“Just buy exceptional businesses.”

Sounds easy when you hear it like that..

it’s look like you need only to:

• Open any screener.

• Sort companies by returns on capital…

• buy the top names…

• and wait.

But reality doesn’t work like that.

A 30% ROIC business is a target.

If the moat is weak, competitors arrive.
Margins compress.
and returns normalize.

We’ve seen it many times…

The real challenge is not reading the numbers.

• It’s understanding why the economics are strong

• whether they are actually protected

• and what could quietly destroy them over the next 5–10 years.

That’s where AI became extremely powerful.

Because AI is very good at:
• Gathering scattered information fast
• Mapping competitors and substitutes
• Stress-testing assumptions
• Simulating competitive reactions
• Generating scenarios you may not think about
• Connecting weak signals across an industry

In other words:

AI helps move the analysis from:

“this business looks exceptional today”

to

“what could stop it from staying exceptional?”

In this article, I’ll give you the exact prompt I use to pressure-test whether a “great business” is actually durable.

Watch this video to see what the output looks like and how to read it:

Subscribe now

How I Pressure-Test “Exceptional Businesses” With AI:

I’m not using a Claude Skill for this.

Because I don’t want to overcomplicate the workflow.

For this use case, one detailed prompt is enough.

The key is to run it inside Deep Research mode.

Why?

• Because I want the LLM to gather fresh data…

• read filings…

• compare competitors…

• pull industry signals…

It’s a research task so i will run a Deep Research inside Gemini

But Mostapha, why not Claude ?

You can use Claude too.

Claude is very strong, and honestly not far behind.

But for this specific workflow, I prefer Gemini because it is especially good at going out, gathering sources, and building a detailed research report from fresh information.

That said, don’t overthink the tool.

Use the one you already pay for.

I have the choice because I pay for all of them…

The real asset is the set of questions inside the prompt and the structure it forces the model to follow.

You are not asking:

“Is this a good company?”

You are asking:

• Why are returns high today?

• Why hasn’t competition destroyed them?

• Can new invested capital still earn high returns ?

• What could push the business back toward average in next 5 years?

That’s the value.

How to use it :

• Go to Gemini: Open Gemini and choose Deep Research mode.

• Use Thinking model

• Paste the prompt:Replace {Insert Company} with the company you want to test.

Read more

A friend tells you:

“ it’s a great business.”

It’s in an industry you’ve never touched.

You tell yourself:

“I’ll figure it out.”

So you do what any serious investor does.

You open the annual report.

“TPV growth…”
“Take rate compression…”

Weird jargon.
Bad explanations..

You skip it.

A year later, the stock is up 35%.

You tell yourself:

“It was outside my circle of competence.”

But it wasn’t.

You just didn’t have someone to explain the basics to you.

At 20, Warren Buffett took a train to Washington, D.C. to understand insurance.

He showed up at GEICO’s office on a Saturday.

He met an executive and asked questions for hours.

By the end, insurance clicked for him.

That understanding stayed with him for life…

The good news is that we’re in 2026.

And we have AI:

In this article, I’m sharing my Claude Skill file.

• You upload it once.

• You type the industry you want to learn.

• Claude creates an interactive guide that helps you get the basics fast, so annual reports stop feeling like a foreign language.

Subscribe now

How this Claude Skill helps you learn any industry 10x faster:

Instead of jumping between scattered notes and reports, everything is grouped in one place.

The Skill does this by creating one interactive learning page.

Inside, you get the full learning path:

1. The jargon explained in plain English

2. A clear 1-page summary of how the industry works

3. A deeper report that breaks down the key drivers and economics

4. A short quiz to test your understanding and reinforce what matters

5. Clear next steps on what to read, what to analyze, and how to go deeper

Nothing explains it better than seeing it in action.

Here’s a video where I run the skill live:

What’s inside the Skill

it‘s 1 ZIP file to download and upload into Claude.

But this is not one prompt.

It’s a structured workflow built from how I actually learn a new industry:

1. SKILL.md
This is the playbook.
It tells Claude what to do, in what order, and when each step runs.

2. The reference files
This is the core.

Each file is a tested prompt from my investing process:

• one to decode the jargon

• one to map supply, demand, and cycles

• one to structure the final learning output

Claude runs everything in sequence.

How to use it :

step 1: install the skill once, use it forover

• Download this file:
  

Read more

I don’t.

This is what my investing research process with AI usually looks like:

1. Understand the business:
AI helps me build an of how the company makes money and what drives economics.

2. Decide if it deserves deeper work:
After the overview, I decide if this stock is worth going deeper or not.

3. Read the annual report properly:
AI helps me go beyond the story and into the real disclosures, footnotes, and accounting.

4. Map the risks before conviction:
While I read the report, AI helps me build the downside map in parallel:

• what already looks fragile

• which disclosures deserve a closer read

• what questions I still need answered before conviction forms
  

5. Build the bull thesis :AI helps me turn the research into a clear upside case.

6. Stress-test it hard :AI helps me challenge my own thesis before I trust it.

This article is about step 4.

I built a Claude Skill called Stock Risk Auditor.

You run it with one command.

Claude then spends ~10 minutes doing the hard work in the background.

And let’s be honest: that’s a much better use of AI than asking it for “the next 10x stock.”

The output is a full Stock Risk Audit. Not a recycled “risk factors” section.

It gives you:

• the top risks ranked by importance across the business, the balance sheet, and the external environment

• a final risk label (Low / Moderate / Elevated)

In this article, I’ll show you how it works first.

Then you can take the skill, drop it into Claude, and use it in your next stock analysis knowing exactly what it’s doing.

Let’s dive in:

Subscribe now

How the Stock Risk Audit Works:

If you’ve never used a Claude Skill before, don’t worry.

A Claude Skill is basically a saved workflow inside Claude.

Instead of rewriting the same prompts every time, you install it once and run it when you need it.

That’s exactly what I’m giving you in this article.

How to use it :

1. Open Claude
   Claude Web is enough.
   You do not need Claude Desktop or Claude Code.

2. Start a new chat

3. Click the “+” button and select the skill
   Choose the skill from your saved skills.
   (You’ll need to install it first by uploading my file, but I cover that at the end of this article.)

   

4. Give it the stock name or ticker
   For example:CrowdStrike

Wait 10 minutes( Go do something else, Claude doesn’t need emotional support)

Result: a 7–8 page Stock Risk Audit, structured the way I designed it inside the skill.

My favorite part of the report is the final section: Next Deep-Dive Questions.

For CrowdStrike for exemple, one question was:

Can CrowdStrike keep charging premium prices if Microsoft makes Defender “good enough” inside Windows?

That’s useful.

Because the report doesn’t just tell you “there is risk.”

It tells you where to investigate next.

But Mostapha, how do I know that this is not just AI slop in a nice wrapper?

Fair question.

I’ve been using AI for investing for more than a year.

And from the outside, it’s hard to tell the difference between:

• a random prompt from someone who never seriously analyzed a stock

• and a real workflow built by an investor, tested across multiple stocks.

The difference is the structure behind it:

This is not one giant prompt asking Claude to “find risks.”

It’s a small research system inside Claude:

1. One main instruction file :It tells Claude the job, the rules, the ranking logic, and the final memo format.

2. Three tested investing prompts: These are prompts I had already refined and tested for 8+ months in my investing process.

• Business risk: what could weaken the operating engine

• Financial risk: what could stress cash flow, dilution, or the balance sheet

• Structural / external risk: what outside forces could change the economics

3. One synthesis step
Claude merges the three passes into one ranked Stock Risk Audit.

So the skill is not magic.

It is:

one main instruction file + 3 tested prompts + one final risk memo.

Now that you know how it works, let’s install the skill.

• Download this file:

Read more

Do you still read annual reports?

A question I got from a subscriber this morning.

Short answer: yes..

But not the same way I used to.

You might expect me to say, “Just let AI handle the whole thing.”

Nah. Not from me.

I believe in a hybrid investing process.

Let AI handle the mechanical work faster.
But the judgment still has to be yours

My investing process looks like this:

• Discovery = mostly AI ( detailed in This artcile )

• Annual report analysis = Hybrid AI + manual

• Decision-making = 100% manual

• Follow-up = Mostly AI, with selective manual review

In this article, I’ll show you 3 AI workflows I use to read annual reports faster and find the few details that can change how you see the business.

I ran it on Uber’s annual report.

It surfaced 3 things many investors would likely miss on the first read:

• Uber reported $10B of profit, but a big part came from tax benefits, not from the core business.

• Uber has a $1.8B UK tax issue tied to a court case that could hit earnings.

• Uber’s cash flow looked stronger in 2025 because it recorded more insurance cost than it actually paid in cash that year

That’s exactly twhat I want AI to flag in 5 minutes, so I can spend my time digging into the few points that actually matter.

How to use AI to find the buried details in annual reports:

There are 3 ways to use AI for the same job: read annual reports faster and catch the buried details that can change the investment case.

1) Using NotebookLM for a Fast Scan

Best for: getting a quick first pass and asking follow-up questions.

Trade-off: fast and useful and often the best first step before going deeper, but it can miss some buried details.

2) Gemini Prompt for a Deep Read

Best for: finding more of the buried details that NotebookLM can miss.

Trade-off: stronger output, but it works best when you give it a clear prompt and a manageable filing

3) Claude Skill for a Structured Review

Best for: getting the most reliable and repeatable version of that same job.

Trade-off: the strongest automated output, with a bit more setup upfront.

You can use any of the 3 depending on your experience with AI, your time, and how much depth you need.

But Mostapha which one should you use?

If you’re starting to use AI, start with NotebookLM.

Over time, you’ll find your own default.

Personally, I use all 3 depending on the stock, the complexity, and how deep I want to go.

Now let me show you exactly how to use each one.

Subscribe now

Level 1 :Using NotebookLM

NotebookLM is useful because it stays tied to the source you give it, which makes it great for annual reports.

It runs on Gemini, but in my experience, it’s still less powerful than using Gemini directly when you want deeper work.

Step 1: Add the annual report as a source

• In NotebookLM, click Search the web for new sources

• Type: “Uber latest annual report”

• NotebookLM will suggest a few sources

• Select the official annual report source

• Then click Import

You can also add it manually by uploading the annual report yourself.

Step 2: Surface the buried details with a custom prompt

This is the key part: we are using the power of NotebookLM’s source-based answers and combining it with a strong prompt to crate a rport that shows the details most people miss.

• Go to Reports,

• Choose Create Your Own:This matters because it lets you use your own prompt and force a more useful structure, instead of relying on the default report templates.

Paste the prompt below.

This prompt is built to do 3 things:

• stop generic summarizing

• force focus on the highest-signal buried details

• turn the output into a usable investor memo

What the report flagged:

For Uber, it generated a 7-page report that looked like this:

highlighting things worth paying attention to like:

• reported profit was boosted by a large one-time non-cash tax benefit

• insurance costs increased, which is an important pressure point to watch

These are exactly the kinds of things I want AI to flag quickly: details that can materially change the investment case.

Paste this prompt to generate the custom report:

Read more

They put cow tissue inside the human body.

I said that out loud between sets to my wife.

She froze.

“Shhh… she’ll hear you.”

Our 3-year-old is in “I loves animals” periode right now.

I was very close to creating a family disaster.

“Thanks. Now finish your set.”

My wife said.

In last Year, AI transformed my gym sessions to my first filter for discovering new stocks.

But that’s only a step of my process.

In this article, I’ll show you how I use AI to find more stock ideas:

• The tools I use at each step

• The prompts I use to decide if it deserves deeper work

• And the exact process you can add to your own investing research.

  ---

The Highest-Leverage Step in Stock Research:

If you only take one AI investing use case from me, make it this one:

Use AI to understand a business fast enough to decide if it deserves a real deep dive

This first filter is where the leverage is.

You don’t need to read the last 3 annual reports just to decide if a stock deserves your time.

well, you can…

But you’ll never see enough businesses to find the real winners.

And on the other side, you can’t just open a screener, look at P/E, revenue growth, and margins… and pretend you understand the business.

You need the basics:

• What does this company actually sell?

• Who buys and why?

• What drives growth?

• What’s the competitive edge if there is one?

That used to take a day of reading.

Now AI make that 10X faster ( with the right process)

How my AI first-pass workflow actually works

I already have more stock names than I can realistically analyze.

I keep them in a note app.

Stocks I’ve heard about.
Stocks I’ve seen on X.
Stocks I found in Substack posts, screens, podcasts, or random conversations.

Each step is built to get the most leverage out of AI with the least extra time.

Let me show you how I actually use it.

It happens in 3 steps:

1. Build Context in Parallel

I never start understanding a business at my desk.

If I did, I’d burn too much focused time too early.

So I use the time I already have.

Driving.
Walking.
Training.

That’s where I build the first layer of familiarity with a stock.

So when I finally sit down, I’m not starting cold.

I create an audio podcast from the a selected sources with the prompts I validated.

That way, I hear the business through the lens I actually care about, not generic ESG or macro talk.

During my latest Gym session, I did this with 2 names from my backlog:

• LeMaitre Vascular (the “cow tissue inside humans” company)

• MIPS (They sell a yellow safety layer inside helmets that prevent brain injury)

By the end of the workout, I already had a rough map of both businesses in my head.

Subscribe now

Is this fast ?

Let’s call it 0 extra minutes… and 2 business model snpashot already in my head.

Not bad I think..

2. Refresh the Context

When I go back to my computer.

Now the business is not new anymore.

I’ve already heard the story once.

So I open the 1-pager I already generated.

This page gives me the business model in the most optimized format for a first pass.

(I’ll show you how to create it at the end of the article.)

This is not where I discover the business.

I already did that with the podcast earlier.

By the time I sit down, some time has usually passed.

So this is just the quick refresh before I go deeper.

Time spent so far: ~5 minutes at the desk (after 0-minute parallel context building).

3. Read the Full Report

If the stock still looks interesting, I open the full report I already generated with AI that covers the exact points I want to undsertnat in a business.

So I can focus on the real question:

• What looks strong.

• What looks fragile.

• What might deserve a real deep dive.

This usually takes 10 to 12 minutes.

So in total you need only 15 to 20 minutes at my desk, I usually know if a stock deserves deeper work.

Often the answer is no.

That’s still a win.

• I learned somthing

• I killed a weak idea fast.

• And I kept the pipeline moving.

Now let me show you the exact tools and steps to have the same thing.

Note: Don’t read this and say “that’s nice,” and then never try it.

Whether you copy my process or not, you need something like this in your workflow.

Try it for one week, this will transform your investing process.

How to Build This 3-Step Workflow Yourself:

This workflow only needs 2 tools:

• Any LLM with Deep Research mode

• NotebookLM ( free is ok)

Quick reminder: Deep Research is non-negotiable.

The whole workflow depends on the quality of the reports.

That’s why I use Deep Research, not a normal chat prompt.

It’s the only reliable way to:

• build a real research plan

• scan hundreds of sources

• write the report in the exact format I want

But none of this works without a good, detailed, validated prompt.

I still use Gemini for this step because it still performs best here.

NotebookLM comes in Step 2.

Step 1: Generate the First-Pass Repports

Read more

“Mohnish Pabrai is a scam.”

My friend was dead serious !

I laughed.

No bro… even if he drops the “I was friends with Munger” line 10x per sentence.. he’s a legit investor.

Yes that’s the problem, my freind said:

“He talks like Munger and then buys Metallurgical coal stocks and offshore drillers.”

I laughed.

But I knew exactly what he meant.

My friend analyzes every stock like it should be a compounder.

His bear case is often:

“ROIC is only 15%… pass.”

But every investment thesis is different.

And different investments fail differently.

• A turnaround breaks when the business never turns.

• A compounder breaks when quality fades or you paid too much.

• A cyclical bear case is when peak earnings get mistaken for normal earnings.

What really matters is pattern recognition.

The kind that usually takes years in the market.

• Years of reading filings.

• Years of watching how stories break.

• Years of learning what actually kills returns.

Unless…

Now we have a machine that can compress years of pattern recognition into minutes.

Yes, that’s AI.

In this article, I’ll show you how to use Claude to challange any stock thesis, find the hidden weak points, and generate a bear case report.

By the end of this article, you’ll be able to:

• download a file

• upload it into Claude

• run it anytime you want to stress-test a thesis

But I don’t want you to use it like a black box.

Before you trust the output, you should understand the system behind it.

So let’s break down:

1. How this agent actually works

2. Why it thinks like an investor, not like a random AI prompt from the internet

The Real Cost of Borrowed Conviction:

Reading a great bull case is dangerous.

Because it can make you believe a thesis before you’ve done the work.

You read a strong write-up from your favorite investor.

And before you know the stock, you’re already seeing it through their lens.

That’s the big trap.

Now, Imaine this:

You read a bull case.

One click later, you get the counter-case.

Not a generic risk list.

A real attempt to break the thesis:

• where it’s weak

• what has to go right

• what could break it, and why

So before you believe the story…

you can test it.

Subscribe now

The 4-Layer Thesis Breaker for Claude

The full workflow is built inside Claude.

You give it a stock thesis.

Claude runs the process the same way every time.

• Layer 1: What kind of bet is this really?

• Layer 2: How does this type of stock usually break?

• Layer 3: What does the thesis really depend on?

• Layer 4: What does the evidence say?

That works because the system lives inside a Claude Skill.

Reminder:

A Claude Skill is a reusable workflow you can install inside Claude.

Under the hood, it is a packaged set of instructions, prompts, and reference files that tell Claude what to do, in what order, and how to format the output.

This is where I store the investing logic behind the system.

At the top is one main file: SKILL.md

Think of this as the control center.

It tells Claude:

• what output to produce

• what steps to run

• what order to follow

• and which files to load next

Then Claude opens the references folder.

This is where the investing logic, failure patterns, and research instructions are stored.

Let’s run this system together:

Here is how it works in practice:

• Install Claude Desktop (Cowork) on your computer

• Turn on Extended Thinking

• Use Opus 4.6

• Upload the thesis file:

  For example I’m using the Pershing Square Meta deck

• Trigger the skill
  Type: Run the Thesis Breaker Skill on this thesis

Now let’s break down what it is actually doing.

Share

Step 1: Identify the Investment Type

The first layer figures out what kind of investment this really is.

This matters more than most investors think.

Because if you frame the stock the wrong way, you build the wrong bear case.

This layer classifies the thesis as a:

• growth story

• turnaround

• cyclical

• serial acquirer

• asset play

• regulated / quasi-monopoly

• or a hybrid

Step 2: Find the Failure Mode

Once the system knows what kind of bet this is, it asks the right question:

Why does this type of investment usually fail?

For example:

• a great business bought at a terrible price

• growth that never becomes real economics

• a turnaround that is really structural decline

• peak margins mistaken for durable quality

• acquisition optics hiding weak returns

Claude uses my instructions to find the most likely way this specific thesis breaks.

Step 3: Rebuild the Bull Case

Before you can break a thesis, you need to know what the bull is actually betting on.

That is what this layer does.

It rebuilds the bull case from the real source material.

Then it strips the story down to the few assumptions that actually matter:

• growth stays strong

• margins keep improving

• returns on capital stay high

• the moat holds

• capital allocation stays smart

• the valuation still makes sense

This is critical.

Because most bull cases sound complex.

But most of them only depend on a few key beliefs.

And if one breaks, the whole thesis changes.

Step 4: Break the Assumptions

Now the system goes looking for evidence against those key beliefs.

Claude uses Deep Research to pull in more information from 50+ sources.

It starts with the official documents and trusted sources I validated in the skill reference files.

Then it turns all of that into one final output:

• what the bull is really betting on

• where the thesis is weakest

• what could break it

• what would prove the bear wrong

• and what to watch next

The Result

Once the skill finishes, you get 2 outputs:

• an interactive web page to explore the thesis

• a structured 8 pages bear-case report

Install the Skill

1. Download the file attached below:

Read more

Investors use it all the time.

Quality business.
Quality compounder.
Quality franchise.

Sounds smart.

But it’s often useless.

Hermès and a random Polish retailer can both get called "quality"on X.

It’s usually just a premium multiple looking for a justification.

And to justify it, they reach for vague buzzwords.

Like:

• Network effects

• Economies of scale

• Strong brand

It’s easy to call something “quality.”

It’s much harder to explain:

• What exactly is better VS the market?

• Why should the premuim persist?

• And how does it show up in future cash flows?

Because answering those questions takes real work.

That’s where AI helps.

By helping you test whether the premium is real or just a story by checking years of reports, transcripts, and industry history.

In this article, I’ll show you the only 4 reasons a stock deserves a premium multiple and how to test them with AI.

Does this stock really deserve a high multiple?

Buying a stock at 30X earnings can make you more money than buying one at 6x.

But only if that higher multiple is justified by math.
Not by a fancy story around “quality.”

A stock price is just 2 things:

Stock price = earnings × multiple

That’s it.

Now, Imagine 2 stocks, both earns $1.

• Stock A trades at 30x.

• Stock B trades at 10x.

Same earnings with very different multiple.

Why?

Because the market believes that $1 is more valuable in Stock A.

And that only makes sense if one or more of these is true:

1. $1 of Stock A earnings will grow much more over time:

   If earnings go from $1 today to $3 in a few years, then paying 30x today is the same as paying only 10x those future earnings.

2. $1 in Stock A will last much longer:

   If one business can earn $1 for 20 years and another for only 10 , the first deserves a higher multiple.

3. Stock A will need less reinvestment to keep growing:

   If one business keeps $0.80 of every $1 it earns and another keeps only $0.20, the first deserves a higher multiple.

4. Stock A’s future cash flows are more reliable:

   If one business is very likely to deliver that $1 of earning he deserves a higher multiple.

Let’s break them down.
And see how AI makes each one easier to check.

Subscribe now

The 4-Part Premium Multiple Test

This framework has 4 buckets.
But each bucket is really a group of questions.

And to answer those questions well, you need 2 things:

• Look across multiple years, not just the latest report

• Cross-check management claims against numbers and industry evidence

That’s why this is a perfect Deep Research task.

Because the job is not finding one answer.
It’s tracing patterns across years of filings, transcripts, and industry history.

Right now, Gemini is still the best Deep Research tool for this.
(Sorry Claude. You still win on automation of tasks.)

The output: a detailed report and a score for each bucket.

Here’s the exact prompt structure I use to inject real investing logic into Deep Research.

Bucket 1: More Future Owner Cash Flows

Will this business produce materially more owner cash per share over time than a market-average business?

What the prompt checks:

• Growth runway: how much room is really left?

• Pricing and mix: can higher revenue turn into higher owner earnings?

• Operating leverage: does growth actually flow through?

• Structural ceiling: what eventually caps the growth?

Bucket 2: Longer Duration of Cash Flows

Will this business sustain strong owner cash flows for much longer than a market-average business?

Duration is not “loyalty.”
It means the cash flows are harder to disrupt, harder to replace, and more painful for customers to walk away from.

What the prompt checks:

• Switching costs: what does it really cost to leave?

• Mission-criticality: embedded workflow or nice-to-have?

• Industry structure: does competition stay rational over time?

• Disruption risk: what could make this less relevant in 10 years?

Bucket 3: Lower Reinvestment Drag

Does this business convert earnings into owner cash materially better than a market-average business?

Owner cash is not earnings.
It’s what’s left after the business reinvests what it must to maintain and grow.

Owner Cash = Earnings − Required Maintanance Reinvestment

What the prompt checks:

• Capex burden: how much capital is needed just to stay in place?

• Working capital: does growth absorb cash as the business scales?

• FCF conversion: how much of earnings becomes real cash?

• Capital intensity: what must be spent to defend and grow the business?

Bucket 4: Lower Fragility

This is one of the biggest traps..

A strong brand doesn’t automatically mean strong returns.

If that were true, Aston Martin would be a legendary compounder.

It isn’t.

The real question:
Are this company’s future cash flows more reliable than those of a market-average business?

Less fragile means less likely to break in a downturn, disruption, or industry shock.
That can justify a higher multiple.

The prompt To Run

• Open Gemini

• Use Pro or Thinking mode. Both do a great job

• Enable Deep Research

• Paste in the prompt bellow

Read more

“Wait… did she just mix 2 languages in one sentence?”

My friend looked at me.

“Yeah,” I said.

“That was the plan.”

We’ve spoken both languages with her since day one.

“Seems to be working,” he said..

When children grow up with 2 languages early, the brain adapts quickly.

Both languages become native.

And the best part?

Learning another languages later usually becomes easier...

“Okay Mostapha… cute. But I came here for investing with AI, not parenting.”

That’s exactly how I think about AI and investing.

Most investors are using AI.

But it still feels foreign.

They test random prompts from X, from a tech bro who’s never invested seriously in his life.

They get inconsistent results.

And they never build fluency..

That’s the whole point of my work.

I don’t want you to just use AI.

I want AI to feel native inside your investing process.

That’s what this article is about.

How to become an AI-native investor:

Like learning a language, there are rules you can’t skip:

• You can’t become fluent by using it occasionally.

• You can’t learn everything at once.

• Random use does not create real skill.

• Half-learning creates almost no edge.

• AI fluency has levels, and you have to earn them one by one.

The 3 levels of the AI-native investor:

Subscribe now

Level 1: Native in fast, source-grounded analysis

This is the foundation.

This is the first level because it feels the most natural.

It feels like a normal conversation.

You ask a question.

You get an answer.

That’s how most people started using AI.

But most investors doest it wrong.

The biggest mistake is that they are not doing it in a source-grounded way.

Best tool Today

Today, the best tool for this is NotebookLM.

Why? Because it keeps the conversation grounded in the source material.

You upload the report, transcript, or presentation, and the model works from the company’s actual words.

If you want exemple how to use NotebookLM, read this article:

This is useful when you want to:

• Understand a specific section of an annual report faster

• Clarify what management actually means in plain English

• Understand a segment, geography, or product line without digging manually

• Verify whether your interpretation matches the source material

For example:
If management says margins improved because of “mix and productivity,” you can ask NotebookLM to explain exactly what that means, where it appears in the report, and what likely drove it.

Level 2: Native in deep, structured research

This is where real leverage starts.

You stop asking isolated questions.

And start running real research workflows.

At this level, the edge comes from 2 things:

1. Deep Research
2. The right prompt

Deep Research can read 100+ sources and build a research file fast.

That’s powerful.

But without the right prompt, it usually gives you more information.
Not better research.

The prompt is what turns raw information into usable investing work.

• It tells the model what to look for.

• What to compare.

• What to challenge.

• What to ignore.

And a real investing prompt is not one line.

On average, it’s closer to 1–2 pages.

Structured.

Tested.

Built from actual investing experience.

Best tool today

Today, the best tool for this is Gemini Deep Research.

The biggest value here is:

• You can go from zero to a solid first-pass thesis much faster.

• You can understand an industry before reading every report.

• You can compare the company, competitors, and substitutes in one workflow.

• You can challenge one idea from multiple angles instead of relying on one answer.

• You can turn broad research into something structured and usable.

Pro tip:

Once you master deepreserch with one strong prompt, the next step is to turn it into a research sequence.

That means using multiple prompts for one specific objective.

That’s exactly what I showed in my growth-rate analysis article.

Level 3: Native in AI-powered investing systems

This is the top layer.

This is where you stop repeating yourself.

And start building reusable investing systems.

You stop rewriting the same prompts.

And start packaging what already works.

Build it once.

Use it again and again.

Best tool today

Today, the best tool for this is Claude with Skills.

It turns a workflow you already trust into a reusable system.

The biggest value here is:

• You stop starting from zero every time.

• You turn your best prompts into reusable investing tools.

• You save time on repeat work like earnings reviews, thesis updates, and company breakdowns.

• You build investing infrastructure that compounds.

Now you know the 3 levels.

The real question is: how do you become native in all three?

Let’s dive in.

Here’s the 3-step process.

Read more

I was in my home gym, training.

It was late..

Because when you’re a father, the rule is simple:

You train early.
You train late.
Or you don’t train.

Between sets, my phone lit up.

A message in our investor group:

MSFT looks cheap here.

I thought, yes , it looks cheap compared to its usual valuation.

Then I remembered something.
And I replied:

They’re making it look cheaper…

It wasn’t just Microsoft.

Alphabet.
Meta.
Amazon.

They all extended the useful lives of their servers.

That means:

• lower depreciation

• higher reported earnings

• and a stock that can look cheaper than it really is

Funny timing, right?

AI capex is exploding…

…and suddenly servers last longer.

Just a coincidence, I’m sure.

That got me thinking:

How one small change in the annual report can quietly change the picture.

• A longer useful life.

• A changed definition.

• A moved segment.

• A footnote most people never notice.

If you only read the last 2 annual reports, you probably miss it.

To spot it, you often need to go back 3 or 4 years.

Honestly, who does that deeply?

Until now, it was too slow.

But today, with AI, that excuse is gone.

What used to take hours is now easy to check.

The real work is no longer just reading the latest report.

The real work is comparing what changed across years.

That’s the shift AI makes possible.

So I built Annual Report Reading 2.0 with Claude.

A 3 steps workflow that:

• Reads across years

• Flags what changed

• Surfaces the quiet details

You’re no longer asking AI to “summarize the report.”

You’re asking it to read the way a serious long-term investor wishes he could:

Across years, not just across pages.

Here’s the workflow:

The 3 Step Annual Report Reading System with AI

The idea is simple:

Go faster and deeper at the same time.

But let me be clear:

This is not the first step in the investing reserch process.

Before reading annual reports, you should already understand the basics of the business:

• How it makes money

• What drives revenue

• What drives margins

If you want that first discovery step, read these first:

A fully automated business overview with Claude:

Or a less automated but powerful option: use NotebookLM to get an overview of any stock.

Once the business model is clear this is when reading annual reports matter.

But not one by one.
Across years. At the same time.
Like a time-lapse.

That’s possible with the right AI workflow:

Subscribe now

The system we are building

We’ll build this inside Claude Cowork.

Why?

Because this workflow needs:

• lots of pages

• lots of reports

• and multiple years compared at once

And that means context matters.

That’s where Claude Cowork is required.

We’ll also use Skills.

Skills are just repeatable workflows.

Instead of re-explaining the job every time, you create a system Claude can run the same way again and again.

I built this based on what I actually look for after years of doing the work manually.

You can download all skills at the end of the article.
But first, let me show you how the system works.

Understanding it matters more than just copying it.

The system is designed to:

1. Collect
Pull the latest annual report, grab prior years, and put everything into one clean workspace Claude can read.

2. Read the Story
Compare management commentary across years to spot changes in tone, priorities, and framing.

3. Read the Business Behind the Numbers
Compare the company across years from 2 angles:

• Business evolution: Compare business descriptions, segments, revenue mix, and KPIs to see how the economic engine evolved.

• Accounting / reporting evolution: Compare policies, estimates, and key disclosures to catch what changed behind the numbers and what may only look better on paper.

Let’s build this :

Step 1 : Collect resources

This step allows us to gather all annual reports, transcripts, investor presentations, and other key official documents.

I created a Collector Skill inside Claude that you can download and reuse later.

Its only job is simple:

Pull the latest annual reports, grab as many prior years as available, and put everything into one clean workspace.

To use it:

• Install Claude Desktop / Cowork on your computer

• Turn on Extended Thinking

• Use Opus 4.6

• Select a working folder (where the files will be stored)

• Type:“Use the filing collector skill to Download all the official documents for Copart.”

Let’s be honest: you can always do this manually.
It’s only a small bonus.
The real leverage is in the next steps.

Step 2 :Management Narrative Evolution Reader

This is where the real work starts.

The goal is simple:

• What story did management repeat?

• What changed in tone?

• What disappeared?

• What suddenly became important?

• What got reframed?

• What is management emphasizing now that they barely mentioned before?

How to run it

• In cowork Select a working folder (where the files are stored)

• Type:“Use the narrative drift skill on XXX stock.”

Claude will run the Skill automatically.

Inside the Skill, I already defined:

• the logic

• the analysis framework

• and the exact output format I want

The output gives you 2 things

1. A 5–6 page Word document
An analysis of how management’s narrative changed across years:

• what stayed consistent

• what changed

• what was added or removed

• what became more important

• and where the story may be shifting before the numbers fully reflect it

  

2. A structured interactiv web page that makes the reading easy

Step 3: Business & Reporting Evolution

A big part of reading annual reports over time is spotting the subtle changes that can completely change your interpretation of the business.

This Skill reads all the documents from Step 1 across multiple years at once.

And it does it from 2 angles:

• Business evolution: how the economic engine changed

• Accounting / reporting evolution: what changed behind the numbers

Business evolution

First, it looks at how the business itself evolved.

It compares:

• business descriptions

• segments

• revenue mix

• KPIs

• product / geographic mix

• operating commentary

Accounting / reporting evolution

Then it scans for the quieter changes.

The ones most people skip.

It compares:

• policies

• estimates

• useful lives

• segment definitions

• non-GAAP bridges

• expense classification

• other changes that affect comparability

This is where you catch things like:

• a longer useful life

• a redefined segment

• a new adjusted metric

• a moved expense

Not fraud.

Optical improvement.

The kind of change that can make margins, earnings, or valuation look better than they really are.

How to run it

In Claude Cowork, select the working folder.

Then type:

“Use the Business & Reporting Evolution Reader Skill on XXX stock .”

Claude will run the Skill using the logic and output structure already defined.

The output gives you 3 things:

• A structured memo explaining what really changed

• A comparison table showing the key changes across years

• A clean web page with the key findings

Install the 3 Skills

1. Download the file attached below:

Read more

My wife said it half-laughing.

We were on vication, sitting at the hotel restaurant.

I had started asking the hotel staff questions.

It’s a habit I’ve developed. I always try to understand how any business actually works.

Why do people choose this hotel instead of the one next door?

Where does the real profit come from?

• The rooms?

• The drinks?

• The upsells?

How many bookings do they actually need each night just to cover the fixed costs?

And what makes customers come back again next year?

When I started investing, I was amazed by how deeply top investors can understand a business and how quickly they can do it.

Buffett once said he can decide on a deal in minutes.

Because he has run this mental exercise thousands of times over decades.

Buffett built that ability through:

• 75 years of studying businesses

• Reading thousands of annual reports

• Seeing hundreds of companies from the inside through Berkshire Hathaway

Without that level of repetition, it is very hard to reach the same mastery.

But today we have something Buffett never had.

AI.

And when it’s used with the right system, it can accelerate that learning process dramatically.

Today I’ll show the exact Claude workflow that produces a 16-page deep-dive report on any company in one click.

Not AI fluff.
A real analysis of how the business actually works.

The $1M Bottleneck in Investing

The biggest bottleneck in investing isn’t finding ideas.

it’s the gap between:

“This stock looks interesting.”

and

“I actually understand how this business makes money.”

Done manually, that takes hours.
Sometimes days..

I built an AI Agent to close that gap.

Instead of spending days digging through reports, the system builds a clear deep-dive in minutes.

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

• A clear understanding of how the workflow runs step by step

• a file you can download and use to run the same system every time you want to understand a new company

The goal isn’t to replace thinking.

It’s to help you understand the business much faster.

The Claude Business Analysis Agent

This workflow runs automatically using a Claude Skill.

Before going further, a quick reminder.

What are Claude Skills?

A Claude Skill is a reusable workflow instructions for the AI.

Think of it as a small analysis playbook.

You give a simple input.
Claude runs the process behind the scenes.

Inside the skill are the steps, prompts, and instructions that guide the analysis.

The skill I built runs a four-step system.

Each step builds on the previous one.

Here is the 4 steps process:

You type the company name in Claude.

Behind the scenes, the system runs the workflow:

• Research the business model: how the company makes money, margins, cash flow, and cycles

• Research the industry: value chain, competition, barriers to entry, and profit pools

• Synthesize everything into one structured report

• Fact-check the key numbers and claims

I could have just given you the file to download.

But that’s not the goal.

I want you to understand what’s inside the system for 2 reasons:

1. So you know exactly what the AI is doing before trusting the output.

2. To see the difference between generic AI workflows and one designed specifically for investment research by an experienced investor.

How the system works :

• Install Claude Desktop Cowork on your computer.

• Activate Extended Thinking.

• Use Opus 4.6

• Select a folder to work in

  

• Type the name of a company ( for example: Copart).

• Trigger the skill: type “Run the Business Overview Skill on Copart”

Let the system run the full workflow automatically

What happens behind the scenes

The picture below shows what actually happens once the skill is triggered.

when you type “Run the Business Overview Skill on Copart”

You are simply telling Claude which company to analyze and what skill to use

Then the system starts reading the skill structure.

At this stage Claude is preparing the research plan.

It is not writing yet.

It is assembling the analytical framework that will guide the investigation.

The structure of my skill

At the top, there is the SKILL md file.

It tells Claude what the skill is supposed to produce, how the workflow should run, and in what order each step should happen.

Then Claude opens the references folder.

This is where the real analytical engines live.

Inside that folder, each file contains a deep research prompt ( the same kind of prompt many of you already know from my Prompt Library.)

So again, this is not Claude starting from zero.

It is Claude loading a research system that I already designed in advance.

You can see the sequence happening:

• it reads the skill file

• it opens the reference prompts

• it loads the company research prompt

• it loads the industry research prompt

Then the real research begins searching 400+ sources.

Step 1 : Run the business overview Prompt

The first layer runs a deep research prompt focused entirely on the company.

It analyzes:

• How the company makes money

• What drives revenue growth

• Where margins come from

• How cash is generated

• Which parts of the business are cyclical

If you have my prompt library, you’ve already seen this one.
It’s the same business analysis prompt many of you are already using.

But at this point you might wonder:

“But Mostapha… why not just run this prompt alone?”

You could, and that’s very helpful.

But a business never exists in isolation.

So understanding the company alone gives you only half of the picture.

That’s why the system adds another layer.

The first prompt explains the company.
The next one explains the industry around it.

Step 2 :Understanding the Industry

The second layer runs another deep research prompt.
But this time focused on the industry.

This layer analyzes:

• How the industry works

• Where value is created along the value chain

• Who captures most of the profit pool

• What barriers to entry actually exist

• What drives industry-level growth

This gives context.
Without the industry, the business model is incomplete.

This Wasn’t Possible Before

This is worth pausing on for a second..

Before Claude Skills, you could run both prompts manually.
Try to connect the 2 reports yourself.

It works.

But Skills inside claude make it faster

Instead of manually copying prompts, running them one by one, and connecting the outputs yourself, the Skill turns the whole process into a single workflow.

You type one trigger.
Claude loads the prompts.
Runs them in the right order.
And builds the full analysis automatically.

What used to be a manual process becomes a repeatable research system.

Step 3 : The Business & Industry Primer

The third layer organizes everything.

Instead of scattered notes , the system produces one clear, structured report by combining the 2 prompts:

It explains, in clear sections:

• What the company actually does

• How it makes money

• How the industry works

• Where the economics come from

• What drives growth

• What risks actually matter

The goal is simple: summry and make complex things clear.

Step 4 : Fact-Checking

There is one last layer in the skill.

After the report is generated, the system runs a verification pass.

This layer checks:

• Important numbers

• Key claims

• Financial data

• Industry facts

It flags anything uncertain.
It corrects what needs to be corrected.

The goal: make the output reliable and trustworthy.

Because a confident-sounding report that’s wrong is worse than no report at all.

The Result

Once the skill finishes, you get one clear output:

Before jumping into valuation.
Before building spreadsheets.
Before forming an investment thesis.

Because serious investing always starts the same way:

Understand the business first.

Install the Skill

1. Download the file attached below:

Read more

15 companies you follow report within days.

You open Claude.

One click.

The system runs.

Earnings call is analyzed.
Results are compared against years of company history.
Management language is analyzed against prior quarters.

Minutes later, you know:

• What changes vs. last quarter

• If your thesis strengthens or starts to break

• What guidance says about the next 12–24 months

• Whether margins and unit economics are getting better or worse

• Whether the moat and management credibility are holding or breaking.

This is now possible with Claude.

Here’s a short video showing what the final result looks like (sound on).

I built this system and tested it across multiple companies.

Here’s exactly how you can build the same setup

What we are Building:

An AI earnings analyst.

Not a summary tool.

Not a recap generator.

A structured system that:

• Knows the business

• Knows your thesis

• Knows the history

• Detects changes and generate a memo.

The core of this system is a Claude Skill

What is a Skill?

A Skill is a reusable analytical engine.

You define the job once.
Claude executes it the same way, every time.

Think of it as hiring a specialist.

• One Skill can be an earnings analyst.

• Another can be a valuation checker.

• Another can write investment memos.

Each Skill = one job, executed consistently.

You build it once.

Then you use it forever.

This article comes with a Skill that I have built specifically to monitor earnings.

you can install directly.

You don’t need to build anything.

But Mostapha, you always use Deep Research… so why this?”

Deep Research creates understanding.
The Skill automates the tracking.

This image bellow explains the difference :

Step1: Build the Analyst’s Brain

Claude is useless without context, you need to give it context.

Not summaries.
The real history of the business and industry.

Before earnings season starts, build the knowledge base.

Create a dedicated folder for each company.

If have run it on Uber for this exemple:

Inside that folder, store:

Latest Annual Report:

• Defines the business model, segments, long-term strategy, and key risks.

Last 6–8 Quarterly Filings:

• Enables trend analysis and guidance comparison.

Earnings Call Transcripts:

Investor Presentations

• What management says vs what actually happens.

Your Investment Thesis:

• a short document of 1–2 pages.

This is critical.

If your thesis assumes 20% growth and the quarter shows 17%,
the system must detect the deviation.

Pro Tip (This Is a Big One)

I also upload 2 internal reports that I generated in Gemini Deep Research mode using prompts from my Prompt Library:

• A Business Overview (how the company truly makes money)

• An Industry Overview (competitive structure, substitutes, pricing power)

This gives Claude structural understanding.

So when margins move, it knows why that matters.
When growth slows, it understands whether it’s cyclical, competitive, or structural.

More context → better interpretation.

Take time for this step. You only need to build it once.

Step 2 : Create Your Claude Workspace

Now bring Claude into the process.

You need:

• Claude Pro or Max

• Skills option enabled

That’s it.

1.Install claude

Option A (Recommended for Most Investors) :Use Projects

This works directly inside claude.ai.

No desktop app required.

1. Go to claude.ai

2. Create a Project

3. Upload your company documents inside that Project

Projects store your files persistently.

Every time you run the Skill inside that Project, Claude references those documents.

This is the simplest and cleanest setup.

Limitation:
Projects have storage limits depending on your plan.

For most investors, this is enough.

Option B (Advanced) : Use Claude Code for Local Folder Access

This is my actual setup.

If you want Claude to read documents directly from a folder on your computer:

You must install Claude cowork desktop app .

Claude Cowork allows:

• Direct access to local folders

• Larger research archives

• More flexible workflows

This is optional and more technical but has the best performance.

2.Enable Code Execution

Inside Claude:

• Click your profile

• Go to Settings → Capabilities

• Turn on Code Execution

This allows Skills to run structured workflows.

You only do this once.

3.Install the Skill

The Skill is a file in a specific format.

Inside it are structured instructions that tell Claude exactly how to analyze earnings from an investor’s perspective.

It defines:

• What to check

• What to compare

• How to structure the output

Then Claude follows the same analytical process every quarter.

1. Download the .skill file attached bellow:

Read more

Kodak controlled 90% of U.S. film and 85% of cameras.

In 1996, Kodak generated $16 billion in revenue.

Gross margins on film were near 70%, closer to Hermès than to a typical manufacturing business.

Because every photo required another roll.

Every roll required processing.

To compete, you needed factories, chemicals, labs, and global scale.

Few companies could replicate that system.

Then digital arrived:

• By 2011, Kodak’s revenue was down to $6 billion.

• In 2012, it filed for bankruptcy.

But the desire to capture moments didn’t shrink !

In 2000, the world took 85 billion photos a year. By 2010, that number had surged past 380 billion.

Demand grew.

But the profit pool migrated to a different part of the value chain.

That’s disruption.

Not falling demand.

Shifting economics.

AI may be creating similar shifts today.

If you invest for the long term, the question isn’t:

“Will this company use AI?”

It’s:

“will AI change the economics of this busineess?”

In this article, I'll show you a simple 3-step framework to answer that clearly:

The 3-Step AI Impact Framework

Here’s how it breaks down.

Step 1: The AI Stress Test How could AI permanently weaken this business?

Step 2: The AI Edge Test How could AI permanently strengthen this business?

Step 3: The Verdict On balance, does AI help or hurt?

Three steps. Each builds on the last.

Let me walk you through each one.

Subscribe now

Step 1: The AI Stress Test

Start with the downside.

How could AI permanently weaken this business?

Run one Deep Research prompt to build an AI bear case.

Question flow:

1. What protected this business before AI?
   (Moat. Scarce asset. Switching costs. Margin structure.)

2. Does AI directly weaken that protection?

3. What structural consequences follow?

   • New entrants?

   • Bargaining power shift?

   • Profit pool migration?

4. What specific bear scenarios emerge?

How to run it

• Open Gemini (or your preferred LLM) in Deep Research mode

• Replace [COMPANY] with your target

• Launch the run

• Wait 5–8 minutes

During those 5–8 minutes, Deep Research works like a real analyst:

• It searches live sources (filings, transcripts, industry reports, credible publications)

• Reads full documents ( 100+ )

• Cross-references data across multiple sources

• Then writes a long-form, evidence-backed report following the exact prompt framework

The prompt to use:

ROLE

You are a senior equity analyst tasked with identifying how AI could structurally weaken this company’s long-term business model.

COMPANY

[Insert COMPANY]

OBJECTIVE

Identify structural scenarios through which AI could weaken or invalidate this company’s economic engine.

Focus only on downside structural risk.

Assess whether AI could:

Undermine the core competitive advantage

Reduce demand for the product or service

Weaken pricing power

Increase structural cost intensity

Compress sustainable margins

Lower long-term return on capital

Replace the company’s core function entirely

Trace the economic sequence clearly:

AI change → loss of advantage → revenue weakness → margin pressure → lower long-term returns

Mechanism-driven analysis only.
No feature descriptions.
No mitigation strategies.

EVIDENCE STANDARD

Use :

officiel filings

Earnings transcripts

Official company disclosures

Regulator publications

Recognized institutional research


1) EXECUTIVE SUMMARY (Start Here)

Provide 8–10 concise bullets including:

Core breakdown thesis (1 sentence)

Most exposed competitive advantage

Main revenue vulnerability

Primary margin risk

Replacement risk (Yes / No / Limited)

Probability of structural impairment (Low / Medium / High)

Expected time horizon (Short / Medium / Long)

Then summarize:

The most credible scenario

Whether early signs are already visible

Or whether risk remains purely hypothetical

Keep it direct and analytical.

2) PRE-AI ECONOMIC ENGINE

Explain clearly:

What makes this business strong today

Why customers pay premium pricing

Where profits concentrate

Why it has historically earned high returns

End with:

Which single condition must remain true for this business to continue earning high returns?

3) FIRST-ORDER AI EFFECTS (Direct Impact)

Assess whether AI:

Reduces the need for the product

Automates the scarce skill behind it

Makes alternatives “good enough”

Expands supply faster than demand

Introduces new structural cost burdens

Key question:

Does AI weaken the basic economic reason this product exists?

4) SECOND-ORDER AI EFFECTS (Industry Structure)

Assess whether AI changes:

Entry barriers

Switching costs

Customer bargaining power

Demand sensitivity to price

Profit pool location

Include table:

| Layer | Pre-AI Power | Post-AI Power | Why |

Map clearly:

AI change → power shift → revenue impact → margin impact

5) DISPLACEMENT SCENARIOS (3 Required)

For each scenario include:

AI trigger

Structural shift

Revenue impact

Margin impact

Long-term return impact

Leading indicator

Probability (Low / Medium / High)

Status: Early evidence visible / Emerging signals / Pure hypothesis

Scenarios must be distinct and non-overlapping.

6) STRUCTURAL REPLACEMENT TEST (Mandatory)

Assess full replacement risk.

Answer clearly:

What function does this company perform today?

If AI removes that function, what replaces it?

Does value shift to:

Infrastructure providers?

Platforms?

AI-native companies?

End users directly?

Describe the new industry structure if this company disappears.

Who captures profits instead?

Does this company have structural access to the new layer?

If replacement is plausible, explain:

Adoption trigger

Behavioral shift

Revenue collapse pathway

Probability estimate

Whether early structural signals are already observable

If unlikely, explain why the function cannot disappear.

7) CAPITAL INTENSITY & RETURN DURABILITY

Assess directionally:

Does AI increase long-term reinvestment needs?

Does it reduce operating leverage?

Does it lower sustainable return on capital?

Does it shorten excess return duration?

Focus on structural direction, not percentages.

8) STRUCTURAL VERDICT

Provide:

One-sentence breakdown thesis

Most exposed moat pillar

Core revenue vulnerability

Primary margin risk

Overall probability of structural impairment

Expected time horizon

Classify:

Margin Compressor

Revenue Durability Collapse

Moat Breakdown

Structural Displacement

Structural Replacement Risk

Here’s an example: Adobe, a stock that fell 41% last year as because of AI risk.

the report start by mapping Adobe’s pre-AI profit engine;

Then test how AI changes the rules through direct and second-order structural impacts.

Finally, the report proposes multiple disruption and replacement scenarios for Adobe’s business model.

Step 2: The AI Edge Test

Now flip the lens.

Step 1 asked: How could AI break this business?

Step 2 asks: How could AI make this business stronger?

Some businesses don’t just survive AI. They become stronger

How?

1. They own something AI needs to work. The business has data, relationships, or access that took years to build. AI turns that into better products, faster decisions, or lower costs. A competitor can copy the AI. They can’t copy what feeds it.

2. Customers were already locked in. AI tightens the lock. People stay because AI made the thing they already use work better for them specifically.

The prompt below tests every way AI could structurally strengthen this business:

• Deepen competitive advantage

• Strengthen pricing power

• Increase switching costs

• Improve return on capital

• Extend the duration of excess returns

Replace [COMPANY] with your target. Run it in Deep Research mode:

Read more

It looked like a compounder. It wasn’t..

Through my years in the market, I’ve seen this story repeat again and again.

A company looks perfect on paper:

• High margins

• Strong growth

• “Recurring revenue”

• Great narrative and premium valuation.

Investors call it a high-quality business.

The word “compounder” gets thrown around in every podcast.

Then, slowly, reality catches up:

• Competition creeps in.

• Customers churn.

• Margins compress.

• The stock is down 50% YTD

And everyone acts surprised..

But did the business change overnight?

No.

The real problem is that most investors confuse today’s quality with durability.

Checking today’s numbers is easy.

Any screener can give you margins, growth rates, and revenue mix in seconds.

But real business quality isn’t what a company looks like today.

It’s whether those numbers still hold 5 years from now.

Answering that used to take weeks: filings, transcripts, competitors, industry research.

So people default to snapshots.

Not because it’s right , because it’s fast.

But with AI, that tradeoff disappears.

You can now test business durability in minutes using real sources and verified evidence.

That’s what The 3 Gates Test is.

3 simple questions, each with a ready-to-use prompt, that help you spot businesses that look like compounders but aren’t:

Subscribe now

Gate 1 : If marketing stops, does the money keep coming in?

Gate 2 : Does time make this business stronger ?

Gate 3 :Can this business survive stupidity?

These gates don’t tell if the stock is cheap.

They tell something more important: is this business built to compound ?

Let’s get into it.

Gate 1 :The Sticky Revenue Gate

This step will answer a very importante question:

If this company stopped marketing today… would money still come in?

“Recurring revenue” is one of the most abused terms in investing.

Investors see”Subscription” and their brain shuts off.

They assume it means stability.

But “recurring” doesn’t mean “captive.”

A gym membership is recurring because people pay monthly.

It looks great on paper no ?

But customers cancel all the time.

The gym has to keep filling the bucket because it never stops leaking.

Now think about a company that installs critical equipment in your building.

The Elevator.

No SaaS metrics.

No subscription hype.

But once an elevator is installed, the building is locked in.

Maintenance is mandatory.

Parts are specialized.

Switching is expensive and risky.

No one wakes up and says: “Let’s change elevator providers.”

The customer decided once and never revisits the decision.

That’s captive revenue.

And it’s the foundation of every real compounder.

Use this prompt in Deep Research to:

• Expose whether revenue is truly captive (or dependent on constant re-selling)

• Break down exactly where churn would come from if growth slows

• Force proof using filings + transcripts + disclosed metrics (not narratives)

• Deliver a clear Gate verdict: Pass / Borderline / Fail, backed by the strongest evidence

This prompt will scan for:

• Revenue structure signals (recurring vs repeat vs one-time, contract length, renewal cycles, vulnerable revenue lines)

• Lock-in mechanisms (switching costs, workflow integration, contract clauses, rebid risk, procurement control)

• Hard retention proof (churn, cohort behavior, backlog, pricing increases + customer reaction)

• Downturn behavior & churn triggers (what happened in past recessions + what would realistically break captivity)

Copy and Insert your stock name:

ROLE
You are a senior equity analyst writing an investment-quality memo for a long-term compounder investor.

COMPANY
[INSERT COMPANY NAME]

OBJECTIVE
Assess how captive this company’s revenue truly is
Core question: “If the company stopped all sales and marketing today, how long would revenue keep coming in at roughly the same level?”

EVIDENCE RULES
Use:

Official company filings, annual reports, and regulatory disclosures

Earnings call transcripts and investor presentations

Competitor filings and disclosures (to benchmark industry norms)

Credible sector research and expert commentary (industry associations, regulators, reputable consultancies, major research providers)

First-principles industry reasoning 

Prefer primary sources. Avoid blogs, social media, and unsourced claims.
Cite sources inline: (source, date).
If something is not disclosed, state “not disclosed” instead of guessing.
Clearly distinguish facts vs interpretation.

WRITING STYLE
Write in a MECE memo style (clear buckets, no overlap, high signal).
Be concise and professional. Avoid generic explanations. Use company-specific facts.

ANALYSIS REQUIRED

EXECUTIVE SUMMARY

Captive Revenue Verdict: Pass / Borderline / Fail

3 key evidence points

3 key risks

2–3 realistic churn triggers

One-sentence answer to the core question

REVENUE STRUCTURE

Revenue breakdown (subscription / contract / transactional / project / usage-based)

Recurring vs repeat vs one-time mix

Typical contract length / renewal cycle / purchase frequency

Identify which revenue streams are most vulnerable if selling stops

STRUCTURAL LOCK-IN

How embedded is the product/service in customer workflows?

Switching costs: time, disruption, operational risk

Does lock-in deepen over time with usage/integration?

Are there substitutes that bypass switching costs entirely? (outsourcing, alternative workflow, in-house)

CONTRACTUAL LOCK-IN

Contract terms, renewal clauses, notice periods

Renewal friction: auto-renew vs renegotiation vs rebid

Who controls renewal decisions (users vs procurement vs CFO)?

Does renewal create repricing or tender risk?

PRIORITY LOCK-IN

Mission-critical vs discretionary spend

Budget category (IT, compliance, operations, marketing, etc.)

Evidence of resilience or vulnerability during downturns (if available)

What would realistically cause customers to cut this spend?

CUSTOMER DECISION FREQUENCY

How often do customers actively re-decide to keep paying?

Is this “decide once, pay forever” or “decide every time”?

What triggers reassessment (renewal cycle, budget review, new CIO, platform migration)?

PRICING POWER (AS A CAPTIVITY SIGNAL)

Evidence of price increases and customer reaction

Is pricing power structural or temporary?

Are contracts indexed to inflation or renegotiated manually?

METRIC EVIDENCE

Retention / churn (gross and net if available)

NRR / GRR if disclosed

Cohort behavior, backlog/RPO indicators if relevant

Sales & marketing spend trends

Do not assume high S&M means weak captivity. Distinguish retention-driven vs growth-driven using evidence and management commentary.

SUMMARY TABLE (MANDATORY)
| Captivity Driver | Strength (High/Med/Low) | Evidence (source, date) | What Could Break It |

Drivers must include:

Structural lock-in

Contractual lock-in

Priority lock-in

Decision frequency

Pricing behavior

Metric support

FINAL VERDICT

Rate: Pass / Borderline / Fail

Justify with the 3 strongest evidence points

Identify the biggest weakness

If verdict = Pass, state 2 reasons it is not “exceptional captivity”

FINAL LINE
End with:
Gate 1 Verdict: [Pass/Borderline/Fail] — [One sentence summary]

What to do with the result

Look at the verdict first.

Pass : Move to Gate 2.

Borderline : some stickiness, but gaps. Flag it. Proceed with caution.

Fail :revenue depends on constant selling effort. The bucket leaks. Stop here. No need to run Gate 2 or Gate 3.

Gate 2 : The Compounding Gate

Now that we know whether the revenue is truly captive, the next question is even harder:

Does time make this business stronger ?

Think about GoPro.

In 2014, it looked unstoppable.
The product was iconic. The brand was everywhere. The stock hit $90.

But the business was sitting on a ticking clock.

GoPro made cameras.

And cameras get commoditized.

Every year, smartphones got better.

Chinese competitors got cheaper.

The “cool factor” faded.

GoPro had to keep launching new models just to stay relevant.

By 2018, the stock was under $5.

Because in consumer hardware, time doesn’t build moats.

It destroys them.

This is the gate that separates true compounders from businesses that just look good in a spreadsheet.

What this prompt will uncover:

• Define the core engine and test whether it has actually stayed stable or quietly changed

• Surface the specific disruption risks that could break the model

• Deliver a clear verdict: Stronger / Stable / Weaker

This prompt scans engine stability, moat direction, and economics to see if the business is compounding or breaking

• Copy it

• Insert your stock name

• Run it in Gemini Deep Research

Already included in the Prompt Library (23+ prompts) — free for paid subscribers.

Read more

A message drops in our investor group.

The kind you read twice hoping you misunderstood

“Intellego’s CEO has been arrested for fraud.”

I didn't own the stock.

But several friends were.

Some of them are experienced investors with 10+ years in the market.

Smart people.

Doing real work.

Still caught in a fraud.

Fraud is one way to lose everything. But it’s not the only one.

After years in the market, I keep seeing the same 3 causes behind every permanent loss.

All are avoidable:

1. You’re invested in a fraud or a company heading for bankruptcy.

2. You bought a cyclical stock at peak earnings.

3. You paid way way too much for a good business.

Screen for these 3 before you buy, and you avoid the vast majority of permanent capital losses.

The problem?

Doing this manually takes hours…

AI changes that.

It runs these checks in minutes and catches blind spots your own bias would miss.

I call it The Capital Killer Checklist.

3 prompts.

Each one destroys capital in a different way.

Let’s break them down.

Subscribe now

Note:

I run all prompts in DeepResearch mode.

In your favorite LLM ( today, Gemini ranks best in my tests for deep analysis)

Been around? You already know why DeepResearch.

New here? I explained why DeepResearch matters in this article:

#1: Is This Stock Going to Zero?

Some companies are lying to you.

Some can’t make money without borrowing or diluting every year.

Some have real businesses, but the have debts they’ll never be able to repay.

3 diffrente problems.

Same result: you lose everything.

This is the first test I run. It screens for all 3:

• Signs of fraud: Are there red flags in public filings or management behavior that don’t add up?

• Broken business model: They can't generate cash without raising debt or issuing shares

• Debt sustainability: free cash flow and assets can't cover what they owe.

Run this prompt to have the answer(Gemini Deep Research):

Role

You are a forensic accounting + credit risk analyst writing for a professional investor.

Target

[Company Name]

Goal

Using verifiable sources only, assess:

Fraud / manipulation risk

Broken business model risk (can it generate real FCF without dilution/debt?)

Debt sustainability risk (can FCF/assets support debt, or is survival refinancing-dependent?)

Be skeptical but fair. Flag issues only when supported by evidence.

Sources to Use

Annual reports, quarterly filings, footnotes, cash flow statements

Debt notes / covenant disclosures

Official earnings call transcripts

Regulator records (SEC / AMF / FCA), exchange announcements

Company press releases, court/litigation filings

Insider trading filings

Credible financial news, broker research

Short-seller reports (claims, not proof)

Mandatory checks (search if not in filings)

Auditor changes / qualified opinions / going-concern warnings

Regulatory actions (SEC, AMF, FCA, etc.)

Major lawsuits / investigations

Insider buying/selling activity

Credible short seller reports (if any)

Debt maturity schedule / refinancing events

OUTPUT FORMAT 
0) Investment Summary (start here, max 10–12 lines)

Overall Risk Rating: GREEN / YELLOW / RED

Most likely failure mechanism (1 sentence): what breaks first?

Top 3 red flags (each with evidence)

Top 2 stabilizers (each with evidence)

Final verdict (one line): SAFE / INVESTIGATE FURTHER / AVOID

1) Risk Dashboard (table)

| Area | Rating | Severity (1–5) | Probability (1–5) | Key evidence |

Areas:

Accounting & reporting integrity

Quality of earnings (cash vs profits / working capital signals)

Business model viability

Balance sheet liquidity

Debt maturities & refinancing dependence

Governance & insider behavior

Legal / regulatory risks

2) Quality of Earnings Quick Scan (MANDATORY)

Check these traps and report only meaningful issues:

Net income vs CFO trend (persistent gap?)

Receivables vs revenue (A/R rising faster than sales? DSO rising?)

Inventory vs sales/COGS (abnormal inventory build?)

Recurring “one-time” items (restructuring, impairments repeating?)

Reserve manipulation (bad debt, warranty, inventory write-downs?)

Capitalization behavior (costs shifted into capex/intangibles?)

Working capital explanations (repeated “timing” excuses in MD&A?)

For each issue provide:

Finding

Evidence (page OR link + date)

Why it matters

Severity / Probability

3) Core Findings (evidence only)

For each area, include only meaningful signals:

Finding (1–2 lines)

Evidence (filing + page OR link + date)

Quote (1–2 lines if possible)

Why it matters (mechanism, not theory)

Scores: Severity / Probability

What to monitor (1–3 concrete indicators)

No generic risks.

4) Required Financial Reality Check

Using the most recent data, report:

Net debt

Net debt / EBITDA

Net debt / FCF

Interest coverage

Current ratio + quick ratio

Free cash flow trend (5Y)

Working capital trend (5Y)

% goodwill & intangibles of total assets

Debt maturity schedule (next 3 years)

If missing: Not found / Not disclosed.

5) Debt Sustainability Assessment (key investor questions)

Answer clearly:

How much debt could be covered by 3 years of normal FCF?

Is the company dependent on refinancing? If yes, when does it become critical?

Are covenants disclosed? If yes, how much headroom exists?

If refinancing tightens, what is the first pressure point?

6) Ranked Outputs (mandatory)
Top 7 Red Flags (ranked)

| Rank | Risk | Severity | Probability | Evidence | Early warning trigger |

Top 5 Stabilizers

Evidence-based factors reducing downside risk.

Falsification Checklist

List 5 concrete developments that would materially weaken or invalidate the bear case.

Rules 

No invented facts.

If a claim cannot be sourced, write: “Not found in provided sources.”

Do not invent page numbers or citations.

If using filings: cite document name + page.

If using web: cite source + date.

Separate Fact vs Inference clearly.

The results will point to potential risks like this one:

The goal isn’t to give you a final answer.

It’s to surface the specific points you might have overlooked , so you can go deeper where it matters.

This is a fast check, but an important one.

And what’s better than an AI that reads 200 pages of filings... before you even open the first one?

#2: Are You Buying a Cyclical at Peak Earnings?

This one is sneaky.

You look at a stock and see a P/E of 5.

Looks cheap, right?

It’s not.

The earnings are at a cyclical peak.

Next year, they could drop 50–70%.

That “cheap” P/E of 5 becomes a real P/E of 30 or 50 once earnings normalize.

This is the cyclical trap.

It catches investors every single cycle.

Commodity producers, shipping companies, auto suppliers, chemical companies, construction stocks.. all of them look cheapest right before they crash.

What happens:

• Earnings peak → Stock looks cheap on trailing P/E

• Cycle turns → Earnings collapse 50-80%

• Stock drops 80-90%

You lose money even if the business survives.

The prompt below helps you spot it before it’s too late.

(Like all prompts in this article, it’s already in the Prompt Library , free for paid subscribers)

Here is the prompt to run in Gemini Deep Research:

Read more

I was invited by The Dutch Investors to share how I use AI in my investing process.

What’s inside:

• Where AI truly adds value

• What most investors get wrong

• How I use AI in my own research process

Listen here:

Mostapha

Subscribe now

My wife looked at me. “The candy part?”

“Yes. All of it.”

It was a quiet Saturday night, 8 years ago.

We were watching a Buffett documentary.

Halfway through, he explained why he bought a small candy shop in 1972.

What struck me wasn’t the numbers.

It was his thinking:

• He understood every detail and made it simple

• He knew his limits and never crossed them

• He thought in decades while others panicked quarterly

That discipline. That honesty. That patience.

That was the night I fell in love with investing.

(My wife takes credit for my returns now. "I picked that documentary," she reminds me.)

So I started digging.

Every shareholder letter.

Every Buffett and Munger book.
Every Berkshire meeting on YouTube. All of them. Every year since 1994.

Then I did the same for all top investors

Joel Greenblatt.Aswath Damodaran.Mohnish Pabrai. François Rochon. Chuck Akre…

But it took months

in 2026 with AI the same work can be done 10X faster

only If you have the right system

That’s what this guide is about:

3 steps to extract legendary investor frameworks 10X faster with AI:

Let’s dig in.

The 3-Step Framework

How to Learn from Legendary Investors with AI:

• Step 1 : Extract the Philosophy
  Compress decades of letters, talks, and interviews into a clear, structured strategy.

• Step 2: Break the Strategy Down
  Decompose the philosophy into parts and understand how each one works in practice.

• Step 3 : Explain New Investments
  Understand any new buy using the investor’s past decisions and strategy.
  

  Subscribe now

Step 1 : Extract the Philosophy

Legendary investors don’t have a single idea.
They have a system of beliefs repeated across decades.

The goal of this step is simple:
turn everything they’ve ever said into one clear strategy.

Here’s how to do it, step by step:

1) Run 1 prompt to get the complete framework

Your output is a single document that answers this question:

• What drives their returns (the 80/20)

• What they say “no” to and why

• How they size positions

• The rules they never break

Run 1 Prompt in Deep Research Mode

Reminder :why Deep Research Mode

Regular chat = shallow, generic summaries
Deep Research = reads 100+ sources, cross-checks facts, builds real reports

For this task, you need Deep Research.

• Open Gemini ( or you favorite LLM)

• Select Deep Research Mode

• Copythe full prompt below

• Change only: [INVESTOR NAME] → your target investor

What Happens ?

• Gemini builds a research plan

• Reads letters, books, speeches, case studies

• Generates 10-15 page report

ROLE & CONTEXT

You are a strategy forensics analyst extracting the complete investment philosophy of [INVESTOR NAME].

Your job: explain the process, framework, guiding principles, and investment philosophy they use consistently across all market conditions.

Not a biography. Not recent news. Not performance review.

This is: explaining the timeless operating system behind their decisions—the beliefs, rules, filters, and mental models that drive how they select stocks, manage risk, size positions, and compound capital over decades.

Approach: precision over completeness, evidence over speculation, patterns over anecdotes, critique over cheerleading.

Separate what they say from what evidence shows.

SOURCES

Primary only:
- Shareholder letters (full archive, multiple cycles)
- Books they authored
- Speeches/interviews (strategic explanations, not market commentary)
- Portfolio holdings across 7+ years (patterns, not single moves)
- Documented case studies

Prioritize foundational content over recent news.

Mark (unknown) if unsupported. Mark (inferred) if derived from patterns.

OUTPUT STRUCTURE

1. Core Investment Pillars (80/20)

2-3 beliefs explaining 80%+ of decisions across cycles.

For each:
- What it is (their words when possible)
- Why they believe it works (mechanism)
- Evidence across multiple investments, 5+ years

Focus: what hasn't changed.

2. Investment Checklist

Non-negotiable conditions before investing.

Frame as: "For investment to qualify, [X] must be true."

Cover: business quality, financials, management, industry structure, competitive position.

Back with examples from different periods.

3. Rejection Rules

What they exclude:
- Business types, industry structures, situations, financial profiles

For each:
- Why rejected (structural reason)
- Evidence from statements/portfolio absence
- What this reveals about their edge

4. Portfolio Construction System

Concentration:
- Typical holdings count (across time)
- How conviction drives position size

Sizing Logic:
- Initial position size rules
- When they add
- Maximum position size

Focus: the SYSTEM, not current portfolio.

5. Decision Framework

Buy: entry triggers, valuation approach, timing

Hold: what keeps them in during drawdowns, noise vs signal

Add: conditions for increasing positions

Sell: exit triggers, mistake recognition, evidence of actual exits

Rules and patterns, not isolated examples.

6. Underlying Market Beliefs

What they believe about:
- Market efficiency
- Time horizon advantage
- Risk vs volatility
- Role of valuation
- Business quality vs price
- Human behavior/psychology

Only beliefs supported by sources or clearly inferred from behavior.

7. Core Mental Models

Frameworks appearing frequently, driving real decisions.

For each:
- What it is
- Where it appears
- How it shapes decisions

8. Strategy Classification

Classify based on evidence: value, growth, quality, GARP, etc.

Explain:
- What makes their version different
- Why labels might be insufficient

9. Five-Point Essence

1. Core edge
2. Primary filter (what eliminates most opportunities)
3. Biggest strength
4. Biggest limitation
5. Transferable lesson

10. CRITICAL ANALYSIS: Blind Spots

MANDATORY SECTION

Analyze:

Structural Weaknesses: What opportunities does this miss? When has it underperformed?

Concentration Risks: Hidden correlations? Shared dependencies? What shift harms multiple holdings?

Assumption Fragility: Most vulnerable assumptions? What if edge erodes?

Behavioral Blind Spots: Confirmation bias? How handled mistakes? Stubborn holding evidence?

Market Cycle Sensitivity: When excel vs struggle? Which environments expose weakness?

Use evidence. Be specific.

End with: "An investor adopting this framework should know [blind spot] is the price of [advantage]."

CONSTRAINTS

- Timeless framework > recent activity
- Patterns across cycles > single examples
- Their language > generic labels
- Evidence > speculation
- Critique > cheerleading
- Decision-useful, not promotional

Say so clearly if undetermined from sources.

This prompt has been added to the Prompt Library and is available to all paid subscribers

10 minutes later:
A detailed 10-15 page report covering their complete philosophy.

I did run it on Terry smith and transformed the report into a web page inside Gemini:

His 3 commandments for investing:

The sectors he avoids and why:

A critique of the blind spots in his strategy:

Save the Report for later

Export to Google Docs

• Click “Share & Export”

• Select “Export to Docs”

2) Build your permanent knowledge base (NotebookLM)

You have the Deep Research report.
Now you create a permanent home for everything this investor knows.

One place where all their wisdom lives.
One place you return to whenever you need to learn from them.

What You’re Building:

A single notebook where you can:

• Ask any question about their strategy

• Get answers pulled directly from their letters, speeches, and books

• Track their thinking across decades

This becomes your personal investing mentor.

Create a new notebook inside NotebookLM:

Gather All Sources (Automated)

This is an important step.

NotebookLM only learns from what you upload.
bad sources = bad answers.

The sources you need:

✓ The Deep Research report (from Step 1)
✓ All shareholder letters
✓ Key speeches (transcripts or YouTube links)
✓ Investment case studies
✓ Interviews (transcripts)

These become your permanent learning library.

stat by uploading the Deep Research Report

Take the report you created in Step 1.

1. Open your NotebookLM notebook

2. Click “Add Source”

3. Upload the Deep Research report

That’s your foundation.

Pro tip:
Upload as Word doc (.docx), not PDF.
AI extracts data much better from Word format.

Next: Find Everything Else (Automated)

You can upload all sources manually,that works fine.

But you can also make it automatic.

Use NotebookLM’s Fast Research function to gather everything for you.

Copy-paste this into NotebookLM:

Read more

They celebrate 10% growth.
They get excited at 15%.

But revenue growth doesn’t drive long-term returns.
Earnings growth does.
That’s the compounding engine.

2 companies can post the same top-line growth and still produce different outcomes.

• Company A grows revenue 15%. Earnings grow 15%.

• Company B grows revenue 15%. Earnings grow 25%.

Same growth rate. Completely different outcomes.

The difference is :

• Company A grew by adding people.

  More revenue yes but also more costs, and more complexity.

  Margins stayed flat.

• Company B grew on what it already built with the same cost base.

The difference has a name: operating leverage.

and the same mechanism explains why some stocks collapse 50% in a single year.

When growth slows, fixed costs remain.

Margins compress and earnings fall faster than anyone expects…

So understanding operating leverage isn’t optional.

The good news is that AI can help to do that faster and deeper.
That’s exactly why you’re reading this.

Used correctly, this AI workflow lets you:

• Surface operating leverage patterns across years of data

• Connect margin behavior to real business drivers

• Stress-test what happens when growth slows

• Do in minutes what used to take hours

Let’s dive in.

Subscribe now

What Operating Leverage Actually Means

Operating leverage is simple:

Does this business get better as it grows?

Imagine 2 businesses.

Both start at $10M revenue. Both make $1M profit.

On the surface, identical.

But they’re built differently.

Business A: High Operating Leverage

The product is already built.

The platform exists. The infrastructure is paid for.

Revenue grows from $10M to $30M.

Costs barely move.

Profit jumps from $1M to $19M.

Revenue: 3×. Profit: 19×.

Business B: Low Operating Leverage

Every new client requires more people. More people means more costs.

Revenue grows from $10M to $30M. Profit rises from $1M to $4M.

Revenue: 3×. Profit: 4×.

Both companies are “growing.”

Only one becomes more valuable per dollar of revenue.

The Scale Test:

I built a 3-step AI workflow to answer one question:

“Does this business get better as it grows or just bigger?”

1. Does leverage exist and why?

2. How strong is it?

3. What happens when growth slows?

Step 1: The Cost Structure Test

Question: Does this business have structural operating leverage?

Some businesses scale by hiring or adding more assets.
Revenue up → costs up → margins flat.

Others scale on what’s built.
Revenue up → costs stable → margins expand.

some signals:

• Fixed costs paid

• Revenue grows faster than headcount

• Built-in distribution

This step answers: “If this business doubles, what percentage of costs stay fixed?”

How to run it:

1. Open Gemini or other LLM in Deep Research mode

2. Paste the prompt below

3. Replace [COMPANY NAME]

4. Run

Copy-Paste Prompt:

ROLE
You are an experienced long-term investor focused on business quality and operating leverage.

YOUR JOB
Assess whether [INSERT COMPANY NAME] has structural operating leverage based on its cost structure and scaling mechanics.

OBJECTIVE
Conduct a Cost Structure Test covering at least the last 5–10 years.

Determine whether operating leverage is structural or temporary, and how it compares to industry norms.
EVIDENCE RULES
Use only reliable sources:

French filings:
URD (Document d’enregistrement universel) / Annual report
Semi-annual financial report
Earnings releases + investor presentations
Earnings call transcripts (if available)
Competitor filings for industry comparison (Fnac Darty, peers, etc.)
Mark any non-factual analysis clearly as (inferred)
No blogs, no opinions, no unsupported claims.

REPORT STRUCTURE
I. EXECUTIVE SUMMARY (max 5 bullets)
Operating margin level and 5-year trend
Primary cost structure: fixed-heavy or variable-heavy?
Industry comparison: above, at, or below peers?
Classification: Structural Leverage / Partial Leverage / Labor-Based Growth
The one-sentence verdict

II. THE COST STRUCTURE ANATOMY
A. Fixed vs Variable Split
Major fixed costs (R&D, platform, brand, infrastructure, leases, overhead)
Major variable costs (COGS, fulfillment, commissions, support)
Estimate: % of total costs fixed vs variable
Comparison to competitors

B. The "Already Paid For" Test
What has already been built that future customers get almost for free?
Quantify if possible: investments, years of build
Examples: platform, logistics network, brand, store network, automation

C. Marginal Cost Reality
Cost to serve one additional customer/order
Does marginal cost decline, stay flat, or rise with scale?
Where does it structurally bottom out?

 D. Margin Bridge (Gross Margin vs OPEX Leverage)
Purpose: avoid confusing temporary margin effects with true operating leverage.

Table:

Year | Gross Margin % | SG&A % of revenue | EBIT %
Decompose EBIT changes:
How much comes from gross margin expansion?
How much comes from SG&A leverage?
Identify:
Temporary drivers (pricing cycle, promotions, supply shortages, mix effects)
Structural drivers (automation, scale efficiencies, cost discipline)
Mark uncertain conclusions as (inferred).

III. THE SCALING MECHANICS
A. The "Double the Business" Question
If revenue doubled tomorrow:

What would need to scale (people, warehouses, IT, complexity)?
What would stay largely fixed?
Estimate: what % of costs would double vs remain fixed?

B. Revenue Per Employee
Table: Year | Revenue | Employees | Revenue per employee
5-year trend: rising, flat, declining?
Compare with 2–3 competitors

C. Operating Expense Behavior
Table: Year | Revenue | SG&A | SG&A %
Is SG&A growing slower than revenue (leverage) or faster (anti-leverage)?

 D. Cycle Stress Test (Structural vs Temporary Leverage)
Purpose: test whether leverage holds in bad years.

Identify strong demand years vs weak demand years
Compare across cycles:
Revenue growth
EBIT margin
SG&A behavior
Key question:

Does margin structure hold when demand weakens, or does profitability collapse?
If leverage disappears in downturns → likely temporary, not structural.

IV. INDUSTRY CONTEXT

A. Industry Operating Margin Range
Typical operating margin range for this industry
Where LDLC sits: bottom / middle / top quartile

B. Industry Cost Structure Norms
Is this industry structurally fixed-cost heavy or variable-heavy?
Is LDLC structurally different from peers?

C. Best-in-Class Comparison
Which player shows the strongest operating leverage?
What is structurally different in their model?

FINAL VERDICT
Cost Structure Score (1–10):

How fixed-cost heavy is the model?
Scaling Score (1–10):

How well does revenue scale without proportional cost increases?
Industry Position:

Above average / Average / Below average operating leverage vs peers
Classification:

Structural Leverage
Partial Leverage
Labor-Based Growth
The Core Answer:

"If this business doubles, what percentage of costs stay fixed?"

Give a specific estimate and justify it with evidence.

What the Results Tell You:

The numbers matter less than the reasoning.

Anyone can see margins expanded. The question is: do you understand WHY?

After running Step 1, you should be able to explain:

• What’s fixed in this business?

• What scales with revenue?

• Why this structure is different from competitors?

If you can’t explain it simply, you don’t understand it yet.

Step 2: The Incremental Margin Test

Answer one question:

For every $100 of new revenue, how much becomes profit?

That’s incremental margin.

Higher than current margin → leverage

Same → no leverage

Lower → business degrades as it grows:

• This is the dangerous case.
  Some businesses get worse as they grow.

• Why it happens:

  • New customers cost more to acquire (CAC rises)

  • More discounts needed to sustain growth

  • Complexity and overhead explode

  • Expansion moves into weaker markets

Run This prompt to have you answer ( Always in Deep research mode)

Copy-Paste Prompt:

Read more

Every serious investor knows this.

But first, what is ROIC?

In plain English: how much operating profit a business generates from every dollar invested.

Invest $100.

Get $15 back.

ROIC = 15%. That’s it

The textbook logic:

High ROIC = good business.
Low ROIC = bad business.

Simple..but wrong.

ROIC is never a fixed number.

• A cyclical business can show 25% ROIC at the top of the cycle and 5% at the bottom.

• A growing business can show low ROIC today but high ROIC once investments mature.

The number alone tells you nothing.

What matters is :

1. Do you understand why ROIC is high?

2. Is it sustainable or temporary?

Here’s the 3-step process to find out that with AI :

• Step 1: understand what drivers gross margins level

• Step 2: Understand the capital turns,and what drives asset efficiency?

• Step 3: Stress-test the ROIC, what would make it collapse?

The Two Engines Behind ROIC

ROIC doesn’t appear out of nowhere.

It’s built from 2 forces:

1. Gross Margin : What’s left from each sale after paying to make the product

2. Capital Turns : How efficiently you use your assets to generate sales

The “High Margin” Path

A luxury watchmaker (like Rolex)

They might only sell a few items, but the Gross Margin is massive (e.g., 80%).

The Math: If your margin is 40% and you only turn your capital over 0.5x a year, your ROIC is 20%.

The “High Turnover” Path

A grocery store (like Costco) or a gas station.

They have tiny profit margins (sometimes only 2-3%). However, they sell their entire inventory every few days.

The Math: If your margin is only 4% but you turn your capital over 5x a year, your ROIC is also 20%.

Why this matters:

If you don’t know which engine is driving ROIC, you can’t know what would break it.

Subscribe now

Step 1: The Gross Margin Reality Check

We want to answer this question; :

Why is this company allowed to keep $0.70 of every dollar it makes?
Why don’t competitors simply lower prices and take its customers?

Run this prompt that will analyze 100+ sources to answer these questions:

• Is the margin protected by customer behavior?

• Is it temporary cost advantage? (fragile)

• Is it linked to industry structure?

Run this prompt in Deep Research mode:

Role
You are an industry expert long-term investor.

Your job
Decide whether INSERT COMPANY HERE gross margin is high or not, and whether it is structural or temporary.

Objective
Conduct a “Gross Margin Reality Check” covering at least the last 10 years (if available). You must determine if the Gross Margin is a structural fortress or a temporary anomaly.

Your entire conclusion must focus exclusively on gross margin mechanics and durability. 
10-K, 10-Q (specifically the MD&A, Risk Factors, and Segment Data) and official filings

Earnings call transcripts and Investor Day strategy decks

Industry reports and competitor disclosures (use the latest data to explain recent changes and context)

Integrity: Mark any non-factual analysis as (inferred). Avoid marketing adjectives.

REPORT STRUCTURE & REQUIREMENTS
I. EXECUTIVE SUMMARY

The Current State: A statistical summary of the 10- to 5-year gross margin trend

The “Money Engine”: A one-sentence explanation of the specific “physics” of how the business model turns into gross margin

Industry vs. Individual: Is gross margin high in this industry, and why? Is this margin an industry standard (sector luck) or individual alpha (company genius)?

The Strategic Verdict: Does this margin represent a “Price Boss” (value-based) or an “Efficiency King” (cost-based)?

II. PART 1: THE QUANTITATIVE ANATOMY (THE “HOW”)

COGS Deconstruction: Identify the top 3 direct costs inside Cost of Goods Sold (e.g., raw materials, labor, depreciation)

The Value Driver: Is the margin driven by high ASP (Average Selling Price) or a structural unit cost advantage (patents, process edge, scale)?

Incremental Margin Analysis: Did gross margin % expand or contract as revenue grew over the last 5 years? (Diagnostic of operating leverage)

III. PART 2: THE QUALITATIVE “WAY BEHIND” (THE “WHY”)

The Mission-Critical Test: Is the product a small cost but high risk if it fails? Quantify the customer’s “cost of failure”

The Switching Barrier: Quantify the “pain” of a customer moving to a competitor (data loss, retraining, loss of long-term reliability)

The “Clone” Question: If a billionaire rival spent $1B today, what specific asset (patent, brand, technical secret, distribution, regulatory moat) would stop them from matching this gross margin?

IV. PART 3: THE PRICING POWER STRESS TEST (THE 5-YEAR AUDIT)

Inflation Pass-Through (2021–2024): Did the company raise prices ahead of cost increases? Analyze the speed of margin recovery

Elasticity Proof: When prices rose, did unit volume drop significantly, or was demand relatively inelastic?


FINAL SCORES & VERDICT (GROSS MARGIN ONLY)

Gross Margin Quality Score (1–10):
How structurally strong the gross margin is (brand power, differentiation, structural cost advantage)

Gross Margin Durability Score (1–10):
Likelihood that the current gross margin level persists over the next 5 years under realistic competitive pressure

Final Classification (choose one):

Structural Fortress — Gross margin is structurally protected and defensible

Cyclical / Temporary — Gross margin inflated by short-term factors (pricing cycle, supply chain, fashion trend, FX, etc.)

Structurally Weak — No evidence of durable gross margin advantage

Final Verdict 
Must answer only this question:

Is the gross margin structurally high, or temporarily elevated?
No commentary on valuation, ROIC, management quality, or overall business attractiveness.

I did run an example for Lululemon Athletica Inc.

The result is a detailed 14-page report that you can export into a PDF file to deeply understand their gross margin mechanisms.

Note: I’ve added this prompt to the Prompt Library, so you can copy-paste it anytime and use it.

Step 2: Capital Efficiency Check

Once you understand why high margins exist, the next question is:

How efficiently does this business turn capital into sales?

What you’re trying to understand?

Not the formula.

The drivers :

1)Asset intensity : How Heavy Is the Business?

• Does the company have to reinvest most of its cash just to maintain operations?

• Where is the capital actually tied up?

• Inventory? Stores? Data centers?

2) Working Capital : Who Funds Whom?

• Does the company pay suppliers after collecting from customers?

• Do customers pay before delivery, at delivery, or long after?

• Does inventory move in days or sit for months?

Run this prompt in Deep Research mode to have all answers:

Read more

The other party was tough.

No compromise.

No logic.

Pure emotional leverage.

The counterparty?

My 3-year-old daughter.

The terms?
15 minutes of Legos. 10 minutes of coloring..

Non-negotiable.

Once the terms were fulfilled, I was finally granted 1 uninterrupted hour.

Door closed.

Phone down.

Silence…

I’d been wanting to write this piece for days. Because I knew it deserved real thinking time.

The kind you don’t get between notifications and half-finished thoughts.

So I sat there and asked myself one question:

If I had to start from 0 with AI today what would I actually do?

Here’s the plan:

• Learn the Basics: Understand what AI can do. Accept what it can’t. Write down the constraints you’ll follow every time.

• Map Your Process: Define how you actually invest. Pick one step. Automate that one first.

• Test it :Run it on a stock you already know to understand how to use both Deep Research (external-source research) and NotebookLM (document-grounded analysis).

• Scale It: Save the prompts that work. Use them on every stock. Build your library.

Step 1 Learn the basics (fast, but correctly)

When you want to learn a new skill, you always face the same choice:

Study the theory first? Or dive in and learn by doing?

The art is finding the right balance.

For AI, If you just dive in and “test and learn,” you’ll make basic mistakes.

But the opposite is also a trap.

You don’t need to become an AI expert.

The right approach is to Learn the minimum necessary to use AI without making basic mistakes.

Here’s what that minimum looks like:

What is an LLM?

A model trained on a very large mix of text:
books, websites, code, articles, conversations, etc.

When you ask an investing question, it does not switch to a “finance brain.”

It uses the same general language model, but steered by:

• your question

• the finance context

• patterns it learned about how people talk about investing

That’s why it can sound like an analyst
without being one.

Examples: ChatGPT, Claude, Gemini, Perplexity ( not an LLM technically but it’s ok)

Waht’s a Prompt:

A prompt is an instruction you give the AI.

It’s not the analysis, it’s the trigger that decides what kind of analysis you get..

Think: briefing a junior analyst. Clearer instructions = better output.

Bad: “Tell me about Apple.”

Good: “Analyze Apple’s Q3 2024 10-K. Focus on services revenue growth. Compare margins to Q3 2023...”

Same LLM + lazy prompt = garbage.

Same LLM + precise prompt = excellent analysis.

The prompt is the difference between AI guessing and AI delivering.

Hallucination:

When AI invents something that sounds true… but isn’t.

Confident nonsense.

Surprisingly common.

Good news:

There are simple ways to avoid hallucinations by choosing the right mode and the right model.

If you want to reduce AI hallucinations, this article shows you the exact method

What are reasoning models?

Regular models start typing immediately.

Reasoning models pause to think first.

Examples: OpenAI “Thikning”, Claude Opus 4.5, Gemini 3 Pro,

For investing, they’re better at:

• Multi-step analysis

• Complex valuations

• Systematic comparisons

• Catching math errors

Trade-off: slower but more accurate.

Use them for complex work.

Use regular chat for quick stuff.

What is Deep Research mode?

You know I run all my prompts in this mode
It’s like hiring a research analyst for a few minutes.

What it does:

• Builds a research plan

• Searches the web 50+ times

• Reads full articles and reports

• Cross-checks multiple sources

• Synthesizes into a structured report

Use it for:

• Due diligence on unfamiliar companies

• Industry deep dives

• Competitive analysis

Don’t use it for:

• Quick answers

• Simple lookups

Chat mode = shallow. Deep Research = useful.

The bottom line :
The goal is to deeply understand 3 things:

• Why AI is useful for investing

• Which tools to use and when

• Why AI can be wrong, and how to avoid that

You should use Deep Research with a reasoning model in your LLM of choice.
Right now, I recommend Gemini.

I keep testing other models, and when one performs better, I’ll share it with you.

Pro tip: Run a Deep Research and watch how Gemini works.
It helps you visualize what the tool is doing and understand it faster.

To Go Deeper:

I already gave short definitions of the most important concepts.

If you want to go deeper, here’s how:
Pick 3-5 key concepts to explore (LLM, Prompt, Hallucination ,token..)

1. Open your favorite LLM in “Thinking” mode

2. Run this prompt for each concept:

Copy-Paste Prompt:

You are an expert AI educator for long-term investors.

Help me understand the concept of {CONCEPT NAME}.

Explain the concept so a smart non-technical investor can clearly understand it.

Guidelines

Keep the real meaning. You may simplify and use analogies, but don’t distort the idea.

Stay current. If something is outdated, explain what changed and give the modern view.

Use real-world investing examples (10-K, annual reports, earnings calls, KPIs, portfolio research).

Avoid jargon. If you use a technical word, explain it simply.

Be honest about uncertainty when it depends on tools, models, or context.

Output format

One-sentence definition (plain English)

Mental model / analogy

How it works (3–5 bullets)

Why it matters for investors

Concrete investing example

Common misconceptions

Quick checklist (how to use it safely)

Key terms glossary (max 8)

What’s changed recently (if relevant)

If I want to go deeper: 3 follow-up questions

Optional (only if useful): Suggest up to 3 related concepts or keywords that would meaningfully help me. Skip this if it adds little value.

Subscribe now

3. Follow the rabbit holes:

The AI will suggest related concepts at the end.
When a concept becomes crystal clear, move to the next.
Then run the same prompt again.

This step should take less than 2 houres of deep work.

Step 2: Map Your Process

You’ve got the basics.

Now what?

Before you touch AI, you need to know what you actually do when you research a stock.

Don’t jump straight into prompts and tools and wonder why nothing sticks.

Here’s the truth:

AI can’t fix a process you haven’t defined.

If you don’t know your own workflow, AI won’t either.

The Process Mapping Exercise

This takes 45 minutes.

1. Open a document.

Or take a pen and paper (I still think better when I write by hand .. maybe I’m old school).

2. Thik about your actual process.

Ask yourself:

“Last time I analyzed a stock, what were my exact steps from start to finish?”

Don’t write what you should do.

Write what you actually do.

3. Identify categories.

Group your steps into buckets. Here are common ones:

• Where ideas come from

• How you understand a new business model

• What you read first (earnings, annual report, investor deck, notes)

• How you summarize and cross-check information

• How you understand an industry

• How you research management

• How you build conviction

• What must be true for you to buy

• What makes you say “no”

• What you track after you buy

4. Reflect on past analyses.

Think about the last 2–3 stocks you studied.

Where did you spend the most time?

5. List every task

be specific.

Not “research the company.”

Break it down.

For example, if one of your steps is “research management”, list exactly what that means for you:

• Check if they delivered on past promises

• Look at compensation structure and incentive alignment

• Review capital allocation patterns (buybacks, dividends, M&A, reinvestment)

• Assess communication style : clear and honest, or vague and promotional?

• Check insider ownership

• Review tenure and any leadership turnover

That level of detail is what makes AI useful later.

6. Select ONE task for AI.

What’s next ?

Use AI for all the steps you listed?

Of course not...

The biggest mistake is trying to use AI for everything at once.

Select a single task and start using AI for it.

Pick one step that involves:

• A lot of reading

• Summarizing

• Qualitative judgment

For example, you might want to outsource to AI the work of evaluating management:

How to judge management quality, incentives, capital allocation, and track record.

7. Implement AI on that single task.

Fix the objective:

learn how to use AI for evaluating management quality.

Step 3: Test the system

You’ve mapped your process. You’ve picked one task to automate.

Now it’s time to test.

But here’s where most people mess up:

They test on a stock they don’t know.

Bad setup.

If you don’t know the company, you can’t tell if AI is wrong. You’ll read the output, think “sounds smart,” and move on.

That’s how bad habits form.

The Right Way to Test

Start with ONE stock you know extremely well.

A company you’ve followed for years.

One where you know the business, the management, the history, the risks.

Then run AI on it.

Compare the output with what you already know.

A familiar stock gives you calibration.

You’re not just testing the output. You’re learning:

• How to ask better questions

• How to constrain the model

• How to detect failure modes fast

That’s the real skill.

What We’ll Test: Management Evaluation

In Step 2, we used “research management” as an example task.

Let’s stick with it.

We’ll test two things:

1. How to evaluate management using Deep Research in Gemini or other LLM to surface their track record, incentives, and history

2. How to track management over time using NotebookLM to follow what they say and compare it to what they do

Evaluate Management with Deep Research

This is a one-time deep dive.

You want to understand:

• Who runs this company?

• What’s their track record?

• Are incentives aligned with shareholders?

• How do they allocate capital?

• Do they say what they mean and do what they say?

How to do it

1. Open Gemini (or your preferred LLM with Deep Research mode)

2. Select Deep Research

3. Open the Prompt Library (you have access as a paid subscriber) and choose the Management Evaluation prompts

   

4. Copy–paste the prompt into Deep Research

Prompt: CEO Track Record (Copy–Paste)

ROLE
Act as an analyst preparing a CEO due-diligence dossier for institutional investors.
Write in a consulting-report style: executive summary first, then structured sections with tables and bullet points.
Keep it factual and evidence-based. Do not speculate about personality or charisma.

INPUT
CEO = {{CEO Name}}
Company = {{Company Name}}

SOURCES (guidance, not strict order)

Official filings (annual reports, shareholder letters, SEC/AMF filings, proxy statements, insider filings).

Investor communications (earnings call transcripts, capital markets day presentations).

Reputable business media (FT, WSJ, Bloomberg, Reuters, Forbes, Fortune).

Professional bios (LinkedIn, company bio, official websites).

Watchdog/regulatory sources if relevant (NGOs, governance trackers, sanctions).

Social media posts only if directly tied to business reputation.

WHAT TO DELIVER (fixed structure)

Executive Summary

CEO Profile at a Glance (career highlights, ownership, expertise).

Synthesized Assessment: 2–3 key strengths, 2–3 material risks.

One-line Verdict: “Long-term steward” OR “Potential risk.”

Career Map & Expertise

Timeline of roles: Years | Company | Role | Sector | Notable contribution.

Classification: entrepreneur, operator, allocator (mark the dominant type).

Core expertise & skills, with real examples of impact.

Track Record of Decisions

Table: Date | Role | Major Decision (M&A, buyback, restructuring, expansion) | Outcome (value created/destroyed).

Identify visible patterns (disciplined allocator, empire-builder, turnaround operator).

Ownership & Alignment

Stock ownership (outright shares vs options/awards).

Insider buying/selling history.

Dilution/compensation impact.

Alignment rating: High / Medium / Low.

Governance, Candor & Risks

Evidence of candor vs spin in communications.

Media reputation (serious press).

Controversies, lawsuits, regulatory actions.

Short evidence table: Date | Source | Quote/Fact | Risk Level.

Final Assessment

3–4 bullets summarizing strengths and weaknesses.

Confidence rating: High / Medium / Low.

Clear conclusion: “This CEO shows evidence of disciplined stewardship” OR “This CEO poses material risks to long-term shareholders.”

RULES

Cover the entire CEO tenure in role.

Where possible, bring in long-term evidence (multi-year patterns, not just recent results).

Every claim must be tied to a specific event, number, or quote.

When possible, include the date + source (filing, article, transcript).

If no reliable source exists, explicitly write “Not disclosed” — never fabricate.

Keep writing concise, consulting-grade, and structured.

Since you know this stock you’ll immediately spot what AI gets right and what it misses.

That’s the point.

You’re not just checking the output. You’re calibrating your judgment of the tool.

When AI nails something you already know, you build confidence.

When it misses something obvious, you learn its blind spots.

Both are essential before you trust it on stocks you don’t know.

Track Management Over Time with NotebookLM

Deep Research gives you a snapshot.

But management quality isn’t static.

You need to track what they say over time and compare it to what actually happens.

That’s where NotebookLM shines.

Why NotebookLM?

• It only answers from documents you upload (no hallucinations)

• It remembers everything across multiple files

• It can compare statements across years

How to set it up:

1. Create a new notebook name it after your stock (e.g., “Microsoft Management Tracker”)

2. Upload the last 3–5 years of:

   • Earnings call transcripts

   • Annual shareholder letters

   • Investor day presentations

3. Once uploaded, ask questions like:

What did management say about margins in 2022 vs 2024? Did they deliver?

How has the CEO's tone about competition changed over the last 3 years?

What capital allocation priorities did management outline in 2021? What actually happened?

List every major promise or target management made in the last 5 years. Which ones were met?

Has management's explanation for growth changed over time? Show me the evolution.

Since you now know the stock, this step will make it easier for you to use the tool and understand its capabilities.

Why This Step Is Non-Negotiable

Testing on a stock you know does two things:

1. You build confidence.

You see AI work correctly and you trust the process.

2. You learn two ways to use AI.

• Deep Research: An open model that searches the web, cross-references sources, and synthesizes analysis from scratch.

• NotebookLM: A closed-loop system that only answers from documents you upload no hallucinations, no outside noise.

Both are essential.

Master both before you scale.

Step 4: Scale It

Your first prompt works.

You’ve tested it on a stock you know.

You trust the output.

Now what?

Scale it.

How to Scale

1. Run it on 2 other stocks.

Pick companies you’re less familiar with.

Master the 2 tools

2. Integrate it into your workflow like a normal habit.

Every time you research a new stock, run the management prompt.

Every quarter, upload the new transcript to NotebookLM.

Make it automatic.

Then Expand

Once the first task is solid:

• Pick one more step from your process map

• use one more prompt

• Create one more repeatable output

One layer at a time.

Over time, you’ll have something most investors don’t:

A research system that gets stronger every time you use it.

Recap: The 4-Step System

If you had to start from zero with AI today, here’s the path:

Step 1 : Learn the basics
Understand what AI can do (and can’t).
Learn key concepts: LLM, prompt, hallucination, Deep Research.
Takes 2 hours, not 2 weeks.

Step 2 : Map your process
Write how you research a stock.
Break it into tasks.
Automate one task first.

Step 3 : Test it
Run it on a stock you already know.
Compare results to reality.
Use Deep Research + NotebookLM.

Step 4 : Scale it
Reuse on other stocks.
Refine prompts.
Add tasks gradually.
Build a system.

PS: If you get stuck on any step or want feedback on your workflow, just reply to this email. I read every message