Dispatch tab. There, I can also find the CSV file generated.

Final Notes

Here are a few things worth keeping in mind:

• Your computer has to stay on: This is the most important thing. If your desktop goes to sleep or you close the Claude app, Dispatch stops working. Toggle on the “prevent sleep” option during setup.

• It’s one continuous thread: There’s no way to start a new conversation or manage multiple threads in Dispatch. Everything lives in a single chat, which means Claude remembers context from previous tasks.

• You can run multiple tasks at once: Send three or five different tasks from your phone and Claude will work on them in parallel (as long as the tasks don’t depend on each other).

• Sequential tasks aren’t supported yet: If task B depends on the output of task A (like “research X, then email the team about it”), you’ll need to send them one at a time and wait for the first one to finish.

• No notifications yet. Claude won’t ping you when a task is done. You’ll need to check back in the chat to see results. Hopefully, this gets added soon.

• Connectors extend to mobile automatically. Whatever apps you’ve connected in Cowork (Gmail, Slack, Google Drive, etc.) are available through Dispatch

Dispatch is still in preview mode, so expect some rough edges. But even in its current state, it’s already changing how I use Cowork.

Subscribe now

Click here to read my best Claude guides

I have too.

But here’s the thing: Claude doesn’t count messages. It counts tokens. Some chats eat through your limit 10x faster than others.

So I put together the best practices to stretch your Claude plan as far as it can go.

You don't have to follow all of them. Think of it like one of those iPhone tutorials to save battery. You could follow every tip, but you wouldn't have as much fun with it if you did.

Pick the ones that fit how you work. Even 2–3 of these will make a noticeable difference.

A new week, a new Claude guide

Click here to browse the full collection

If you want to keep these Claude guides coming, consider supporting me by becoming a paid subscriber 👇

Subscribe now

#1 Edit your prompt. Don’t send a follow-up

When Claude’s answer misses the mark, most people send another message like “no, I meant...” or “can you try again but...”

Every follow-up message adds to the conversation history.

Claude re-reads the entire conversation every single turn. So by message 10, you’re not just paying for message 10 — you’re paying for messages 1 through 9 all over again.

This adds up fast.

Instead of sending a new message, click the edit icon on your original message, fix the prompt, and regenerate. The old exchange gets replaced instead of added.

Here’s how to do it

Go to the prompt you wrote → click the pencil icon → rewrite your prompt → hit send. Claude regenerates as if the bad response never happened.

#2 Start fresh every 15–20 messages

Again, Claude re-reads your entire conversation history every single turn.

Your first message costs hundreds of tokens. By message 20, a simple question can cost thousands of tokens. The longer the chat, the more expensive every message becomes.

Long chats are expensive chats.

Start a new conversation every 15–20 messages.

If you need context from the old chat, ask Claude for a summary, copy it, and paste it into the new chat as your first message.

Here’s how to do it

When a conversation starts getting long → ask Claude: “Summarize everything we’ve discussed so far” → copy the summary → open a new chat → paste it as your first message → start fresh

You get fresh context without the bloated token cost.

Subscribe now

#3 Combine multiple questions into one

Sending three separate messages means Claude loads the conversation context three separate times.

Batch your questions into a single message.

Instead of sending “Summarize this article,” then “List the main points as bullets,” then “Suggest a headline,” send all three at once.

Write one message like: “Summarize this article, list the main points as bullets, then suggest a headline.”

One message. Three answers. One context load.

By the way, the answers are often better when Claude sees the full picture.

#4 Upload recurring files to projects

If you’re uploading the same file in multiple chats, Claude is re-counting those tokens every single time.

The solution? Use Projects. Upload your files there once, and they get cached so they don’t re-cost you on each conversation.

This is a huge saver for anyone who works with long documents regularly.

Here’s how to do it

Go to the sidebar → click Projects → create a new Project → upload your recurring files there. Now every conversation inside that Project can reference those files without re-paying the token cost.

Upload once. Stop wasting tokens every time.

Subscribe now

#5 Set up memory

Every conversation you start without context burns setup messages just re-explaining who you are and how you work.

“I’m a marketer, I write in a casual tone, I prefer short paragraphs...”

You’re spending tokens on the same intro over and over.

A simple way to fix this is to set up memory:

Settings → Memory → add things like your job, your writing style, your preferences.

A better approach is to create an .md file with context about you.

Read this guide to generate .md files with context: artificialcorner.com/p/md-file

Create it once. It runs forever.

#6 Turn off features you’re not using

If you’re not using web search, research mode, connectors, and others, turn them off

They add tokens to responses even when you don’t need them.

Same goes for extended thinking. Leave it off by default and only switch it on when you want Claude to think longer for better responses.

Here’s how to turn them off

Click + → Turn off research, web search, or connectors

Click model selector → Turn off extended thinking

#7 Use Haiku for simple tasks

Most people use the same model for everything — whether they’re brainstorming ideas or doing deep research.

The bigger the model, the more tokens it costs per message.

Using the right model is the single highest-impact decision you can make.

Use the right model for the job:

• Haiku (very low cost) → Quick answers

• Sonnet (medium cost) → Everyday tasks

• Opus (high cost) → Complex tasks

Here’s how to switch models

Just click the model selector at the top of the chat and pick the right one for the task.

#8 Spread your work across the day

I used to sit down at 9 and power through everything in one session. By 11, I’d already hit Claude’s limit. Done for the day.

What I didn’t realize: Claude’s limit runs on a rolling 5-hour window. It doesn’t reset at midnight. Instead, your usage gradually falls off — messages you sent at 9 AM stop counting against you by 2 PM. So if you burn through everything in one burst, you’re leaving most of the day’s capacity on the table.

Split your work into 2–3 sessions per day, instead of 1.

Here’s what I do.

I do a session in the morning, another after lunch, and another in the evening. By the time I come back for each session, my earlier usage has already aged out of the window, and I’ve got a fresh allowance to work with.

Space them out so your rolling window resets between sessions.

Claude counts tokens, not messages. Now you know how to make every token count.

Some of these best practices were taken from Claude Support. For more, read this

Click here to read the best Claude guides

Subscribe now

You chat with it, write some prompts, use the default model, and wonder why your outputs feel no different from ChatGPT.

Why? You haven’t set up Claude properly.

In this guide, we’ll fix that. You’ll learn how to:

• Set up Claude chat (the features that separate it from ChatGPT)

• Set up Claude Cowork (the files you need before every session)

• Set up Claude Code (for non-technical folks)

New to Claude? Click here to read my best Claude guides

If you like my Claude guides, consider becoming a paid subscriber to make more guides like this possible👇

Subscribe now

How to set up Claude chat

“Claude chat” is the default way to interact with Claude, whether you use the web or desktop app. Here’s how to get the most out of it.

#1 Models + extended thinking

Claude gives you three models to choose from. Think of them as three employees with different strengths.

Opus 4.6: It’s the smartest. It takes its time, thinks deeper, and gives you the most thorough answers.

• When to use it: For complex tasks (long documents, strategy, analysis, etc)

• Downside: It’s slower and eats through your usage limits faster.

Sonnet 4.6 (default model): It’s fast and smart

• When to use it: For quick writing, brainstorming, summarizing, and daily work

• Downside: It’s not as thorough as Opus

Haiku 4.5: It replies almost instantly

• When to use it: For quick questions or simple tasks (Claude won’t think too hard)

• Downside: It’s the least powerful of the three

To switch models, click the model selector.

To be completely transparent, I use Opus 95% of the time.

When I need AI for casual tasks, I use ChatGPT or Gemini instead of wasting usage on Sonnet or Haiku. If you have access to either, I’d recommend doing the same. If not, use Opus wisely.

I also prefer to keep “extended thinking” turned on. When it’s on, you’re telling Claude: “Don’t rush. Think this through before you answer.”Leave it off for everyday stuff (quick questions, casual writing, simple requests).

#2 Projects

Think of a project as a folder where Claude already knows the context. You upload documents, add instructions, and from that point on, every conversation inside that project starts with context.

Here's how to create a project:

1. Click “Projects” in the left sidebar

2. Click “New Project”

3. Give it a name

4. Add custom instructions and upload files (PDFs, docs, etc)

Pro tip: Create an .md file with context about you or your work and upload it to your project, so you don’t have to repeat it over and over again.

📚 How to generate an .md file with context: artificialcorner.com/p/md-file

Subscribe now

#3 Skills

A skill is a set of instructions you save once, and Claude loads it automatically whenever it's relevant.

Say you write weekly reports and they always follow the same structure. Instead of explaining that structure every time, you create a skill for it. Next time you prompt “write my weekly report,” Claude recognizes the task, loads your skill, and follows your format from the start.

To work with Skills, enable it in settings.

Go to Settings → Capabilities → Turn on Skills

You don't need to be technical to create a skill. Claude has a built-in skill creator that builds the skill for you. Here’s how to enable the skill creator:

1. Go to preferences in the left sidebar

2. Click “Skills“

3. Select “skill-creator“

4. Toggle skill-creator on (leave it always on)

If you toggle on the skill brand-guidelines and try the prompt below, you’ll get slides with the Anthropic colors, design, etc.

create a quick presentation on claude skills. use anthropic brand guidelines

📚 Learn more about skills: artificialcorner.com/p/claude

#4 Artifacts

Artifacts turn ideas into shareable apps, tools, or content. Common examples of artifacts are diagrams, flowcharts, websites, and documents.

To trigger an artifact, just describe what you want (an app, diagram, document, or chart) and Claude will automatically create an artifact for it

To turn artifacts on, do this:

1. Click your initials or name in the lower left corner

2. Navigate to Settings > Capabilities

3. Find Artifacts and toggle it on or off

One video is worth a thousand words. Watch artifacts in action:

Subscribe now

How to set up Claude Cowork

Claude chat is like texting a smart friend. You ask, it answers.

Cowork is like hiring an assistant. You describe the task, point it to your files, and it gets to work — planning steps, reading documents, creating files, and delivering results directly to a folder on your computer.

Cowork lives inside the Claude desktop app. Here’s how to set it up:

1. Download the Claude app → claude.com/download

2. Open the app and sign in

3. At the top of the window, you’ll see three tabs. Click Cowork

4. Create or pick a folder on your computer where Claude will work

5. Point Claude to that folder

   

6. Try this prompt:

You are going to help me create a context file I can use with Claude Cowork.

Ask me questions one at a time to understand my work: who I am, what I do, how I like things done, what I never want to see in outputs, and what “good work” means to me specifically. Use AskUserQuestion for it.

Ask me 40-50 questions total. One at a time. Push back if my answer is vague — ask for a specific example. When we’re done, compile everything into a .md file I can save and use with Claude.

You’ll get an .md file with context about you. Put it inside the folder Cowork has access to.

Next time you use Cowork, add this at the end of your first prompt:

[your prompt]

Before you start, read [your filename].md first

Now you don’t need to repeat context to Cowork.

Pro tip: To get the most out of Cowork, it helps to keep these three .md files ready:

• about-me.md — everything about you, your work, etc

• voice-and-style.md — your tone preferences, formatting rules, words you like/dislike, writing samples, etc

• working-rules.md — how Claude should behave

📚 How to generate these .md files: artificialcorner.com/p/md-file

Cowork offers something that you won’t find in Claude chat: plugins. They’re pre-built instruction sets that make Claude good at specific professional tasks.

Here’s how to install and use plugins:

Cowork → Click + → Add plugins → Pick the plugin you like → Click install

📚 Learn more about plugins: artificialcorner.com/p/cowork

Subscribe now

How to set up Claude Code

Claude Code is Cowork on steroids.

Both can read your files and execute tasks. But Claude Code goes deeper. It can edit code, run commands, manage entire projects, and work across multiple files at once. It’s the most powerful way to use Claude.

Now, here's the part that scares most people: Claude Code was originally built for the terminal (that black screen with the blinking cursor that feels like hacking in a movie)

The good news is that after the installation, you don’t have to touch the terminal!

Here’s how to install Claude Code:

1. Go to the Claude website → “Install the latest version“ → Look for your operating system → copy and paste the installation command into your terminal → execute it

If you’re new to the terminal, here’s how to open it:

• Windows: press the Windows key, type “CMD” and press enter

• Mac: press Command (⌘) + Spacebar, type “Terminal“ and press enter

After running the command, you should see this message: “✅ Installation complete!“

To verify that Claude was installed successfully, type “claude” and press enter. You should see the image below in your terminal

2. Install VS Code. It has a friendlier interface to work with Claude Code

3. Install the Claude extension within Visual Studio Code: Open VS Code → View → Extensions → search for Claude → click Install

4. After installing the extension, click on the Claude icon to log in to Claude

5. Finally, open a folder you will use as your working directory inside Visual Studio Code: File → “Open Folder” → Select a folder

Inside the folder selected, add the structure below.

my-project/
├── CLAUDE.md
├── context/
│ └── content-context.md
├── profiles/
│ └── audience-profile.md
├──skills/
│ └── SKILL.md

Download the files here: Click on the green “Code“ button → download ZIP

Viewed inside VS Code, the structure will look like this:

This is just a basic Claude Code setup, though.

For more, check out our Claude Code series here.

Click here to read the best Claude guides

Subscribe now

When you switch from ChatGPT to Claude, the weird part isn’t the model.
It’s the reset.

You open a new chat, and everything you worked to build starts to disappear.

• The tone shifts.

• The format drifts.

• You end up re-explaining decisions you already made.

That’s not a prompt problem. It’s a memory problem.

Claude Code fixes this because you don’t “work inside a chat.”
You work inside a project, and the project loads your context from files every time.

This is the setup I wish I had from day one.

In this guide, you’ll get:

• 3 prompts to extract your existing chat history and turn it into reusable files

• A simple folder structure that keeps Claude Code consistent across every session

• A system you can set up in minutes and keep building on over time

No coding. You run the prompts in your current AI tool (ChatGPT, Gemini, etc), copy the outputs, and paste them into your Claude Code project.

Our series of guides on Claude, Claude Cowork and Claude Code keeps growing every week!

You can find all my Claude guides here. Consider becoming a paid subscriber if you like these guides 👇

Subscribe now

The shift that matters: from chat to project

Most people use AI like this:

Open a chat → ask for something → correct the output → repeat tomorrow from scratch

That works fine for one-off tasks.

However, once you start producing the same kind of work on a regular basis (guides, decisions, documents, analysis), two things start breaking:

• Inconsistency (each day it answers differently)

• Repetition (you keep re-explaining who you are, what you want, and how you work)

Claude Code fixes this because it reads from files you set up, not just your latest message.

So the goal isn’t to move your prompts over. It’s to move your structure.

It’s to build a stable structure Claude Code can load automatically.

Here is the workflow :

Instead of relying on the chat to "remember," you make the project carry the memory.

The minimum structure that actually scales

This setup is enough to make the system work — and simple enough to actually maintain. As your project grows, it grows with you.

my-project/
├── CLAUDE.md
├── rules/
│ ├── quality.md
│ └── boundaries.md
└── context/
├── you.md
├── project.md
└── writing-voice.md

Why does it work?

• It's not about having a lot of files. Each file just has one clear job.

• CLAUDE.md acts as the project’s memory.

• One thing worth knowing: CLAUDE.md doesn't replace your system prompt. It just adds context on top — things that are always true about your project. That's what makes it so consistent.

Think of it this way:

• CLAUDE.md = the essentials that always apply

• rules/ = longer policies and checklists (quality, boundaries, style)

• context/ = project facts, how you work, your voice, your audience

• skills/ = repeatable procedures (only when there is truly repetition)

This keeps you from stuffing everything into one file until it becomes a mess you can't manage.

Subscribe now

Prompts to extract your chat history and turn it into reusable files

The whole point is to pull out what you’ve already built across your history and turn it into project memory.

Here are three prompts, each focused on a different layer:

• Prompt 1: Extractor from a ChatGPT export

• Prompt 2: Project blueprint

• Prompt 3: Writing voice compiler (articles, emails, notes, and more)

The idea is simple: don’t try to do magic, do setup. You don’t need to run all three every time. Treat them like modules — use only what you need.

You run these prompts in your current AI chat tool.

What you get back are Markdown blocks you copy and paste into .md files in your Claude Code project.

Prompt 1: Extractor from a ChatGPT export

Read more

Hi!

I wrote today’s post in collaboration with . Ilia writes Prosper, a newsletter focused on practical AI workflows, systems, and frameworks for professionals who want to use AI more thoughtfully.

If you’re interested in building systems for working with AI, I highly recommend checking out Prosper.

Picture this: you need help with your business, so you open Claude (or any other AI) and start the same way: “Hey, quick context: here is what my business does, this is what I need help with today … ”

Claude helps you out. But then the next time comes, and you’re doing it all over again: rewriting context, re-explaining your situation, updating what was discussed last time.

Every. Single. Time.

That used to take up a lot of my time until I started using .md files. Once I did, I stopped re-explaining myself from scratch.

An .md file is a simple text file that holds context about you, your work, or your projects. You create it once. You reuse it everywhere.

In this guide, you’ll learn:

• How to create your own .md file

• 5 interview-style prompts that generate ready-to-use .md files for business context, writing voice, project work, content creation, and analysis

• How to use your .md file in Cowork

You can upload .md files to Claude, ChatGPT, or any other AI, but I recommend using it on Claude Cowork. In the next section, we’ll see why.

New to Claude? Click here to read our best Claude guides

Claude is powerful, but most people barely scratch the surface.

Every week, paid subscribers get step-by-step guides just like this one. No technical skills needed. No ads. Just the stuff that actually matters (+ these perks) 👇

Subscribe now

Cowork makes .md files more powerful

In regular Claude chat, you paste a .md file into the conversation. Works fine, but you do it every time.

Cowork is different.

Drop a .md file into a folder and Claude reads it automatically whenever you work in that folder. Open the folder, Claude already knows the context. No pasting. No re-uploading.

Here’s what that looks like in practice.

Ilia writes a newsletter called Prosper. His writing folder contains a voice profile with banned words, sentence patterns he avoids, positions he’d never take.

“When I open that folder in Cowork and ask for a draft, it doesn’t ask who I am. It reads the file and produces something that sounds like I wrote it on a Tuesday morning”

- Ilia Karelin

Same principle, any folder. Your client folder knows your delivery standards. Your research folder knows how you think about data.

Zero re-explaining. Claude starts already knowing your work.

Here’s how to use Cowork:

1. Download Claude for desktop

2. Open the app and choose “Cowork“

3. Give access to the folder containing the .md file

In the next section, we’ll learn how to create an .md file. Once you connect the file to Cowork, it’ll have it as a reference for future responses.

📚 Claude Cowork - Complete Guide: artificialcorner.com/p/cowork

📚 Move from ChatGPT to Claude: artificialcorner.com/p/switch-to-claude

📚 Best Claude Setup: artificialcorner.com/p/claude-setup

Subscribe now

How to create your .md file: the interview method

You don’t sit down and write the file completely from scratch. Claude interviews you and builds it from your answers.

Instead of staring at a blank document trying to articulate how you work, you let Claude interview you, and we’re going to give it a starting point.

It asks the questions. You answer honestly. At the end, it hands you a finished file.

The insight that makes this work: your profile is mostly about what you “reject”, not what you prefer. What you never want to see. What you’d never say.

That’s where the real signal is - not “I like direct communication” but “I’d never send a bullet-pointed email to a client.”

Here’s the base prompt:

You are going to help me create a context file I can use with Claude Cowork.

Ask me questions one at a time to understand my work: who I am, what I do,
how I like things done, what I never want to see in outputs, and what
“good work” means to me specifically. Use AskUserQuestion for it.

Ask me 40-50 questions total. One at a time. Push back if my answer is
vague — ask for a specific example. When we’re done, compile everything
into a .md file I can save and use with Claude.

Claude takes it from there.

Note: For the examples below, I told Claude to ask me 3 questions and then compile a first draft of the file. In practice, 40-50 questions produce a much richer profile — but this is a good way to see what you’re building toward before committing to the full interview. You can always run it again or ask Claude to update the file with more details later.

The key is answering specifically. “I want professional outputs” is useless. “I want outputs that sound like they came from a CFO - direct, numbers-first, no filler” gives Claude something it can actually apply.

In the previous prompt, we used AskUserQuestion to collect your preferences through preset options, but short multiple-choice answers only scratch the surface. Claude works best when it has rich, detailed context about you

For a more tailored experience, choose one of the five role-specific prompts below and answer as thoroughly as you can. The more detailed and thoughtful your responses, the more context Claude will have to work with.

📚 Claude Opus 4.6

📚 Claude - The Setup Most People Never Do

5 interview prompts that generate ready-to-use .md files

1. Business context

For owners, freelancers, and consultants who want Claude to help with anything business-related: client calls, contractor briefings, meeting prep, proposals.

You are going to help me create a business context .md file for Claude Cowork.

Ask me questions one at a time. Cover:

- What my business does - what I sell, who I serve, what problems I solve

- How I describe my business in my own words (not my website version)

- My team structure - employees, freelancers, partners, and who does what

- My biggest recurring challenges: meetings I dread, tasks I repeat, decisions I struggle with

- My communication style - how I write emails, how I run meetings, what I sound like

- My financial setup - how I track money, who I report to, what I need to know for accounting

- My non-negotiables - things I never want Claude to suggest or assume about my work

- What a great output actually looks like for me, with a specific example

Ask 40-50 questions. One at a time.

Interview rules:
- Push back on vague answers. If I say something generic, ask for a specific example.
- Call out contradictions. If something conflicts with an earlier answer, flag it.
- Don’t accept “I don’t know.” Reframe the question or approach it differently.
- Go deeper on interesting threads. If something unusual comes up, follow it before moving on.
- Ask for real samples - actual emails, documents, or decisions. Descriptions mean less than examples.
- Don’t validate my answers. No “great point” or “that’s interesting.” Just ask the next question or push deeper.
- If my answer sounds like how I wish I worked rather than how I actually work, call it out. Ask me for proof.
- If my answer could describe anyone in my profession, flag it. Ask what’s specific to me.

When we’re done, compile everything into a .md file with clear section headers.
Include the specific examples alongside the rules — they’re what make the file useful.

With this one, you will definitely see your prep time that used to take 20 minutes of re-explaining, now takes 2.

2. Writer’s voice

For newsletter writers, bloggers, and anyone who creates written content and wants Claude to draft in their actual voice — not a polished AI version of it.

You are going to help me capture my writing voice in a .md file.

Ask me questions one at a time. Cover:

- What I write and who I write for

- What my writing sounds like at its best — ask me to share an example paragraph

- What my writing sounds like when I’m just going through the motions

- Phrases I use naturally that feel like “me”

- Words and phrases I hate seeing in my own writing

- Topics I always come back to, even when I’m not trying to

- What I refuse to write about, or positions I’d never take

- My formatting instincts: long or short paragraphs? headers or prose? lists or sentences?

- Writers I admire — and specifically what I’d steal from each one

- What feedback I keep getting on my writing

- What feedback I keep ignoring, and why

Ask 40-50 questions. One at a time.

Interview rules:
- Push back on vague answers. If I say “I like directness,” ask me what that looks like in a specific paragraph.
- Call out contradictions. If something conflicts with an earlier answer, flag it.
- Don’t accept “I don’t know.” Reframe the question or approach it differently.
- Go deeper on interesting threads. If something unusual comes up, follow it before moving on.
- Ask for real samples — actual paragraphs I’ve written, sentences I’d never publish. Descriptions of style mean less than examples of it.
- Don’t validate my answers. No “great point” or “that’s interesting.” Just ask the next question or push deeper.
- If my answer sounds like the writer I wish I were rather than the writer I actually am, call it out. Ask me for proof.
- If my answer could describe any writer in my space, flag it. Ask what’s specific to me.

When we’re done, compile everything into a .md file with clear section headers.
Include the specific examples alongside the rules — they’re what make the file useful.

That’s the one I have in the folder I work in. Adjectives like “direct, warm, conversational” mean nothing to Claude. You end up with actual rules - hard nos, specific examples, the exact phrases you’d never use - the kind of guardrails that make Claude’s drafts sound like yours.

The rule that makes this work: 80% of a good voice profile is about what you reject. What sounds wrong. What you’d never publish. That’s what separates your voice from everyone else’s.

3. Client and project work

For consultants, agencies, and project managers whose Cowork folders hold client deliverables, project notes, and anything where working-style context matters.

You are going to help me create a client work .md file for Claude Cowork.

Ask me questions one at a time. Cover:

- The type of work I do with clients — deliverables, services, projects

- How I onboard clients and what I need to know before I start

- My delivery standards — what “done” looks like, how I format deliverables,

 what I never send without reviewing myself

- How I communicate with clients — tone, frequency, what I always and never say

- My biggest recurring frustrations with client work

- What makes a project go well vs. badly — in specific terms, not general ones

- How I handle scope creep, missed deadlines, or difficult conversations

- Tools and systems I use (where projects live, how I track things)

- What I’d want a smart assistant to know before helping me with any client task

Ask 40-50 questions. One at a time.

Interview rules:
- Push back on vague answers. If I give a process-level answer, ask me for a real example from a recent project.
- Call out contradictions. If something conflicts with an earlier answer, flag it.
- Don’t accept “I don’t know.” Reframe the question or approach it differently.
- Go deeper on interesting threads. If something unusual comes up, follow it before moving on.
- Ask for real samples — actual client emails, deliverable excerpts, or project decisions. Descriptions mean less than examples.
- Don’t validate my answers. No “great point” or “that’s interesting.” Just ask the next question or push deeper.
- If my answer sounds like how I wish I handled clients rather than how I actually do, call it out. Ask me for proof.
- If my answer could describe any consultant or PM, flag it. Ask what’s specific to me.

When we’re done, compile everything into a .md file with clear section headers.
Include the specific examples alongside the rules — they’re what make the file useful.

Drop this file into a client folder and Claude stops being a blank-slate assistant.

It knows your delivery standards, how you talk to clients, and what you’d never send without reviewing. One file, reused across every project.

4. Content creator

For YouTubers, podcasters, and social media creators who want Claude to help with ideation, scripting, and repurposing - in a way that actually fits their content.

You are going to help me capture my content identity in a .md file.

Ask me questions one at a time. Cover:

- What kind of content I make and where I publish it

- Who my audience is — not who I think they are, but what they actually respond to

- My content pillars: the 3-5 topics I return to again and again

- My on-camera or on-mic personality vs. my off-camera personality

- Content I’m most proud of and why it worked

- Content I’ve made that flopped — and what I think went wrong

- Content trends I refuse to follow even if they’d probably perform

- My production constraints: time, team size, tools I have access to

- The 3 creators I admire most and specifically what I’d take from each one

- What I want people to feel or do after consuming my content

- What I’d never make, even if someone paid me to

Ask 40-50 questions. One at a time.

Interview rules:
- Push back on vague answers. If I say “I want to be authentic,” ask me what that looks like in a specific piece of content I’ve made.
- Call out contradictions. If something conflicts with an earlier answer, flag it.
- Don’t accept “I don’t know.” Reframe the question or approach it differently.
- Go deeper on interesting threads. If something unusual comes up, follow it before moving on.
- Ask for real samples — actual video scripts, post captions, or content decisions. Descriptions mean less than examples.
- Don’t validate my answers. No “great point” or “that’s interesting.” Just ask the next question or push deeper.
- If my answer sounds like the creator I wish I were rather than the creator I actually am, call it out. Ask me for proof.
- If my answer could describe any creator in my niche, flag it. Ask what’s specific to me.

When we’re done, compile everything into a .md file with clear section headers.
Include the specific examples alongside the rules — they’re what make the file useful.

Claude stops suggesting ideas that don’t fit your channel. Scripts stop sounding like they could belong to anyone.

When you repurpose a video into a thread or a newsletter, the thing that made the original work doesn’t get lost in translation.

5. Analyst and researcher

For data analysts, researchers, and strategists whose Cowork folders contain reports, datasets, and research notes where output format and analytical standards matter.

You are going to help me create an analyst context .md file for Claude Cowork.

Ask me questions one at a time. Cover:

- What I analyze and for whom — who reads my outputs and what they do with them

- How I define a good analysis — what questions it answers, what format it takes

- My output preferences: how I structure reports, what I always include, what I always cut

- The frameworks or mental models I use most, even informal ones

- What “rigorous” means to me — where I draw the line on confidence and evidence

- My biggest recurring frustrations with how analysis gets done or presented

- Tools and data sources I work with regularly

- What I want someone to do after reading my analysis: decide, act, question, understand?

- What I never want to see in an analysis output — vague language, unsupported claims,
specific phrases that signal lazy thinking

Ask 40-50 questions. One at a time.

Interview rules:
- Push back on vague answers. If I describe a preference, ask me for an example of when it mattered in real work.
- Call out contradictions. If something conflicts with an earlier answer, flag it.
- Don’t accept “I don’t know.” Reframe the question or approach it differently.
- Go deeper on interesting threads. If something unusual comes up, follow it before moving on.
- Ask for real samples — actual report excerpts, analysis decisions, or data interpretations. Descriptions mean less than examples.
- Don’t validate my answers. No “great point” or “that’s interesting.” Just ask the next question or push deeper.
- If my answer sounds like how I wish I analyzed rather than how I actually do, call it out. Ask me for proof.
- If my answer could describe any analyst in my field, flag it. Ask what’s specific to me.

When we’re done, compile everything into a .md file with clear section headers.
Include the specific examples alongside the rules — they’re what make the file useful.

No more “here are the findings” summaries that don’t match how you think or how your audience reads.

Outputs formatted and reasoned the way you’d do it yourself - because the file tells Claude exactly what that looks like.

Subscribe now

Do This Now

Follow these steps:

1. Open the Claude app

2. Go to the Cowork tab.

3. Paste one prompt from the list above into Cowork

4. Answer every question.

When Claude pushes back and asks for specifics, give them. That’s where the file gets good. When it hands you the finished file, save it with a descriptive name - [voice_profile.md], [business_context.md], whatever fits what you built.

Drop the .md file in your most-used work folder and give Cowork access to that folder:

At the end of your first prompt, type this:

[your prompt]

Before you start, read [your filename].md first

Here’s an example from just a couple of words I wrote in Cowork:

Done.

Every time you open that folder and give Cowork a task - prep a meeting, draft an email, analyze a dataset - it reads the context file before doing anything. It already knows who you are.

If you want to take this to the next level, use the scheduled tasks feature:

📚 Claude scheduled tasks: artificialcorner.com/p/claude-automates

Subscribe now

Don’t forget to subscribe to Prosper, where you can find practical AI workflows, systems, and frameworks for professionals who want to use AI more thoughtfully.

Here are some of Ilia’s best posts:

• Commands vs Skills vs Agents in Claude Code

• Official Claude Prompting: Everything You Need To Know

• I Stopped Scrolling for AI News

I used to spend hours writing scripts to automate the web, Excel reports, etc.

Thanks to Claude, now anyone can do this in minutes without writing code!

The best part? You can schedule tasks directly in Claude, so it runs them automatically on a recurring basis. Set it once and it just works, while you focus on everything else.

In this guide, I’ll show you step by step how to use Claude to schedule tasks like:

• Weekly research: Track topics, competitors, or industry news

• Daily reports: Compile data from spreadsheets into a formatted summary

• Daily file organizer: Sort and clean up files in your downloads folder

For this, we’ll use two tools: Claude in Chrome to automate the web, and Claude Cowork to automate your files.

Twice a week, I share the best ways to get more out of Claude.

Click here to read our best Claude guides

and consider becoming a paid subscriber if you like my Claude guides 👇

Subscribe now

Claude in Chrome: Automate the Web

Claude in Chrome is a Chrome extension that automates the web. To work with it, follow these steps:

1. Open Google Chrome

2. Go to Claude in Chrome → Click “Add to Chrome” to install

3. Sign in with your Claude account

4. Grant permissions so Claude can interact with your browser

The Claude icon will appear in your Chrome toolbar. Click it to open Claude in a side panel that stays visible while you browse.

Anything that involves repetitive web actions can be automated with this extension.

Just click the Claude sidebar while browsing any site, describe the task, and Claude will control your browser until it’s done.

Here’s one way Claude in Chrome has automated my work.

Every week, I track which Substack newsletters break into the top 200. I analyze what's working and spot potential collaborations.

To get data for my analysis, I have Claude scrape Reletter’s Substack charts with this prompt:

Build a table with the following columns for the first 250 elements in the table:

• Ranking

• Newsletter

• Date (today’s date)

• Type = free/paid

Give me a link to download the file in csv format

Here’s a video demo.

If you do this manually (clicks, copy-paste, etc), it takes forever. With Claude, it takes a couple of minutes.

It gets even better with the scheduled tasks feature. You can turn a prompt that works into a task that runs any time you want. Just click the three dots (at the top of the chat) and select “Convert to task.“

You’ll be redirected to the Claude in Chrome settings with the window below. Mine is empty, but if you clicked on “convert to task,“ the fields should be automatically filled based on your prompt.

Claude might tweak your prompt to make it more robust. Review it and adjust if necessary. Also, toggle on “schedule” and pick when the task should run.

Once you’re happy with it, click on “Create shortcut.“

To edit a shortcut later: Go to Claude in Chrome settings (under the three dots) → Settings → Shortcut

You’ll see all your scheduled tasks listed there. I schedule mine to run every Tuesday.

Now every week, Claude scrapes the site and downloads the data for me.

That’s how you automate the web. Now let’s see how to automate your desktop.

Important note: Scheduled tasks only run while your computer is awake

Subscribe now

Cowork: Automate Your Files

Claude Cowork feels like a virtual employee, capable of managing files, organizing data, and performing multi-step tasks directly on your computer.

Recently, Cowork also got scheduled tasks!

It works a lot like Claude in Chrome. Write a prompt once, and Cowork can turn it into a scheduled task.

Here’s how to create a scheduled task from scratch:

1. Download Claude for desktop

2. Open the app and choose “Cowork“

3. Click “Scheduled” in the left sidebar

4. Click “+ New task”

5. Fill in the details (prompt, task name, etc)

6. Click “Save” to add a new task to the Scheduled tasks page.

We can also turn a working prompt into a task that runs whenever we want.

In the previous example, I used Claude in Chrome to set up a task that runs every Tuesday. It scrapes the data and generates a fresh dataset with a Substack newsletter ranking.

Now I’ll have Cowork analyze that data and generate a weekly report showing the 10 newsletters with the biggest gains and biggest drops.

Here’s the prompt I use:

Based on the data collected in the folder “Substack Newletter Ranking” generate an analysis that covers:

• Biggest rank gains: Top 10 newsletters with the largest improvement (largest drop in rank number)

• New entrants: Top 20 newsletters that appear in the rankings only in the most recent dataset

• Biggest rank losses: Top 10 newsletters with the largest decline (greatest increase in rank number).

The folder contains two types of datasets:

• Paid: rankings based on ARR

• Free: rankings based on number of free email subscribers

Create the same analysis separately for each dataset (Paid and Free)

After running the task, Cowork drops a spreadsheet right in my folder with the full analysis. Congrats to the newsletters climbing fast … hope to be on the list soon :)

Now I can turn this working prompt into a scheduled task.

In the chat, click the dropdown arrow and select “schedule.”

The /schedule command will appear. Type something like “run this every [day] at [time]“

My previous automation generates datasets on Tuesday around 5 PM, so for this one, I’ll schedule it a bit later: “run this every Tuesday at 5:10 PM.”

Cowork will ask for confirmation. Just press enter.

Your scheduled task has been created!

To view and manage all your scheduled tasks, open the left sidebar and select “Scheduled.”

Important note: Scheduled tasks only run while your computer is awake and the Claude Desktop app is open.

Congrats! Now you know how to schedule tasks in Claude.

Check out the guide below for more Cowork use cases.

Final thoughts

• Not everything should run on autopilot. Review your chat history and identify which tasks actually benefit from being scheduled

• You can schedule as many tasks as you like (Claude will happily work in the background), but each one draws from your daily usage allowance. The $20 Pro plan may be enough if you’re running a few lightweight tasks, but heavier workflows might call for the $100 or $200 tier

• If this was helpful, consider becoming a paid subscriber. It helps me move up the Substack rankings (like my fellow writers!) and keeps these guides coming

I’m thinking of making a guide packed with Claude and Cowork use cases. If there’s a specific one you’d like covered, drop it in the comments.

Click here to read the best Claude guides

Subscribe now

You know Claude is best at following instructions.
You want to try Claude’s skills and Cowork

But you have too much built up in ChatGPT to just walk away. You don’t want to lose all your chats, memory, custom instructions, etc, so it feels like you’re trapped.

In this guide, we’ll fix that.

In a few minutes, you’ll learn the best ways to move from ChatGPT to Claude:

• Option #1: Transfer ChatGPT memory to Claude (with Anthropic’s prompt)

• Option #2: Transfer ChatGPT projects, GPTs & custom instructions to Claude

• Option #3: Transfer all your data from ChatGPT to Claude

New to Claude? Click here to read our best Claude guides

Subscribe to get my upcoming Claude guides right in your inbox 👇

Subscribe now

Option #1: Transfer memory to Claude (Anthropic’s prompt)

Anthropic just released a method to transfer your preferences and context from other AI providers to Claude.

You only have to follow two steps.

Step 1: Copy Anthropic’s prompt below and paste it into ChatGPT

I’m moving to another service and need to export my data. List every memory you have stored about me, as well as any context you’ve learned about me from past conversations. Output everything in a single code block so I can easily copy it. Format each entry as: [date saved, if available] - memory content. Make sure to cover all of the following — preserve my words verbatim where possible: Instructions I’ve given you about how to respond (tone, format, style, ‘always do X’, ‘never do Y’). Personal details: name, location, job, family, interests. Projects, goals, and recurring topics. Tools, languages, and frameworks I use. Preferences and corrections I’ve made to your behavior. Any other stored context not covered above. Do not summarize, group, or omit any entries. After the code block, confirm whether that is the complete set or if any remain.

After a few seconds, you’ll get the information ChatGPT has been collecting about you in its memory across every chat you’ve had: who you are, what you do, what you like/dislike, and other preferences.

Step 2: Copy and paste the results into Claude’s memory settings.

Done! Your ChatGPT memory has been transferred to Claude

But that’s only the memory!

If you have other important information in your custom instructions, projects, GPTs, and chats, options 2 and 3 will come in handy.

Subscribe now

Option #2: Transfer GPTs, Projects & custom instructions

This is my favorite way to move from ChatGPT to Claude.

Why? Because most of us don’t need every single thing we’ve shared with ChatGPT. If you go with option #3, you’ll get a massive .zip file with all your ChatGPT data. There are workarounds to connect this file to Claude, but they come with some downsides:

• ChatGPT might’ve saved inaccurate info about you in its memory

• Both useful and useless chats will get exported (which might confuse Claude later)

For most users, transferring the following will be enough:

GPTs & Projects → Claude Projects

If you’ve created your own GPT or project and want to transfer it to Claude, there’s an easy way to do it: transfer its instructions to a Claude project

1. Go to your GPT or Project

2. Copy the GPT/Project instructions

3. Download the GPT/Project knowledge files

4. Go to Claude → Projects (left bar) → New Project

5. Paste the instructions and upload the files

If you want to transfer your project’s memory, do this:

1. Go to your ChatGPT project

2. Use this prompt “Based on everything discussed in this project, create a comprehensive summary and context“

3. Copy the response

4. Go to your Claude project

5. Click on the + inside Files → Add text content → Paste the response → Click “Add content”

ChatGPT Memory / Custom Instructions → Claude Memory

You can quickly transfer ChatGPT’s memory and custom instructions to Claude. Here’s how.

1. Click on ChatGPT settings

2. Go to Personalization

3. Scroll down to Memory → Manage (or Custom Instructions → copy text)

4. Copy your saved memories (tip: read the memory and verify that the info about you is accurate. If something doesn’t sound quite right, delete it)

5. Go to Claude settings → Capabilities → Turn on Memory

6. Open a new chat → Type “Update your memory about me with this“ → Paste the saved memories (or custom instructions)

That’s it!

Now Claude has the full memory of what ChatGPT has learned about you over the years. If you want to review or edit your memory in Claude, go to Settings → Capabilities → Memory → Memory from your chats

📚 If you’re new to Claude, you might find these guides useful:

• Best Claude features: artificialcorner.com/p/claude

• Don’t use Claude wrong: artificialcorner.com/p/claude-levels

• How Claude Cowork saves me time: artificialcorner.com/p/cowork

Subscribe now

Option #3: Transfer all your data from ChatGPT to Claude

This method involves exporting your ChatGPT data and uploading it to Claude.

1. Go to ChatGPT settings

2. Select “Data controls”

3. Click on export data

After some hours, you’ll get an email with your ChatGPT data (I got mine after 24 hours)

You’ll get a zip file. Download it and unzip it.

The data export from OpenAI is big (mine was 3.5GB). Here are the two best ways to connect it to Claude:

• Create a Claude project and upload only the chat.html file

• Point Claude Cowork to the whole folder

#1 Upload only the chat.html file to a project

chat.html is a smaller file inside the data export that contains our chat history

We can upload this file to a Claude project. The only issue is that project files have a 30MB limit in Claude.

If your chat.html file is under 30MB, do this:

1. Go to Claude → Projects (left bar) → New Project

2. Click on the + inside Files and upload the chat.html file

If your chat.html file is over 30MB, do this:

1. Open the chat.html file

2. Select all the text (Ctrl + A) and copy everything

3. Go to Claude → Projects (left bar) → New Project

4. Click on the + inside File → Add text content → Paste the text → Click “Add content”

After following the steps, Claude will have knowledge of your ChatGPT chats.

#2 Point Claude Cowork to the whole folder

Cowork is a tool available in the desktop app. It’s different from regular Claude.

One key difference is that Cowork can access your local files directly (without manual file uploads).

Although Claude Cowork can’t connect to your Claude memory or projects, it’s still useful when you want to find specific information in your ChatGPT data.

Just download Claude for desktop, install it, and open the app.

Choose Cowork and give it access to the folder that contains your ChatGPT data.

After that, Cowork will have all your ChatGPT knowledge.

That said, regular Claude still doesn’t have this knowledge. You could ask Cowork to create a blueprint from your ChatGPT data and then upload it to Claude's memory. If you need that, here’s a Notion repository where you can find a long migration prompt that generates that blueprint.

If you have any questions, let me know in the comments!

And don’t forget to check out our best Claude guides

The people getting the most out of Claude Code aren’t better at prompting.

They build systems.

If you use AI today the way most people do, you probably follow the same loop:

Open a chat → write a prompt → fix the output.

Then the next day, you start from scratch.

You’re not building an assistant. You’re improvising. Over time, that has a real cost:

• Results become inconsistent

• The tone shifts from one request to the next

• You keep re-explaining things you thought were already settled

Claude Code is different.

Instead of a chatbot that just replies, it works as an environment that can read files, run tools, generate documents, and improve outputs — all inside a single project.

That shift from chat to project is what stops everything from restarting at zero every time you open it. This is also why building a CLAUDE.md file matters (think of it as the project’s memory)

You design it once, and it shapes every interaction with Claude from that point on. You get consistent judgment, the same response structure, and clear boundaries.

By the end of this guide, you'll have:

• A practical definition of what CLAUDE.md is and what it actually changes

• A simple checklist for what to write so Claude stops improvising

• CLAUDE.md templates you can copy-paste (and adapt)

• A plug-and-play, interview-style prompt that generates your own CLAUDE.md

Our Claude Code series grows every week with a new guide.

Click here to read our best Claude guides

And consider becoming a paid subscriber if you like my Claude guides 👇

Subscribe now

What is CLAUDE.md and why it matters

A CLAUDE.md file is, quite literally, a Markdown file.

Plain text. Simple structure. Designed for writing instructions that are readable and easy to organize.

It is not code. You don’t need to know how to code to use it.

In Claude Code, CLAUDE.md acts as the project’s memory. It’s a special file that Claude reads at the start of every conversation to:

• Load your rules

• Follow established conventions

• Keep context consistent across sessions

The goal isn’t documentation. The goal is to operate.

Claude gets a stable framework — so it stops improvising every time you open a new chat.

Think about the difference between these two scenarios:

• A chat with no memory: each session depends on what you remember to ask for, and how clearly you phrase it that day

• A project with memory: your “way of working” lives in files, not in your head

In Claude Code, that memory isn’t just one file. It’s a hierarchy, and it loads automatically when you start. That hierarchy also determines which rules carry more weight.

One thing worth knowing: CLAUDE.md doesn’t replace Claude Code’s default system prompt. It gets added as consistent content inside the conversation — on top of the base prompt, not instead of it.

That might sound like a small detail. It’s not.

It means CLAUDE.md is one of the most stable levers you have because:

• It’s present in every session of your project

• It’s written as rules your whole team can read and scale

Subscribe now

What to put in your CLAUDE.md so Claude stops improvising

My recommendation is to keep it:

• As human as possible

• Compact, so it can scale over time

• Focused

This content loads in every session. Every word counts.

Here’s the simple version:

Identity: who you are here and what you are here to do

Defining the role is not a small detail. It’s how you control behavior.

Your identity section should answer three things:

• Role: editor, PMO, manager, writer

• Goal: the outcome you want consistently

• Operating principles: how you prioritize and how you handle uncertainty

The more specific you are here, the less Claude improvises.

Output: a delivery contract that stays stable

Instead of asking “write me a report” every time, define the structure once in CLAUDE.md. For example:

• Summary in at most X lines

• Risks and constraints

• Compared options

• Next steps

In practice, this cuts down revisions. Claude stops picking a new format every time it responds.

Quality: how the work is checked

Claude performs much better when it knows what “correct” means. Give it criteria, expected outputs, or examples. Lines like:

• Include an introduction that previews the value

• Use concrete examples

• Do not invent data — ask for it instead

• Avoid claims you cannot verify

Boundaries: what Claude should not do, even if it sounds reasonable

Claude Code already has a permission system. But your CLAUDE.md should set its own boundaries too — it reduces risk and prevents drift.

Some non-technical boundaries that work really well:

• Do not assume numbers, dates, or policies if they are not in the files

• If critical information is missing, ask before proposing a final plan

• Separate facts from hypotheses

• Avoid recommendations with material impact unless you clearly present risks and trade-offs

Navigation: where the truth lives inside the project

This one surprises people who don’t code.

AI does not guess your context. It reads it.

Claude Code lets you pull in files from CLAUDE.md using @route syntax. This keeps CLAUDE.md from turning into an endless file — and tells Claude exactly where to look.

Instead of pasting three pages of brand guidance in CLAUDE.md, keep a separate file and reference it:

• Brand guide: @docs/brand-voice.md

• Glossary: @docs/glossary.md

• Article template: @docs/article-template.md

This keeps the whole design modular (much easier to maintain long-term)

Example structure:

my-skills/
├─ CLAUDE.md
├─ rules/
│  ├─ rules.md
│  └─ boundaries.md
└─ skills/
   └─ weekly-decision-brief/
      └─ SKILL.md

Here’s a practical rule to keep CLAUDE.md minimal, while still making it scalable:

a) Identity

• Keep in CLAUDE.md: the “who you are here” that should always be present

• Move to context/: long project history, quarterly goals, detailed audience definitions

b) Output

• Keep in CLAUDE.md: the default format you expect most of the time

• Move to skills/: repeatable procedures, or anything you want to trigger by intent instead of running on every request

c) Quality

• Keep in CLAUDE.md: 3 to 7 universal rules that apply across the project

• Move to rules/quality.md: longer checklists, anything you want to version and maintain in parts

d) Boundaries

• Keep in CLAUDE.md: simple, universal project boundaries (usually 2 to 6 bullets)

• Move to rules/boundaries.md: specific policies, or when you want boundaries isolated so they don’t get diluted

e) Navigation

• Keep in CLAUDE.md: high-level pointers about where things live

• Move to context/: glossaries, sources of truth, templates, examples, assets

How to keep CLAUDE.md current without letting it degrade

If you write CLAUDE.md once and never touch it again, you will miss most of its value.

A better approach:

• Review it on a regular basis

• Keep “memory” specific, structured, and easy to maintain

A simple and realistic cycle looks like this:

1. Start with /init to generate an initial CLAUDE.md based on your project. Then edit it so it reflects what the project actually needs.

2. When you notice the same mistake happening again, don’t fix it only inside a one-off prompt. Turn it into a rule or import a reference file Claude can reuse.

3. If the file grows, move rules into a rules/ folder and split them by topic.

4. For anything that doesn’t apply every time, use skills. Skills load only when needed, so they don’t consume context in every session (I rely on this a lot)

One more thing: keep security in mind. Design your CLAUDE.md with clear boundaries, and avoid placing sensitive information inside it.

Subscribe now

CLAUDE.md templates you can copy-paste (and adapt)

CLAUDE.md should be short, readable, and practical. Anything too specific or too heavy belongs in imported files or separate rule files.

Here are three templates you can use:

📄 CLAUDE.md template for decisions and briefs

my-skills/
├─ CLAUDE.md
├─ rules/
│ ├─ rules.md
│ └─ boundaries.md
└─ skills/
 └─ weekly-decision-brief/
 └─ SKILL.md

# Identity
You are my thinking partner for executive decisions.
Prioritize clarity, trade-offs, and risk. Do not fill space with empty opinions.

# How to work
- If critical information is missing to make a decision, ask questions first.
- Separate facts (given) from hypotheses (inferences).
- Do not invent numbers, dates, or policies.

# Output standard: Decision brief
ALWAYS deliver using this structure:
1) Context (5 to 7 lines)
2) Key signals (max 7)
3) Risks and constraints (max 7)
4) Options (min 2, max 4) with pros and cons
5) Open questions (always)
6) Next steps (concrete actions)

# Navigation
- Reasoning rules live in: rules/rules.md
- Boundaries live in: rules/boundaries.md
- Decision outputs live in: skills/weekly-decision-brief/SKILL.md

Why it works:

• It uses specific rules, a clear structure, and imports for heavier context — no bloat in the main file

• CLAUDE.md is read at the start of every conversation, so the restart cost drops to almost zero

📄 CLAUDE.md template for writing articles in your voice

claude-md-article/
├─ CLAUDE.md
├─ context/
│  ├─ audience.md
│  └─ article-goal.md
├─ rules/
│  ├─ output.md
│  ├─ quality.md
│  └─ boundaries.md
└─ skills/
   └─ article-outline/
      └─ SKILL.md

# Identity
You are my editor and content strategist.
Your job is not only to write, but also to shape an article that is useful and actionable.

# Editorial rules
- Avoid generic lines. Use concrete, verifiable examples.
- If something is not supported by a source or a project file, say so explicitly.
- Keep the logic clear: problem, framework, example, application.

# Required structure
- Introduction: sell the article (the promise plus what the reader will achieve).
- Maximum of 4 main sections (not counting the intro).
- Must include at least:
- 1 diagram
- 1 mind map
- 2 applied examples based on real scenarios

# Navigation
- Audience: context/audience.md
- Goal and angle: context/article-goal.md
- Output contract: rules/output.md
- Quality checks: rules/quality.md
- Boundaries: rules/boundaries.md
- Article Outline Skill: skills/article-outline/SKILL.md

rules/output.md:

# Output Contract
- Strong intro: the promise plus what the reader will get
- Max 4 main sections (excluding intro)
- Must include:
- 1 diagram
- 1 mind map
- 2 concrete examples (include "Why it works")

rules/quality.md

# Quality Checklist
- No generic filler
- Every concept has a concrete example
- Separate facts from assumptions
- If something is missing, ask or state the gap

Why it works:

• CLAUDE.md stays lightweight — identity and pointers only, nothing competing with the actual content

• Output and Quality live in rules/, so they’re easy to version and update independently

• The Skill is called only when needed (outline and structure), so it doesn’t eat up context on every request

📄 CLAUDE.md template that turns meeting notes into a plan

Good starting point if the project is new and there isn't much complexity yet.

# Identity

You are my operations assistant.

Your priority is to turn ambiguous information into clear actions.

# Information discipline

- Separate confirmed decisions from open items and risks.

- If an owner or a date is missing, ask for it. Do not invent it.

# Required output: Action plan

1) Meeting summary (max 10 lines)

2) Confirmed decisions

3) Actions (mental table: what / who / when / definition of done)

4) Dependencies and blockers

5) Questions to close gaps

# Boundaries

- Do not assume missing data

- Avoid speculation

# Navigation

If this grows, split it into rules/ and skills/

Why it works:

• It reduces friction at the start. No over-engineering before you need it

• As the system grows, you can migrate cleanly into the same architecture (rules/, context/, skills/)

Generate your own CLAUDE.md with this prompt

There’s a faster and more effective way to create a CLAUDE.md: let Claude Code run an interview with you. At the end of the interview, it’ll generate a CLAUDE.md and every file your project needs.

By answering simple interview questions, you’ll define:

• who Claude is in your project

• how it should respond

• the quality bar it must meet

• the boundaries it must respect

• how it should navigate the rest of your files

Think of it as guided onboarding. You answer once, and those answers give Claude Code enough structure to build your full system.

After that, you never have to explain it again.

Using it is straightforward:

1. Open a Claude Code chat

2. Paste the interview prompt below

3. Answer one question at a time. Take your time (it’s worth it!)

When the interview is done, type:

Now generate my folder structure and all files

Claude returns each file’s content, ready to copy-paste into your project or generate directly in your workspace.

Paste this prompt into Claude Code (67 interview questions)

If you want more guides like this, support us by becoming a paid subscriber today and get full access to our best guides and other perks.

Subscribe now

You are Claude Code acting as an interviewer.

Goal: run a structured interview to generate my project system:

- CLAUDE.md

- rules/output.md

- rules/quality.md

- rules/boundaries.md

- context/project.md

- context/audience.md

- context/glossary.md (if needed)

- context/sources.md (if needed)

- 1–3 skills/*/SKILL.md based on repeatable workflows

- optional: examples/, templates/, assets/

Interview rules:

- Ask ONE question at a time.

- Wait for my answer before continuing.

- If my answer is vague, ask a short follow-up (max 1 follow-up).

- Accept “N/A” and “Not decided yet” without pushing.

- Do not write the files until the interview is complete.

After the interview:

1) Summarize what you learned in 8–12 bullets.

2) Propose a folder structure.

3) Output each file in a separate block labeled with its path, like:

[CLAUDE.md]

…content…

[rules/output.md]

…content…

Then start the interview with Question 1.

Here are the interview questions (do not show them all at once; ask one by one):

A) Identity and purpose

1) What’s the name of this project (one line)?

2) What are you trying to produce in the end?

3) What’s the main outcome you want, in one sentence?

4) What role should Claude play in this project?

5) What are the top 3 responsibilities Claude must handle every time?

6) What are 3 things you never want Claude to do in this project?

7) What should the default tone feel like? Pick 3 adjectives.

8) What tone or style should Claude avoid completely?

B) Quality bar

9) What does a “perfect” response look like here? Name 3 concrete traits.

10) What makes a response bad or unusable? List red flags.

11) What 5 criteria will you use to judge quality?

12) When tradeoffs happen, what should Claude prioritize? Put in order.

13) How rigorous do you want this by default (practical / medium / high)?

C) Output defaults

14) What output format do you want by default?

15) What sections must show up every time?

16) What sections should appear only sometimes?

17) What should the default length be, and when should it switch?

18) What types of examples should Claude include?

19) What kind of ending do you prefer (next steps / checklist / open questions / decisions)?

D) Uncertainty

20) If key info is missing, what should Claude do first?

21) What is Claude allowed to assume without asking?

22) What should Claude never assume?

23) How should Claude label uncertainty?

E) Boundaries

24) What topics are out of scope or off-limits?

25) Any confidentiality/compliance constraints?

26) What decisions should Claude not make for you?

27) Any style boundaries?

28) Any operational limits (tools/links/workflows allowed)?

F) Project context

29) Describe the project context in 8 bullets.

30) What parts of context change often?

31) What parts are stable truths?

G) Audience

32) Who is the end reader/user?

33) Biggest pain points?

34) What are they trying to do quickly?

35) What confuses them?

36) What examples would feel instantly relevant?

H) Vocabulary and sources

37) Mini glossary: 10 key terms + one-line meanings.

38) Trustworthy sources?

39) Sources to avoid?

40) What claims must be backed up or clearly qualified?

I) Navigation

41) Where should rules/context/templates/examples/assets live?

42) What’s the read-first order of files?

43) Which files are required vs optional?

J) Skills (repeat 1–3 times)

44) What’s one repeated task that drains time/energy?

45) Typical input?

46) Ideal output?

47) Common failure modes and prevention?

48) Trigger: when to use / when not to use?

49) 3 success criteria?

50) Want a second Skill? (repeat 44–49)

51) Want a third Skill? (repeat 44–49)

K) Examples/templates

52) Provide 2 real input samples.

53) Provide 2 ideal output samples.

54) What reusable template do you want?

55) Best before/after transformation?

L) Diagrams/mind maps

56) 2 ideas that need a diagram?

57) 2 ideas that need a mind map?

58) Preferred visual style?

59) What should never show up in visuals?

M) What stays in CLAUDE.md vs moves out

60) 5 rules that always apply, no exceptions?

61) Which rules are situational, and what do they depend on?

N) Defaults

62) Default format if user doesn’t specify?

63) Default depth if user doesn’t specify?

64) Default behavior if critical context is missing?

O) Source-of-truth hierarchy

65) If files disagree, what’s the hierarchy?

66) Which files are source-of-truth vs reference-only?

P) Skill triggers

67) What repeated work deserves a Skill, and how will you recognize it?

Start simple. With this interview, you can build a clean, functional CLAUDE.md in minutes.

Then, as your project grows, you can keep moving specific details into rules/, context/, and skills/ without breaking anything.

It’s the best AI I tried so far.

If it weren’t for the short daily usage limit (the $20 plan isn’t enough), I would’ve cancelled ChatGPT a long time ago.

Most people can’t get the most out of Claude because they’ve been using ChatGPT for a long time, so they:

Use Claude like ChatGPT. Like it’s 2023!
Think of prompt crafting, rather than systems
Have a big catalog of prompts that they rarely use

I’ve been there. You’re probably there too, and there’s nothing wrong with that.

ChatGPT was our introduction to AI, so we learned the ChatGPT mindset. In 2026, we need to shift this mindset. In this guide, we’ll do that.

We’ll learn when and how to go from:

• Prompts → .md files

• Prompt libraries → Skills for repeatable workflows

• Chat windows → AI that executes

I’ll break this into levels. Not everyone needs to reach the final level, but I want to give you a sense of what’s possible with a mindset shift.

Click here to read our best Claude guides

Claude is powerful, but most people barely scratch the surface.

Every week, paid subscribers get step-by-step guides just like this one. No tech skills needed. No ads. Just the stuff that actually matters 👇

Subscribe now

Level 1: Prompts → .md files

Since ChatGPT's release, we’ve been obsessed with prompt engineering.

I wrote many guides teaching the best prompt engineering techniques. They’re still relevant. But here’s the thing.

Your best prompts saved on Notion are at the “what to do” level.

Since 2023, the context window and agentic capabilities evolved. Now we can go deep into the “exactly how to do it” level.

This means we can go far beyond ‘Act as a writer with 10 years of experience. Write about XYZ ’

We can teach the AI how to create the article end-to-end: our writing voice, how to structure each section, which words to avoid, the format, etc. The entire workflow becomes instructable, not just the final ask.

All this info goes into an .md file (aka markdown file).

An .md file is simply a text file.

The difference? A .txt file is flat. Markdown gives you headings, code blocks, bold/italic text, and organized lists. This structure helps AI (and us) quickly find sections without having to read the entire file top to bottom.

Here’s what a simplified .md file looks like for capturing your writing voice:

# Blog Post Instructions

## Structure
- Start with a personal story (1-2 sentences max)
- 3 tips only. No fluff. Each tip should have a real example.
- End with one clear takeaway the reader can apply today

## Writing Style
- Short paragraphs (2-3 sentences max)
- Use "you" more than "I"
- No corporate jargon. Write like you're texting a smart friend.
- Never use: "In today's world", "Let's dive in", "Game-changer"

## Formatting
- No bullet points inside the body (use them only for lists of tools)
- Bold only the first sentence of each tip
- Total length: 800-1000 words

Not all your prompts should be turned into an .md file, though.

Think of .md files as recipes: write the instructions once, and the AI follows them every time it runs the task.

There are different ways to create an .md file:

• You can create one manually from scratch (like drafting a brief)

• You can have AI interview you and generate it from your answers

• You can generate them with Claude skills

I spent over an hour answering a 100-question interview designed to capture my writing voice in an .md file. The result is a detailed instruction set for how I write guides. Now, anyone on my team can upload that file and rewrite any draft in my voice.

📚 Here’s a guide with interview-style prompts you can copy-paste to generate .md files for business context, project work, and more: artificialcorner.com/p/md-file

Writing an .md file takes time, but you only do it once. And here’s the thing — you only need to write a few manually (the ones that capture something personal, like your voice or your process).

The rest? Claude skills can generate them in seconds

📚 Move from ChatGPT to Claude: artificialcorner.com/p/switch-to-claude

📚 Best Claude Setup: artificialcorner.com/p/claude-setup

📚 Turn Claude into a 24/7 Employee: artificialcorner.com/p/claude-automates

Subscribe now

Level 2: Prompt libraries → Skills

Skills are Claude’s cheat code.

They’re powerful when you have repeatable workflows.

Instead of re-explaining your preferences and processes in every conversation, skills let you teach Claude once and benefit every time

Let’s take the “brand-guidelines” skill as an example. With this skill, we can generate slides that copy Anthropic brand guidelines in one shot (colors, design, etc):

Skills produce outputs that are more consistent and aligned with our standards

The best part? You don’t have to save skills in your Notion database. No more copy-pasting. Claude triggers a skill automatically when needed.

Here’s how to use skills in Claude

To work with Skills, enable it in settings.

Go to Settings → Capabilities → Turn on Skills

Go to “example skills“ to browse skills created by Anthropic. Turn on any skill you want. Just make sure “skill-creator“ is always enabled (that’s what lets you build your own skill)

To create your own skill, just describe what you need in plain English. Claude takes care of the rest.

Here’s a prompt you can try:

Help me create a skill that applies Google’s official brand colors and typography to any sort of artifact that may benefit from having Google’s look-and-feel. It should be used when brand colors or style guidelines, visual formatting, or company design standards apply

Claude will generate a zip file (that’s the skill). To add your new skill to Claude, do this:

Settings → Capabilities → Skills → Upload the zip file

Once a skill is added, you can either click on “try on chat“ or let Claude trigger the skill when necessary.

📚 Learn more about Claude skills: artificialcorner.com/p/claude

Level 3: Chat windows → AI that executes

Using ChatGPT is like being stuck in a chat bubble.

You type, ChatGPT responds. You upload a file, ChatGPT processes it. Back and forth.

But what if AI could actually do things on your computer? Not just generate text and code, but read your files, run scripts, and build things end-to-end.

That's what Claude Code does.

Claude Code is a command-line tool where Claude doesn’t just answer. It executes!

Claude Code allows us to structure our most valuable tasks into well-defined projects. This means you stop repeating the same instructions over and over again. Here’s Claude Code’s high-level structure to turn raw input into meaningful output.

This might look intimidating, but it’s a combination of many elements we’ve seen so far:

• Remember the .md files from Level 1? Claude Code uses a CLAUDE.md to store your project structure, preferences, and past decisions. It gives Claude context before a session starts (without it, every interaction resets to zero, like opening a fresh ChatGPT conversation)

• Remember the skills from Level 2? Claude Code uses those too. Skills tell Claude how to execute specific tasks.

With Claude Code, we build systems that give us more control over what Claude produces. We create pipelines that work the same way every time.

Many non-technical users are afraid of Claude Code because of the “code“ in its name.

You shouldn’t.

I have a series of Claude Code guides for non-coders: Click here to check them out

I know many of you aren’t technical, so I’ll show you what Claude Code can do using Claude Cowork as an example. Cowork is a simplified version of Claude Code. It was built for non-technical people who are afraid of working with the terminal.

While ChatGPT requires you to be the middleman who gathers and uploads context, Claude Cowork does it itself.

Point Cowork at a folder on your machine and it gets to work. If that folder has your reusable instructions from Level 1, even better. Claude reads them at the start of every session, so you’re never starting from scratch. From there, Cowork can:

• Sort your downloads

• Extract data from screenshots into spreadsheets

• Draft reports from scattered notes

• Execute multi-step workflows on its own

Here’s an example:

This is just one example. For more, check out this guide I wrote.

Here’s how to use Claude Cowork

Download Claude for desktop, install it, and open the app.

At the top, you’ll see 3 options: Chat, Cowork, Code. Choose Cowork. Choose Cowork

Try this prompt:

Click on “Organize files” (from the prompt templates) and select a folder. You’ll get a prompt like this:

Help me organize my Downloads folder. Scan the contents and propose a plan:

• Categories/folders to create

• How files should be sorted

• Any naming conventions to apply

• Files to flag for review or deletion

Show me the plan before making changes. Only proceed after I approve.

After approving the plan, my downloads folder was quickly organized.

Note: If you’re working with important files, back up first. Just like any other AI, Claude Cowork can make mistakes.

📚 My complete guide to Cowork: artificialcorner.com/p/cowork

Note: Sooner or later, you might hit a wall with Cowork. Cowork is a simplified version. Claude Code is the full, unrestricted experience.

So, how to use Claude better than 99% of people?

• Adopt the new AI mindset

• Explore each level. Without realizing it, you’ll build the foundation for level 3

• Instead of crafting the perfect prompt, build systems that give you consistent, high-quality results

• Subscribe so you don’t miss my upcoming Claude guides (and become a paid sub so you always get full access!)

Subscribe now

Click here to read the best Claude guides

Claude Code Skills have changed the way I’ve been working lately.

I have scattered notes, half-read documents, metrics pointing in different directions, and the feeling that something important is there… but it’s still not clear.

It’s not that there isn’t enough information. It’s the opposite.

I could ask Claude to summarize everything. I’ve done that before. The result is usually correct, but not necessarily useful. A summary doesn’t tell you what truly matters, what risks you’re taking on, or what real options you have.

Instead, I started doing something different.

I created a Skill whose job isn’t to “write better,” but to impose order on chaos, the same way, every time.

What’s a Skill? Instead of re-explaining your preferences and processes in every conversation, skills let you teach Claude once and benefit every time

Skills are powerful when you have repeatable workflows

By the end of this article, you’ll learn:

1. When something deserves to become a Skill

2. How to work with skills in Claude Code

3. A Skill that automatically builds new Skills

I’m creating a series of guides on Claude, Claude Cowork and Claude Code.

You can find all my Claude guides here

Subscribe now

1. When something deserves to become a Skill

The Skill doesn’t come first. What comes first is a pattern.

Every time I faced messy strategic notes, I was unconsciously applying the same mental structure:

Clarify the context → Extract key signals → Surface risks → Compare trade-offs → Present structured options → Define next steps

That repetition is what changed everything.

Because repetition is the signal that something deserves structure.

Here are the problems I had before using skills (and what they led to):

• No defined role → Inconsistent responses

• No rules → Shifts in tone

• No boundaries → Unnecessary creative drift

• No environment → Constant repetition

That shift didn’t happen because I created a Skill. It happened because I stopped improvising.

Structure came first. Automation came second.

The hidden step most people skip can be summarized like this:

The difference is massive. In the first case (left side), you automate an idea. In the second case (right side), you automate a proven framework.

A Skill is not a clever idea. It is a structure that has already proven it works multiple times.

Claude Skills on the Web App (why it’s not enough)

This approach is based on a simple idea: instructions can be reused to carry out specialized tasks.

With Skills, you can:

• Build custom workflows

• Save hours on repetitive tasks such as reports and documentation

• Generate content that follows your company’s brand guidelines, chart styles, font selections, visual design standards, and more

In a previous guide, we’ve seen how to easily create and use skills in the Claude web app. That’s the easiest way to work with skills. However, this comes with certain limitations:

• Skills are downloaded as files

• They start to multiply (especially in more complex projects)

• Staying organized can become a headache due to the web’s limited environment

In short, creating a Skill on the web is easy.
Keeping it alive, consistent, and reusable is not.

Enter Claude Code Skills

This Claude Code feature offers a different way to integrate prompts into your project. Unlike the Skills created through the web interface, the difference is not in the responses themselves, but in where the rules are defined and maintained.

The core ideas behind Claude Code Skills are:

• They operate with project-level rules

• They have a clearly defined scope

• They support versioning, which makes them scalable

• They can be applied automatically

For many of us, working within the Claude Code Skills environment may seem challenging at first. However, that perception is far from reality. As you will see later, we can use Claude Code itself to help us build a Skill that generates other Skills.

2. How to work with skills in Claude Code

A Skill in Claude Code is simply a structured file inside your project.

Here is how you create one.

Create the folder structure: Project folder → Create skills folder → Create a subfolder with the Skill name → Create SKILL.md (or download it here)

my-skills/

├─ CLAUDE.md

└─ skills/

└─ weekly-decision-brief/

└─ SKILL.md

Now that the structure has proven stable, we can encapsulate it.

It is important to understand that what we will be creating is a file that defines the Skill. This file should be structured clearly and include well-defined boundaries so that its purpose and behavior are fully understood.

A practical and reliable Skill framework should include the following elements:

This ensures that the structure remains flexible enough to support both simple and complex Skills.

To follow the next steps, download the files CLAUDE.md and SKILL.md:

Click here to download the files

Let’s analyze the SKILL.md file:

🧭 Identity and Scope: Frontmatter, Purpose, When to Use, When NOT to Use

• Clearly defined scope

• Context-based activation, not improvisation

• Explicit boundaries

🧠 Reasoning Discipline: Rules, Inputs

• Separates facts from assumptions

• Forces trade-offs to be made explicit

• Handles incomplete information without fabricating

🏗 Repeatable Structure: Output Structure, Definition of Success

• Fixed order

• Mandatory sections

• Evaluable outcome

This is not about automating a response. It is about automating a stable structure.

Perfect. Now let’s put it into action inside the Claude Code environment.

First, install Claude Code and VS Code, and then install the Claude extension. For details on the installation, check out this guide.

Then, select a working folder inside VS Code and paste the files you downloaded before. VS Code should look like this:

The folder structure would be organized as follows:

my-skills/

├─ CLAUDE.md

└─ skills/

└─ weekly-decision-brief/

└─ SKILL.md

This structure is intentional. It separates identity, defined in CLAUDE.md, from capabilities, contained in the skills/ directory. As a result, you can add new Skills without disrupting what is already working.

In this case, we are developing the weekly-decision-brief Skill, which is designed to turn scattered notes into a clear brief that supports better decision-making.

Now let’s test this Skill.

We will use the following prompt:

Turn these notes into a Weekly Decision Brief.

- Considering hiring a Head of Sales
- Revenue flat last 3 quarters
- Current team = 3 junior reps
- Cash runway = 8 months
- CEO feels we need “strong leadership”
- No clear sales process documented
- Investors asking about growth plan

Here are the results we got:

The result is not just a well-written answer.

It is a fixed structure:

1. Context Summary

2. Key Signals

3. Constraints & Risks

4. Options

5. Open Questions

6. Next Steps

The order does not change.
The sections do not disappear.
The reasoning pattern stays consistent.

That stability comes from the structure defined in SKILL.md.

What matters here is visible:

• Sections are mandatory

• Order is fixed

• Trade-offs are explicit

• Missing information is surfaced

• Decisions are structured

This is not a better prompt.

It is enforced reasoning discipline.

While most AI tools begin each request from scratch, a Skill applies the same predefined structure consistently.

Subscribe now

3. A Skill for Creating New Skills

This is not a Skill meant for daily use. It is activated only when we want to design a new skill, ensuring that it has a clear purpose, defined boundaries, and a solid structure before implementation.

Why does this matter?

Because it helps prevent:

• Creating unnecessary Skills

• Duplicating existing capabilities

• Adding unnecessary complexity to the system

• Turning every task into a Skill

In other words, not everything should become a Skill.

With a Skill creator in place, the structure would look like this:

my-skills

├── CLAUDE.md

└── skills

├── weekly-decision-brief

│ └── SKILL.md

└── skill-architect

└── SKILL.md

Now we will create a new Skill. For that, we will use the following prompt:

I want a skill that turns messy meeting notes into structured action plans with clear ownership and deadlines.

With this, we will have created a new Skill for our project:

Now that we have added our new Skill, let’s test it. To do so, we will use the following prompt:

These are messy notes from today’s meeting:

- Marketing wants launch in March

- Dev says backend not ready

- Maybe split release?

- Someone mentioned budget constraints

- Ana will check numbers

- Need landing page

- Legal review pending

- Follow-up next week?

- Who owns the onboarding flow?

- Customer feedback unclear

Without explicitly telling it which Skill to use, we get the following results:

These results show that:

• The purpose is clearly defined

• The intent is explicit

• The structure encourages additional, meaningful analysis

What we are really doing is designing behavior, and that fundamentally changes the way we work with AI.

We can also test whether our project is robust. To do that, we will use the following prompt:

We need to decide whether to delay launch or do a split release.

Backend not ready, marketing pushing March.

Budget unclear.

What should we do?

As expected, Claude Code was able to identify and apply the appropriate Skill:

With this, we can conclude that our system demonstrates three things so far:

• It can switch Skills based on intent

• It can refuse to make a decision when there is not enough information

• It can generate a consistent structure without being reminded

At that point, this is no longer just prompting. It is behavior design, and from an architectural perspective, it works.

Click here to read the best Claude guides

If you’re already using Claude Code, or thinking about it, you’ve probably felt overwhelmed trying to figure out which components, folders, and files you’re actually supposed to create.

Claude Code is impressive. That’s undeniable. However, if we don’t understand how to organize its internal architecture, we won’t get the results we expect.

It’s not Claude Code’s fault.

Many of us try to solve everything through prompts alone. That approach is not entirely wrong, but it becomes far more effective when it’s supported by a clear structure.

So what should your first project include?

You need to think about the following:

1. A clearly defined role

2. Written rules

3. Explicit boundaries

4. The environment is intentionally designed

At this point, you already understand the central idea. This is the foundation that will support every project you decide to build on your own.

When I first started, I spent hours going through documentation and watching countless tutorials. The reality is that it can be frustrating not to fully understand how projects are structured.

When working with Claude Code, clarity around structure makes all the difference.

That’s why, by the end of this article, you’ll have:

1. Clarity on which files to create first

2. A simple but scalable structure (highly underrated)

3. An understanding of how to think like Claude Code—not just how to use it

This will become your starting point for any project, from the most basic to the most advanced.

I’m creating a series of guides on Claude, Claude Cowork and Claude Code.

You can find all my Claude guides here

Subscribe now

Before you build anything, learn these 3 foundational concepts

.md Files (Markdown)

An .md file is simply a plain text file with a lightweight structure for organizing information.

It is not complex code.
It is not something highly technical.
It is a clean and structured way to write instructions, rules, and context.

An .md file exists so that Claude understands the environment.

And that is the key difference. In a traditional chat, every conversation starts from scratch. In Claude Code, .md files allow context to live inside the project.

Project Structure

Claude Code does not operate on a single file. It reads the entire project.

That project may include different types of files, but the most important elements typically live inside .md files.

How you organize those files into folders is not just a technical detail. It is a way of giving structure to the system.

When you separate rules, context, and content, you are creating clarity.

It is not about having many files. It is about ensuring that each file has a clearly defined purpose.

Context Persistence

In a traditional chat, each conversation begins almost from zero. In Claude Code, however, context remains within the project.

This means the role, rules, and objectives do not need to be repeated in every interaction.

In other words, the system already knows where it is and what it is building.

And when that happens, the experience stops being improvisation and starts becoming architecture.

Consider becoming a paid subscriber, so you get full access to our next Claude guides 👇

Subscribe now

Every Claude Code architecture stands on 4 pillars

You do not need an endless list of concepts. You only need four essential elements:

1. A Defined Role

Claude Code needs to know who it is within your project.

• Is it a strategic assistant?

• A technical writer?

• A data analyst?

• A demanding editor?

If the role is not clearly defined, the system will improvise. When it improvises, its tone, focus, and level of depth begin to shift.

Defining the role does not limit Claude Code. It gives it direction.

2. Written Rules

Rules are not unnecessary restrictions. They are clear criteria. For example:

• Should it ask questions before answering?

• Should it avoid generic responses?

• Should it structure everything in lists?

• Should it be concise or go deeper into analysis?

When rules are written inside an .md file, they no longer depend on your memory.

They become part of the system itself, which significantly reduces inconsistencies.

3. Explicit Boundaries

This is rarely discussed, yet it is critical.

• On which topics should it avoid making assumptions?

• How far can it go when proposing ideas?

• Should it avoid speculation?

• Should it request missing data before proceeding?

Boundaries do not limit creativity. Instead, they prevent misalignment.

4. Environment Design

This is where everything comes together. It is not enough to have a single central file.

The environment should clearly separate:

• Rules

• Context

• Content

• Production

When everything is mixed together, the system loses clarity.

When each element has its own place, the project becomes scalable.

Most importantly, you stop thinking about which prompt to write next and start thinking about how your system is designed.

Claude Code does not produce better results simply because your prompts are longer. It produces better results when the project environment is clearly and deliberately structured.

Subscribe now

The Minimum Scalable Architecture

Once you understand the four pillars, a scalable project structure could look like this:

my-project/
│
├── CLAUDE.md
│
├── rules/
│   ├── rules.md
│   └── boundaries.md
│
├── context/
│   ├── project.md
│   └── audience.md
│
└── skills/
    └── article.md

This is not a complex technical structure. These are simply folders and text files that you can create on your own computer or laptop

In this structure, anything that ends in .md is a text file. The items without an extension such as rules, context, and skills are folders used to organize those files.

Defined role → CLAUDE.md
The role lives in one single place to prevent ambiguity.

Written rules → rules/rules.md
Rules are separated so they do not compete with content.

Explicit boundaries → rules/boundaries.md
Boundaries are isolated to keep them clear and prevent them from being diluted among other instructions.

Environment design → context/ and skills/
The separation into folders makes the system’s structure tangible and organized.

Subscribe now

A System You Can Adapt and Scale

The structure defines order, which is what we have covered so far. However, what truly determines the quality of the outcome is how you define each piece within that framework.

A system that is poorly written, even if it is well organized, will still improvise.

Structure organizes. Content directs.

Now let’s walk through a practical example that shows how structure and content work together inside a real project:

📄 CLAUDE.md: This is where the system’s identity lives

# System Identity

You are a strategic thinking partner embedded in this project.

# Primary Objective

Support high-level decision-making, clarify trade-offs, and structure complex problems.

# Operating Principles

- Prioritize clarity over speed.

- Surface assumptions explicitly.

- Identify risks and constraints.

- Distinguish facts from hypotheses.

- Ask clarifying questions when context is insufficient.

# Tone

Calm, analytical, structured. Never dramatic or exaggerated.

📄 rules/rules.md: This defines how the system reasons

# Strategic Reasoning Rules

- Always separate context, analysis, and recommendation.

- Highlight trade-offs before proposing solutions.

- Avoid generic advice.

- Use structured frameworks when helpful.

- Keep responses concise but deep.

📄 rules/boundaries.md: This sets clear boundaries

# Boundaries

- Do not assume data that has not been provided.

- Avoid speculative claims.

- If direction is unclear, request clarification.

- Distinguish clearly between analysis and opinion.

📄 context/project.md: This is where the project frame lives

# Project Context

This environment is designed for structured strategic thinking.

Focus areas may include:

- Product positioning

- Resource allocation

- Growth decisions

- Organizational design

The goal is disciplined reasoning, not open-ended brainstorming.

📄 context/audience.md:

# Audience Profile

Readers value:

- Clear thinking

- Explicit trade-offs

- Risk visibility

- Practical insight

They prefer:

- Structured analysis

- Direct language

- Minimal noise

📄 skills/article.md:

# Skill: Strategic Brief

## Purpose

Generate structured analysis for complex decisions.

## Output Structure

1. Context Summary

2. Key Signals

3. Risks & Constraints

4. Trade-offs

5. Strategic Options

6. Recommended Direction

## Standards

- No filler.

- No repetition.

- Structured reasoning only.

Spending time defining the content of each file is not a minor detail. The clarity you put into writing these files is the same clarity you will get back in every response, so it is absolutely worth investing time here.

Subscribe now

Final Thoughts

Claude Code is not difficult. What often creates confusion is trying to use it without a system. The difference does not lie in writing longer instructions, but in defining the framework where those instructions live.

You do not need a perfect architecture from day one. You need clarity.

So start simple. Define the role, write the rules, make the boundaries explicit, and organize the environment.

From there, allow the system to evolve and scale with you.

Click here to read the best Claude guides

It’s an amazing tool for automating tedious computer tasks!

I’ve been wanting to share this guide for a while now, but Cowork was macOS-only. No point publishing a guide that half of you can't use.

Today, it was just released for Windows. So here we go!

Here’s what makes Claude Cowork different from other tools:

• It has direct access to your local files (no manual file uploads!)

• Takes on multi-step tasks and executes them autonomously

• Creates professional deliverables (Excel, PowerPoint, Word docs)

• You can add plugins for specialized work (sales, legal, finance, productivity)

Working with ChatGPT is like being stuck in a chat bubble. There’s a lot of back and forth because it has no context about what you’re trying to do.

Claude Cowork is different.

Its context is the folder and files you give it access to.

You point to a folder & give it a task → Cowork plans the steps & executes on its own

By the end of this article, you’ll learn:

• How to use Claude Cowork

• How to save hours with Cowork (ChatGPT can’t do this)

• How to make Cowork even more powerful with plugins and connectors

You can’t ignore Claude in 2026!

So I’m creating a series of guides on Claude, Claude Cowork and Claude Code.

You can find all my Claude guides here

P.S. These first Claude guides cover the basics (free for everyone). The upcoming guides are where it gets interesting: the workflows and use cases that separate casual users from professionals who get real results using AI

The next guides will be exclusive to paid subscribers. Subscribe today and get these welcome gifts 👇

Subscribe now

How to use Claude Cowork

Just download Claude for desktop, install it, and open the app.

At the top, you’ll see 3 options: Chat, Cowork, Code. Choose Cowork.

Note: You might have to turn on Developer Mode on Windows to install and use Cowork: Settings → Update & Security → For developers → Developer Mode

At first look, Claude Cowork looks like the ChatGPT app, but it’s not. Here’s a quick overview of the Cowork interface:

In addition to that, inside the + icon, you’ll find connectors and plugins that make Cowork even more powerful (we’ll review them a bit later).

Let’s see some things you can do with Cowork

Note: If you’re working with important files, back up first. Just like any other AI, Claude Cowork can make mistakes.

Subscribe now

How to save hours with Cowork (ChatGPT can’t do this)

#1 It can organize your messy desktop

Click on Organize files (from the prompt templates) and select a folder.

You’ll get a prompt like this:

Help me organize my Downloads folder. Scan the contents and propose a plan:

• Categories/folders to create

• How files should be sorted

• Any naming conventions to apply

• Files to flag for review or deletion

Show me the plan before making changes. Only proceed after I approve.

After approving the plan, my downloads folder was quickly organized.

Here’s a video demo

You can also type your own prompt. For example, you can rename all your screenshots using this prompt:

Rename all the screenshots in my desktop based on what they are

#2 It can classify hundreds of video files

Next time you come back from vacation, you won’t have to manually organize your video files. Just ask Claude Cowork to do it for you

With the prompt below, you can classify video files by location and orientation (horizontal, vertical).

I’ll give you access to a folder with hundreds of videos taken at [place].

First, separate files by location. One folder for each location. Then, inside of each folder, create a subfolder separating horizontal videos and vertical videos. Finally, rename the files based on the [location] they were taken.

Please confirm what you’re going to do before doing it.

Here’s an example of Cowork organizing video files recorded at Disney World.

#3 It can turn months of receipts into a categorized spreadsheet

Organize my receipts based on the type of expense

Claude Cowork will ask how you want to organize the receipts. You can choose “move into folder,“ create a spreadsheet or both!

Subscribe now

How to make Cowork even more powerful

Plugins and connectors unlock a whole new world of possibilities in Claude Cowork

How to work with plugins

So far, we’ve seen that Cowork is very powerful.

Plugins make it even more powerful. They’re pre-built instruction sets that make Claude good at specific professional tasks like reviewing contracts, writing marketing copy, or doing financial analysis.

Think of them as “expert modes” for different industries.

Here’s how to install and use plugins:

Click on the + icon → Add plugins

By default, you’ll see plugins by Anthropic, but you can also add external plugins. For this example, we’ll install the Data plugin by Anthropic.

Next time you click on plugins, the “Data“ plugin we installed will be listed. Inside it, you’ll find the commands you can use.

For this example, we’ll use the command “build-dashboard,“ which builds an interactive HTML dashboard with charts, filters, and tables.

Let’s say you have a folder with multiple CSV files about your business expenses, revenue, and customer satisfaction.

Use the command “build-dashboard“ → point to your folder → copy the prompt below

Based on the files in this folder, show me how our marketing spend, revenue and customer satisfaction are trending. Interactive, dark theme.

How to work with connectors

Connectors let Claude interact directly with third-party tools like Google Drive, Gmail, Notion, and more to read data or perform actions.

Here’s how to install a connector.

Click on the + icon → Connectors → Manage connectors → Browse connectors

You’ll see dozens of connectors for the web and desktop. Select any of them and click install.

Next time you click on connectors, you’ll see the connector you installed in the list.

Make sure you read what each connector does before installing it. I installed the Read and Write Apple Notes connector, hoping it could help me organize my messy notes, but it turned out that it can only create, read, and update notes.

Important Notes

• Back up your files before working with Claude Cowork

• Make sure you review Claude’s plan before approving

• Cowork consumes more tokens than a regular chat, so use it for complex work

• Read all our Claude guides and become a paid subscriber to get full access to ALL our content (guides, videos, courses) 👇

Subscribe now

Find my Claude and best AI guides here

Claude Code is one of the best AI tools I have used in recent years.

Forget about getting generic answers with ChatGPT.

Claude Code’s real value emerges when you organize your work into projects. It understands your intent, manages the information you provide, and delivers results that are genuinely useful.

The key? Providing as much context as possible.

The more clarity Claude has about what you want to achieve, the better the outcomes will be. Believe me, spending a bit of time setting this up is entirely worthwhile.

Here’s what makes Claude Code different from ChatGPT or Gemini: Claude Code offers significantly more control over the results you want to obtain, leading to

• Fewer errors

• Fewer repetitive responses (I love this one in particular)

• Greater precision

• Better focus

And you do not need to be a programmer to use Claude Code

I’ll say it again: Claude Code isn’t exclusive to programmers!

A lot of people are intimidated by the name Claude Code (and that’s totally understandable), but in reality, Claude Code handles the technical complexity for you.

In this article, you will learn:

• How to understand Claude Code and its core components

• How to install Claude Code using VS Code (the easiest way to use it)

• How to use Claude Code using a real-world example

By learning Claude Code, you’ll be far ahead of the average AI user who relies only on traditional web apps.

Hi! I’m Frank. If you’re new here, welcome! I’m creating a series of guides on Claude, Claude Cowork and Claude Code.

You can find all my Claude guides here.

Subscribe so you don’t miss my upcoming guides 👇

Subscribe now

How to install Claude Code in 4 steps

1. Visit the Claude website → Look for your operating system → copy and paste the installation command into your terminal → execute it.

If you’re new to the terminal, here’s how to open it:

• Windows: press the Windows key, type “CMD” and press enter

• Mac: press Command (⌘) + Spacebar, type “Terminal“ and press enter

After running the command, you should see this message: “✅ Installation complete!“

To verify that Claude was installed successfully, type “claude” and press enter. You should see the image below in your terminal

2. Install VS Code. Note that Claude Code can be used directly from the terminal, but I find VS Code the easiest way to use it

3. Next, install the Claude extension within Visual Studio Code: Open VS Code → View → Extensions → search for Claude → click Install

   

   After installing the extension, click on the Claude icon to log in to Claude

   

4. Finally, open a folder you will use as your working directory inside Visual Studio Code: File → “Open Folder” → Select a folder

Subscribe now

What Claude Code is and what it is not

After spending several hours working with Claude Code, I found myself wondering why I had not adopted it earlier. Claude Code allows you to structure everyday tasks into well-defined projects, which eliminates the need to repeat the same instructions over and over again.

Many people assume that Claude Code is designed exclusively for programmers.
This assumption, however, is incorrect. As you will see throughout this article, Claude Code is capable of handling even the technical aspects on its own.

Before diving deeper, let’s look at the high-level structure that explains how Claude Code turns raw input into meaningful output.

a) Input

This is the initial information that Claude needs in order to operate: documents, data, instructions, etc.

b) System Configuration (how Claude thinks)

This layer defines how the system should reason and operate, rather than specifying a particular task to execute. It contains:

• CLAUDE.md: The central file that defines the rules, principles, and criteria Claude follows consistently.

• Context: Structured information that ensures coherence and continuity across tasks.

• Profiles: Structured configurations that adapt the system’s behavior based on the type of work, the audience, or the intended focus.

In practical terms, this functions as Claude’s operating manual.

c) Strategic Reasoning (the reasoning layer)

At this stage, the system applies the previously defined configuration. This layer includes:

• Claude Code: The environment in which structured reasoning takes place.

• Skills: Specific procedures that guide how certain types of tasks are solved within a project.

Here, the system combines the provided input with its global configuration, as defined in CLAUDE.md, along with project-specific instructions. This allows it to analyze, structure, and resolve the requested work effectively.

d) Output

The final results may include: well-organized documentsm clear and concise responses, analytical insights, code, actionable work plans, etc

Setting up Claude Code: the system brain

We’ll need “System Configuration” and “Strategic Reasoning” from the previous structure. Inside the folder selected in step 4 of the Claude Code installation, add the structure below.

my-project/
├── CLAUDE.md
├── context/
│ └── content-context.md
├── profiles/
│ └── audience-profile.md
├──skills/
│ └── SKILL.md

Download the files here (click on the green “Code“ button → download ZIP). We’ll use these files in the next sections.

Viewed inside VS Code, the structure will look like this:

At this stage, the core structure is nearly complete.

Subscribe now

Your first chat in Claude Code

Let’s ask Claude Code a few questions about the files we’ve shared with it.

To open a chat window with Claude Code and have everything ready to start a conversation, click on the Claude logo → select the Explorer tab (top left) → enter your prompt in the chat box.

At this point, everything is ready to start a conversation with Claude Code.

Keep in mind that the environment in which the responses are generated is centered around the files we previously loaded for this project.

I ask Claude Code to explain how it understands the project configuration based on CLAUDE.md and the defined skills, helping verify that the setup is interpreted correctly.

Tell me everything I should know about CLAUDE.md

Here’s its response:

I ask another question to confirm the type of output the system is designed to produce with this configuration.

This confirms that the entire workflow will work correctly (taking into account the tools we’ve provided to Claude Code)

Using skills in Claude Code

Now let’s put everything into action. We’ll assign Claude Code the role of Content Strategist.

We’ll use the Article Outline Skill from the SKILL.md file and the following prompt:

Using the Article Outline Skill, propose a structured brief that clarifies a potential initiative before execution. The goal is to define purpose, scope, key considerations, and open questions—not to produce final recommendations. The audience consists of decision-makers with limited time.

This produces a very specific type of output.

From here, we can take different directions depending on how we want to guide Claude Code. For this particular case, we’ll use the following prompt:

Apply this structure to a hypothetical initiative focused on improving how teams document decisions.

As we can see, the workflow generates a concise initiative summary designed to clarify thinking before execution.

This same approach can be applied to create additional projects for any use case you have in mind.

If you are looking for additional ideas for skills, you can:

• Explore the official Claude Code repository

• Use available templates as a starting point

• If you’re new, read this beginner-friendly introduction to Claude Skills. You’ll find what it is, how to create a skill, and how to use it in the Claude app.

Final Thoughts: What makes Claude Code special

The key difference between Claude Code and traditional chatbots is this:

In ChatGPT or Gemini, the structure, role, and rules usually live inside the prompt and must be repeated each time. With Claude Code, however, this logic is already defined in files like CLAUDE.md and the skills, and it is automatically applied to this specific project. This enables consistent and reusable results without having to restate the context in every interaction.

Subscribe now

Click here to read the best Claude guides

This is an AI tool grounded in your own documents that will save you hundreds of hours when learning new subjects in 2026.

I use it regularly. It’s extremely easy to use, highly effective, and free.

Unlike most AI tools, NotebookLM only works with the data you provide.

You can feed it with: Google Docs, YouTube videos, websites, PDFs, text, audio, Word documents, and more!

This means: no hallucination in responses, full control over sources, and more accurate and verifiable results (it provides inline citations)

By the end of this post, you’ll:

• Know why NotebookLM is the best tool to learn new topics

• Get my 1-hour course to learn anything with NotebookLM

• Learn how to use NotebookLM for school, university, work, and business

and I wrote this article together to share different perspectives on learning with NotebookLM. I’m a self-taught learner who uses it to dive into new subjects and create “experts” across fields to guide my business, while Diana relies on it for complex topics in university and at work.

Subscribe now

What makes NotebookLM different from other tools

Go to NotebookLM → Try NotebookLM → Create new notebook

At first sight, it looks similar to ChatGPT or Gemini

1. In “sources” you can see your uploaded materials (PDFs, docs, etc)

2. In “chat” you can chat with it (just like any other chatbot)

3. In “studio“ you can transform uploaded sources into different content formats

NotebookLM is more powerful than it looks.

The fact that its responses are grounded in the materials you upload makes it less likely to hallucinate about those topics than ChatGPT.

Let me show you most of the things you can do with a simple example.

Let’s say I’m preparing for an exam on Spanish history.

If I just Google things or ask ChatGPT directly, it’s easy to drown in an overwhelming amount of information, like different sources, conflicting facts, and general confusion.

So I chose the source I want to learn from, whether it’s a YouTube video, articles, or even an original museum history book that would be hard to read in just a few days, and simply upload it into NotebookLM.

Based entirely on the sources I’ve chosen, I can generate all of this:

Among them, the two below are exclusive for testing your knowledge:

• Flashcards: Key concepts, definitions, and facts are automatically turned into flashcards. It makes it easier to prepare for exams, presentations, or discussions using your materials

• Quizzes: It generates a quiz based on your sources. You get instant feedback with clear, detailed explanations, so you don’t just see whether you’re right or wrong, you understand why.

The rest can turn your materials into friendly, easier-to-digest content:

• Mind Maps: Helps visually organize information from sources.

It automatically extracts key concepts, topics, and relationships, turning complex materials into a clear, structured map you can explore and expand.

• Audio podcast: The AI generates a conversation between two hosts who discuss the content of your sources, simulating a natural, live discussion!

How is this useful? Hearing different perspectives helps you avoid assuming there is only one “right” answer. By listening to real discussions, you see what experts actually debate and disagree on and when different approaches make sense.

Here’s how to generate an audio podcast with NotebookLM:

Choose a topic, length, and language, and get a podcast in minutes. While listening, tap “Join” to ask questions and get real-time answers, just like in a live conversation.

Subscribe now

• Video overviews: You choose the topic, and NotebookLM generates a video-style lecture where experts walk through key ideas, explain concepts in simple terms, and more

Instead of reading long materials, you can watch and listen as ideas are broken down step by step, making learning feel more natural and easier to follow.

Here’s how to generate a video with NotebookLM:

• Slide deck: It generates clear, well-structured slides that feel cohesive, balanced, and easy to follow. Each presentation is designed with clarity in mind.

• Infographics: It converts your content into clear and easy-to-understand visual summaries

If you love reading or want to help your child enjoy it, try this:

1. Upload a book to NotebookLM

2. Click “Customize infographic”

3. In the prompt, write: Create an infographic focused on the keywords in Chapter 1

I tested it with The Little Prince in Spanish, and the keyword-focused infographic made Chapter 1 easier to understand and a lot more fun!

How we actually use NotebookLM to learn faster

So far, we’ve seen the 80% of NotebookLM that matters most.

Now comes the best part:

1. My NotebookLM course from scratch: A 1-hour video tutorial on how I use it to learn topics faster

2. How to use NotebookLM to build experts in different fields (AI, marketing, real estate, etc) that advise in your career and business

3. How Diana uses NotebookLM to learn complex stuff at university and work (it saves her around 10 hours per week)

Think of that one thing you’ve always wanted to learn. NotebookLM makes it a whole lot easier, more accessible, and interactive.

This is exclusive to paid subscribers. Support us by becoming a paid sub, clicking on the button below :)

Subscribe now

Read more

Clawdbot is an AI assistant that not only chats, but actually does tasks for you 24/7

This feels like a leap into the future. Some have even bought dedicated machines, like Mac Minis, just to run this AI agent (you don’t need to buy one, though)

In this article, we’ll look at:

• What Clawdbot is and why it’s popular

• How to install it

• How you can use it

Fun fact: Clawdbot went viral the same day I published my article “Forget AI Tools. AI Agents Will Automate Your Life in 2026“

Here’s a Google Doc with my best guides (agents, prompting, etc): Check it out here

P.S. Anyone can access the doc link, but only paid subs can read the content behind the paywall. They also get other welcome gifts. Subscribe today 👇

Subscribe now

Clawdbot: What it is and why it matters now

Clawdbot is an open-source AI personal assistant that goes far beyond a simple conversational chatbot. It runs locally on your own devices and, unlike ChatGPT or other AIs that live in a browser tab, Clawdbot runs in the background as a service on your computer and integrates with the chat apps you already use.

But the question we might ask is: why does it matter so much now?

Partly because it combines several recent advances in AI in a very smart way:

• It has full system access (files, shell commands, etc)

• It offers persistent memory

• It’s proactive

All of this while keeping a unified integration where your same assistant lives across multiple messaging platforms at once, with the same conversation and context synced everywhere.

What you can do with Clawdbot

Practically any digital task you entrust it with. It’s designed to execute real-world automations, and early users have pushed their creativity to the limit with surprising use cases:

1. Analyze comments on X and create personalized replies according to a given style

   

2. Organize tasks to run 24/7, such as replying to emails or making changes to their projects on GitHub

3. Many say Clawdbot can become their full-time AI employee

The use cases are diverse and depend heavily on the approach and customization you want to give Clawdbot. Apparently, there are no limits.

Subscribe now

How to install Clawdbot in under 5 minutes

Disclaimer: Ironically, one of Clawdbot strengths is also its biggest risk. For Clawdbot to do tasks for you, you’ll have to grant it permissions. If the AI misinterpreted a command or was compromised by an exploit, it could, for example, delete important files, leak sensitive information, or send inappropriate messages.

What I recommend is using Clawdbot in isolated environments, such as on a secondary machine or an account dedicated to the bot.

Here’s how to install Clawdbot:

Step 1: Go to the Clawdbot page and choose the operating system for your installation.

Step 2: Copy the installation code and paste it into your console.

Step 3: Choose from the available options to connect your AI model.

Clawdbot is an orchestrator that leverages existing models. My personal recommendation is to use Claude.

To get a look at Clawdbot’s skills, use the following command:

clawdbot skills list

Final Note: Although Clawdbot is free, running it isn’t entirely costless. If we need to scale to larger projects, we’ll have to use an API. If we don’t monitor usage, the number of processed tokens could drive costs up quickly.

Thanks for reading!

You can find my best guides clicking here

P.S. Anyone can access the link, but only paid subs can read the content behind the paywall, which is the best part of my guides. Become a paid subscriber today 👇

Subscribe now

A quick search on Google Trends reveals how the worldwide interest in AI agents skyrocketed in 2025.

You can’t ignore AI agents in 2026!

But what exactly are AI agents?

To put it simply, an AI agent can autonomously perform tasks on your behalf.

Say you need to create a presentation. Instead of having AI create text for the slides, you can now give an AI agent a goal, and it’ll do online research (read sites, copy text, organize info, etc) and then create the slides with text, images, design, and other elements.

Many AI agents were released in 2025, but only a few people know about them or how to get real value from them. In this guide, I’ll show you the best AI agents and how to make the most of them.

I created a Google Doc with my best guides (prompting, AI for professionals, etc)

Check it out here

P.S. Anyone can access the doc link, but only paid subs can read the content behind the paywall. They also get other welcome gifts. Subscribe today 👇

Subscribe now

#1 Claude in Chrome: Web Automation

Claude is an AI chatbot just like ChatGPT or Gemini. In 2025, it released a Claude in Chrome extension that has agentic capabilities.

I’ve tried agent mode in ChatGPT and Perplexity and found Claude to be the more reliable.

This extension works directly in your browser. It can automate tasks and take actions on your behalf (clicks, scrolls, navigation, etc). Here are some tasks it can automate on your browser:

• Extract information from sites (my favorite)

• Organize your inbox and calendar

• Fill forms and handle repetitive data entry

• Run multi-step processes across multiple tabs

• Record common browser workflows and let Claude handle them automatically

Anything that involves repetitive web actions can be automated with Claude.

Just click the Claude sidebar while browsing any site, describe the task, and Claude will control your browser until it completes your task. In the example below, I collect data on the top 200 tech newsletters:

If you do this manually (clicks, copy-paste, etc), it’d take you too much time. With Claude, it takes a couple of minutes.

I’ve tried different agents and dozens of tasks. Here’s what I love about Claude’s automation approach:

1. It starts by creating a plan on how it’ll handle your task. You can modify the plan

This is great because other agents just start, and when you come back to check the progress, you realize they got stuck in a step because their approach wasn’t the best

2. If your task involved generating a file, it’ll automatically save it to your computer

If you’ve worked with ChatGPT agent, you know that you might encounter the message “Session expired“ when trying to download a file. This rarely happens with Claude in Chrome

3. It’s more reliable than ChatGPT agent and Perplexity Comet

When it comes to web automation, ChatGPT agent isn’t good (we’ll see what’s good for later). It gets stuck selecting elements, clicking buttons, etc. Perplexity Comet is better, while Claude in Chrome is way better.

An important step in data collection is extracting accurate data (no hallucination, wrong data, etc). Claude is very reliable for this and other web tasks

In case your task involves irreversible actions like making a purchase, Claude will always ask for confirmation.

To try it, just install this Claude extension on Chrome and sign in (you need a paid plan).

Subscribe now

#2 ChatGPT Agent: Slides and Report Automation

Before I said that ChatGPT is bad at web automation. That’s true, but there are other tasks it’s still good at. To see all the tasks it can do, press the + button and select “Agent Mode.“

In theory, ChatGPT agent can generate reports, spreadsheets, presentations, and perform web actions. But, in my experience, it’s only good at generating reports and presentations.

Let’s say you want to build a presentation about AI implementation in a small business. You can simply give ChatGPT agent the prompt below, and it’ll do live research and then build the slides.

Build a presentation about the ROI of AI implementation in a small business. Cover at least these points:

• Case studies

• Market data

• Costs

• Projected returns

I used the same prompt on Claude in Chrome, and it’s less autonomous. It asks for permission every time it goes to a new page (and it clutters your browser with multiple tabs). In contrast, ChatGPT handles everything in its own environment, and it rarely stops until it gets the task done.

Pro tip: If you want to build a report, I’d recommend using ChatGPT’s deep research instead of ChatGPT agent.

Why? ChatGPT Plus users are limited to 40 messages/month in agent mode and 25 messages/month in deep research. Generating presentations is something exclusive to agent mode, while building reports can also be done with deep research.

Subscribe now

#3 Claude Cowork: Automate your folders

Claude Cowork is like an agent mode inside your computer folder.

It can perform tasks on your behalf, such as organizing your downloads folder, renaming the screenshots saved on your desktop, and more!

Just download Claude for desktop (only available for macOS now), install it, open the sidebar, and select “Cowork.“ Then give access to a folder and click on one of the prompt templates.

I clicked on Organize files → Download folders and got this prompt

Help me organize my Downloads folder. Scan the contents and propose a plan:

• Categories/folders to create

• How files should be sorted

• Any naming conventions to apply

• Files to flag for review or deletion

Show me the plan before making changes. Only proceed after I approve.

After approving the plan, my downloads folder was quickly organized.

You can also type your own prompt. For example, you can rename all your screenshots using this prompt:

Rename all the screenshots in my desktop based on waht they are

Here’s a video where Cowork organizes months of receipts into a categorized spreadsheet with monthly breakdowns.

Claude was just released a few weeks ago, and it’s exclusive to paid Claude users. Let me know if you want a complete guide on how to use Claude Cowork + best use cases

#4 Clawdbot: Automate your personal assistant

Clawdbot is an open-source “personal AI assistant” you run on your own computer (or a server) that you can talk to from WhatsApp or any other chat apps.

What makes it different? It’s designed to take actions, not just chat

Here are some of its features:

• Persistent memory: It remembers you, your preferences, and context

• Browser control: It can browse the web, fill forms, and extract data from sites

• Full system access: It can read and write files, and execute scripts

• It’s private: It runs on your computer (Mac, Windows, Linux). You can choose your AI model

After you install it on your computer, you can ask it to clear your inbox, send emails, manage your calendar, check you in for flights, and more!

An X user asked Clawdbot to make a reservation for him. Clawdbot called the restaurant and completed the reservation!

Thanks for reading!

You can find my best guides clicking here

P.S. Anyone can access the link, but only paid subs can read the content behind the paywall, which is the best part of my guides. Become a paid subscriber today 👇

Subscribe now