Maison Cler · Claude Code Bookclub · Self-Study

FOUNDATIONS

From idea to built.

A beginner's guide to Claude Code — and why it changes everything for founders, operators, and builders.

SHIFT

Something changed

Not long ago,
if you had an idea —

you had two options.

Learn to code. Months. Maybe years.
Or hire someone to build it. Expensive. Slow. Dependent.

Most ideas died between the idea and the execution.

That gap just closed.

The new reality

It's already
happening.

Founders are building their own Shopify apps.
Operators are automating entire sales flows.
People with zero coding background are shipping real software — not prototypes, not demos — real, working products.

This isn't hype.

The question isn't whether to learn this.

The question is: how long can you afford not to?

In this module — you'll learn

Six things.

Why agentic AI is a different category — not just a better chatbot.
The difference between Claude.ai, Cowork, and Claude Code.
How to think of Claude Code — as a contractor, not a search engine.
How to brief it so it builds the thing you want.
How to make it remember you across every session (CLAUDE.md).
How Skills turn great one-off results into a system that runs.

By the end, you'll be able to explain Claude Code to anyone — and start building your first real thing.

Let's
begin.

Part A · A new category

A

What is
Agentic AI?

Two kinds of AI help

You've used AI before.

You type something. It replies. You ask for an email. It writes one. You paste a problem. It explains.

That's conversational AI — useful, but reactive. It waits for you. It answers one question at a time.

Agentic AI is different.

What "agentic" actually means

Imagine you ask an assistant:

"Plan my customer onboarding flow, build the emails, connect it to my CRM, test it, and tell me when it's done."

Conversational

Helps you think through it. Step by step. With you guiding every move.

VS

Agentic

Just goes and does it. Reads files. Makes decisions. Takes actions. Comes back only when there's a real choice.

Agentic AI acts toward a goal — not just toward the next sentence.

Why this matters for you

This is what
agentic AI unlocks.

If you run an ecom brand: brief a tool to build your upsell flow, test three versions, and push it live — while you focus on your next product launch.
If you're building an app idea: describe it in plain language and watch a working version take shape in hours, not months.
If you run sales: automate the sequences, follow-ups, and data pulls — built exactly the way you'd describe it to a team member.

It's the difference between having a conversation and having a contractor.

Where you are now

You now know:

Conversational AI = reactive; it answers one question at a time
Agentic AI = acts toward a goal; it just goes and does
Claude Code is agentic — it's not a chatbot, it's a contractor

↓

Next: Anthropic makes several tools. Which one is which — and which one builds real software?

Part B · The Claude family

B

What's what?

You already know this one

Claude.ai

The chat interface in your browser. Conversational. Great for thinking, writing, problem-solving.

But Claude.ai is still a chat window. You type. It replies. Nothing gets built. Nothing gets saved to a project. Nothing gets deployed.

It's the equivalent of texting a really smart friend.

Then there's this one

Cowork

Claude's desktop tool — designed for people who want to automate tasks on their computer without writing code.

Think: file management, document workflows, repetitive tasks across your apps.

More powerful than Claude.ai — but still at the level of your desktop.

And then there's this one

Claude Code

This is where things get serious.

Not a chat window. A full AI agent that lives in your terminal — and can read your project files, write code across multiple files, run tests, fix errors, commit changes, and ship working software.

You describe what you want to build. It builds.

It understands your entire codebase — not just the message you just typed.

The simple map

All three, side by side:

	Claude.ai	Cowork	Claude Code
Where it lives	Browser	Desktop app	Terminal / Browser
What it does	Thinks & writes with you	Automates desktop tasks	Builds actual software
Best for	Thinking, drafting, researching	File & task automation	Shipping real products

Claude Code is the one that builds.

The rest of this module is about Claude Code.

Where you are now

You now know:

Three Claude products. Different tools for different jobs.
Claude Code is the one that builds real software.

45 min

Pull from Stripe. Pull from Shopify. Paste into a sheet. Format. Format again. Realize you missed Friday's data.

→

With Claude Code

3 min

"Pull this week's numbers and format my report." It does. Same template, every time.

SHIFT

The pattern

If you do it
by hand today,
you can automate
it tomorrow.

That's the whole shift. Not "code instead of work." More like:
brief, review, ship — five times faster, every time.

Part D · Meet your contractor

D

How it
actually works

Meet your contractor

Imagine you're
building a house.

You don't need to become a builder. You don't need to learn how to lay bricks or wire electricity.

What you need is a great contractor — someone you hand a brief to, who knows their craft, and who comes back with results.

Claude Code is that contractor.

You describe the house. It builds.

But it lives in a terminal — what's that?

Don't worry about
the word "terminal."

The terminal is just a text-based way to talk to your computer — like talking to your file system directly. Claude Code lives there because it needs access to your actual project files. It reads them. Edits them. Creates new ones.

Good news: Claude Code also runs in your browser at code.claude.com.

So you don't have to be technical to start. Just a browser, a folder, and a brief.

The loop · how a session works

Six steps.
Same every time.

You open a project.
Claude Code reads your files and understands the context.
You describe what you want — in plain language.
It plans. It codes. It runs. It fixes errors. It iterates.
It asks your permission before anything risky.
You end up with working, committed code.

You describe the outcome. It figures out the path.

Kitchen analogy

A skilled chef
in your kitchen.

You don't tell them how to cook. You say:

"Dinner party for 8. Italian. Nothing too heavy."

They handle the shopping list, the prep, the timing, the plating. You stay focused on the experience you want to create.

You are the chef-owner. Claude Code is the head chef in your kitchen.

What it looks like on screen

Three windows.
One view.

your-project — VS Code

Files

▸ company-context/

CLAUDE.md

index.html

styles.css

README.md

Editor
1# My Coffee Brand
2
3## Brand voice
4Warm, editorial,
5understated.
6Never exclamation.

Claude Code
› build me a landing page
  for our coffee brand.
claude is thinking…
✓ created index.html
✓ created styles.css

Files on the left. Code in the middle. Claude Code on the right. Everything in one place.

Where you are now

You now know:

Claude Code is a contractor. You describe; it builds.
It lives in a terminal — or in your browser at code.claude.com.
Every session is a loop: read, describe, plan, build, ask, ship.

↓

The rest of this module is about how to talk to your contractor well — so what comes out actually matches what you wanted.

Section 1

01

The Mental Model

In this section — you'll learn

Two things:

How to think about Claude Code so you stop expecting it to read your mind
Why your brief — not your typing speed — decides what you get back

First — what it isn't

Claude Code is
not Google.

With Google, you ask a question and get an answer.
With a chatbot, you have a conversation.
With Claude Code, you hand over a job and get back a deliverable.

BRIEF

The mental model

Claude Code is
a contractor.

You brief it. It builds.

The catch

Contractors don't
read minds.

They complete what you ask for.

The quality of what you get back is directly proportional to the quality of your brief.

The freelancer test

Imagine you hire a designer.
Two ways to brief them:

Bad brief

"Make me a website."

What you get: garbage. Stock photos. Lorem Ipsum. Generic template.

VS

Good brief

"One-page site. Hero, 3 features, pricing table with two tiers, CTA at the bottom. Cream background, dark text, red accents. No animations. No login. No blog."

What you get: something usable. Same designer, different brief.

Kitchen parallel

Same thing
at a restaurant.

Bad order

"Bring me
something good."

The chef has no idea what you want. Sweet? Savory? Vegetarian? You'll get whatever's coming up.

Clear order

"Spaghetti
carbonara."

"4 eggs, guanciale, pecorino, black pepper. No cream. Twenty minutes." The chef does it right — and fast.

Where you are now

You now know:

Claude Code is a contractor — not a search engine, not a chatbot
The quality of your brief decides the quality of what you get back

↓

But: even with the right mindset, most builds still go wrong. Here's the #1 way it happens.

Section 2

02

The Boat With Wings Problem

In this section — you'll learn

Two things:

The #1 way builds with Claude go wrong (it happens to everyone)
One simple rule that prevents 90% of the damage

How it plays out

You tell Claude to
build you a car.

A simple car. Four wheels, engine, drives forward.

Ten minutes in, you get an idea:

"What if it could also fly?"

Claude adds wings. Now you're excited.

"What if it could also float?"

Claude adds a hull.

The result

A boat with wings
and wheels —

that doesn't drive,
doesn't fly,
and sinks.

Meanwhile, Claude spent the first 20 minutes building a login system and a user database — before it built the car.

Why builds fail

Two traps killed
this build.

Trap 1 — Scope creep

You kept adding features before the first one worked. Every new idea felt small ("just add wings!") but each one made the whole thing more fragile.

Trap 2 — Infrastructure first

Left to itself, Claude builds plumbing first — auth, databases, error handling. The thing your user actually touches comes last, if at all.

The rule

Build the thing
your user touches
first.

Everything else comes later.

What the rule looks like in practice

Email generator?

First thing that should work: type a topic, get an email.

Product page?

First thing: the page renders with real content.

Not a CMS. Not an admin panel. Not a login.

And when a new idea hits mid-build?

Say this:

"Good idea. Let's add that after this works."

Write it down. Come back later. Nine times out of ten, the idea is still good — the timing was wrong.

Kitchen parallel

Same problem
at dinner.

Imagine you order pasta. Five minutes in, you change your mind — "actually, make it pizza." Halfway through, you say — "you know what, sushi instead."

You'd get a plate of raw, half-cooked everything.

Same dynamic with Claude. Pick the dish. Let it finish.

Where you are now

You now know:

Quality in, quality out — the brief is the work
Build what your user touches first — no infrastructure before workflow
New ideas mid-build go on a list — not into the build

↓

Next: Now we get into how to actually write the brief. There's a name for it. It's called a spec.

Section 3

03

The Spec —
How to Brief Claude

In this section — you'll learn

Three things:

What a "spec" actually is (it's not as scary as it sounds)
Why constraints matter more than instructions
How your existing taste in apps is already enough to write one

A spec is just three things

Don't be intimidated.

01

What it does

The job it performs.

02

What it looks like

The format, the style, the layout.

03

What it must NOT do

The boundaries. The walls.

That third one is the secret weapon.

The secret weapon

Constraints are
more powerful
than instructions.

Kitchen first — bad order vs good order

Tonight's dinner — two ways to ask:

Bad order

"Make me dinner."

You'll get something. Maybe a stew. Maybe a salad. You won't know until it lands.

VS

Good order

"Spaghetti carbonara. 4 eggs, 200g guanciale, pecorino, black pepper. No cream. 15 minutes."

The cook knows exactly what to make — and what to skip.

Now the same idea, in code

The build:

A tool that generates product descriptions for your ecommerce store.

Same idea. Two briefs. Two very different results.

Bad spec vs. good spec

Bad spec

"Build me a product description generator."

What Claude builds: a full Next.js app with a database, user accounts, history, settings page, and a paid API integration — all before you've seen a single description.

VS

Good spec

"Single HTML page. One input for product name, one textarea for bullets, one Generate button. Calls Claude API, shows result below. No login. No database. No history."

What Claude builds: exactly that. A working tool in under 5 minutes.

Same idea

Wildly different
output.

The only difference was the brief.

Your product intuition is the technical skill

You already have
the most important skill.

You've used thousands of apps. You know what good feels like. You know when a checkout flow is annoying. You know when a dashboard is confusing.

You know what "clean" means, what "fast" means, what "too much" means.

That intuition IS the spec.

Use what you already know

Reference the apps
you already love:

"Add a like button like Instagram — heart animation, count, the whole thing"
"Make the sidebar collapsible like Notion"
"Add a search bar like Shopify's admin"
"Make the cards draggable like Trello"
"Style it like Apple's product pages — whitespace, big images, minimal text"

Each one is a complete spec. Claude has seen all those apps. Your reference library is your technical vocabulary.

Where you are now

You now know:

A spec is three things: what it does, looks like, must NOT do
Constraints beat instructions — that "must NOT" line is the secret weapon
Your taste in apps is already enough — reference what you love

↓

But here's the catch: every new session, Claude forgets all of this. Even your perfect spec is back to zero. How do we fix that?

Section 4

04

CLAUDE.md —
Giving Claude a Brain

In this section — you'll learn

Three things:

The problem: Claude has no memory between sessions
The fix: a single file called CLAUDE.md
How to make Claude feel like it knows you — your brand, your stack, your "never do this" list

The goldfish problem

Every session
is day one.

Here's something that trips up almost everyone the first time they use Claude Code:

Claude forgets everything between sessions.

It doesn't remember your project setup. It doesn't know your tech stack preferences. It doesn't know that your checkout page has a quirky architecture, or that you always want TypeScript.

So it has to rediscover all of this every single time. That's slow. And it leads to inconsistent results.

What that's like

Like hiring a new
contractor every
single day.

And re-explaining everything from scratch.

Brand voice. Tech stack. The rule about never deploying without permission. Every. Single. Time.

The fix

CLAUDE.md

A simple text file in your project folder. Claude Code reads it automatically at the start of every session.

Think of it as the onboarding document you'd give a new contractor before they touch your project.

Still abstract? Let's try a few angles.

Angle 1 · Cooking

You and your friend
both make pasta tonight.

Same dish. Spaghetti carbonara. Same kitchen. Same ingredients on the counter.

Two very different dinners come out of it. Why?

Same goal · different prep

You wing it.
Your friend doesn't.

You

No recipe.
Pure vibes.

You've made pasta before. How hard can it be? You guess at quantities. Forget the pepper. The eggs scramble.

Edible. Not memorable.

Carbonara

By Gordon Ramsay · pinned to wall

200g guanciale, diced

4 large egg yolks + 1 whole egg

80g pecorino romano, grated

Cracked black pepper · lots

400g spaghetti

Never · cream. Never · garlic.

Same kitchen. Wildly different dinner. The recipe was the only difference.

Still not landing?

Try another
angle.

Angle 2 · Buying a house

Imagine you're
house-hunting.

You walk into an agent's office. They ask: "What are you looking for?"

If you say "a place to live" — they'll show you 50 random listings. Most won't fit. You'll waste your weekend.

If you hand them this:

"3 bedrooms · garden · under €600k · within 10 min of a metro · NEVER ground floor, NEVER on a main road, NEVER fixer-upper."

Now they show you 5 perfect matches.

CLAUDE.md is your house brief — but for your project.

Halfway there

You with me?
One more,
just for you.

Angle 3 · The contractor walkthrough

Before the hammer
starts swinging.

A good contractor doesn't just show up and start working. Before they start, you walk them through the house:

"We're keeping the original floorplan. The electrics in the east wing are old — don't touch those without asking. We always use this supplier for materials. The client hates open shelving."

That walkthrough is your CLAUDE.md.

Not magic. Just a really good brief — written once, used every time.

Three angles · one idea

Same goal,
same people,
different result.

The brief is what changes everything.

What goes in a CLAUDE.md?

The more specific,
the better the output.

Your project's tech stack and architecture overview
Coding conventions and standards you follow
Libraries and frameworks you use (and avoid)
The context Claude needs to not ask the same questions twice
Anything a smart contractor would need to know before day one

Plus: your brand voice, your "never do this" rules, and what Claude got wrong before.

Make the recipe even better

Split it into
context files.

Instead of cramming everything into one CLAUDE.md, split your business context into a few files. Like a chef who keeps separate notes for ingredients, dietary rules, and supplier lists.

your-project/ ├── CLAUDE.md # Rules and preferences ├── company-context/ │ ├── COMPANY.md # Who you are, your numbers, your team │ ├── BRAND-VOICE.md # How you sound, good and bad examples │ ├── PERSONAS.md # Your 3 main customer types │ └── COMPETITIVE.md # Who you compete with

Claude reads all of these at the start of every session. Now when you say "write a product description," it already knows your brand voice, customer, and competitive angle — without you explaining.

The progressive disclosure trick

Don't write the perfect
CLAUDE.md on day one.

Start with 10 lines. The basics.

Then, every time Claude does something you don't like — builds a login when you didn't ask, uses the wrong colors, deploys without permission — add a rule.

After two weeks, your CLAUDE.md will be perfectly calibrated. It will have caught every bad habit Claude has and corrected it.

This is the file that makes Claude feel like it knows you.

Where you are now

You now know:

Every session is day one — Claude has the goldfish problem
CLAUDE.md is the fix — a file pinned to your project
Start small. Add a rule whenever Claude does something you don't like

↓

But: even with CLAUDE.md, sessions get long and messy. Claude starts making weird choices halfway through. Why?

Section 5

05

Context Management —
The Invisible Skill

In this section — you'll learn

Three things:

Why long sessions go bad — even when your brief was perfect
The single highest-ROI habit you can build (it's one command)
When to reset, when to compress, and when to push through

The whiteboard mental model

Imagine Claude's brain
is a whiteboard.

Every message you send, every response Claude gives, every file it reads, every error it encounters — all of it gets written on the whiteboard.

Watch the whiteboard fill up

Over the course of one session:

0:00 Whiteboard is clean. Claude is sharp, focused, precise.

0:20 First request, third revision, an error from earlier — all logged.

0:50 Claude is juggling files, code, contradictions, and an instruction that conflicts with one from 30 min ago.

1:30 The whiteboard is full. Claude forgets constraints. Contradicts itself. Builds things you already told it not to.

Same as a chef during dinner rush. Order #1, #2, #3, "table 6 is allergic to nuts," send back the steak… eventually they forget which one was medium-rare.

The #1 reason sessions go bad

Not because
Claude got dumber.

Because the whiteboard is full.

The fix

/clear

Erases the whiteboard. Completely. Fresh start.

This is the single highest-ROI habit you can build. When a session is degrading, type /clear and start over.

Your CLAUDE.md gets reloaded automatically. Claude still knows your project, your rules. It just loses the clutter.

Two rules of thumb

Rule 1 — One task per conversation

Building a product page? That's one conversation.

Need to fix a bug? /clear, new conversation.

Want to add a feature? /clear, new conversation.

"Continue where we left off" is a trap

It feels efficient. It's not.

Claude has to reconstruct context from a messy whiteboard. A fresh start with a good CLAUDE.md is almost always faster.

The middle ground

/compact

Sometimes you can't afford a full reset. /compact squeezes the whiteboard — keeps the important context, removes the noise.

Use /compact when

Deep in a build and switching tasks would lose momentum. Long but still productive. Want Claude to refocus.

Use /clear when

Starting a new task. Claude is making mistakes. You're frustrated. (Seriously — /clear and a calm re-prompt fixes most problems.)

Where you are now

You now know:

Sessions go bad because the whiteboard fills up
/clear is the highest-ROI habit — fresh start, CLAUDE.md reloads
One task per conversation. /compact when you can't reset; /clear when you can

↓

Next: The handful of commands you'll use every day.

Section 6

06

Essential Commands

Three commands · the ones that matter

/clear

Fresh start.

The most important one. Use between tasks. Use when things feel off. Use liberally.

/compact

Compress, don't reset.

For long sessions still going well — but getting heavy.

/init

Generate a starter CLAUDE.md.

Run in any project folder. Claude scaffolds one based on what it finds. Edit from there.

Giving Claude a screenshot

This is 80%
of debugging.

When something looks wrong, don't try to describe it. Show it.

Drag and drop an image into the Claude Code terminal
Paste a screenshot directly (Cmd+V on Mac)

Then say:

"This is what it looks like. The button should be on the right, not the left. Fix it."

Like showing the chef the dish that came back to the pass — they can taste the problem instantly.

Making Claude plan first

Don't let Claude
just dive in.

When you're starting something complex, tell it to think first:

"Plan how you would build this. Don't write any code yet."

Claude lays out the approach — what files, what structure, what order. You review, adjust, then say "go."

Why this works: you catch bad decisions before any code is written. This prevents the boat-with-wings problem.

Where you are now

You now know:

The three commands: /clear, /compact, /init
Drag-and-drop screenshots — 80% of debugging solved
"Plan first" prevents most disasters

↓

Next: Even with a great CLAUDE.md, you'll do the same kinds of tasks over and over. There's a name for that — and a fix.

Section 7

07

Skills —
Repeatable Playbooks

The repetition problem

Even with a great
CLAUDE.md —

you'll find yourself doing the same kinds of tasks over and over.

"Review this for SEO."

"Build a new product page using our standard layout."

"Generate a customer email in our brand voice."

You can describe these from scratch every time. But there's a better way.

What are Skills?

Saved, reusable
instructions.

Pre-packaged workflows you can trigger on demand. Think of them like standard operating procedures — but for your AI contractor.

Instead of re-explaining your SEO process every time, you create a skill called /seo-review. Claude Code knows exactly what to do. Every time. Consistently.

The restaurant SOP analogy

A great restaurant
has a recipe book.

It doesn't rely on the chef inventing the bolognese fresh every Tuesday.

They have a recipe. A process. A standard. New staff learn it. Experienced staff follow it. The result is consistent — because the knowledge is baked into the operation, not just kept in one person's head.

Skills are your restaurant's recipe book.

They turn great one-off results into reliable, repeatable output.

What Skills enable

Four big things:

Your team can use Claude Code without re-learning your processes
Results are consistent across sessions and team members
You build a library of capabilities specific to your business
Complex, multi-step workflows become a single command

Skills are how Claude Code stops being a tool you use — and starts being a system that runs.

Section 8

08

The Stack Decision

What tool for what job

You don't need to
answer this in detail.

But a rough sense helps:

What you need	What to use	Example
Show an idea, mockup, prototype	HTML file	Character card, landing page draft
Build a tool people use in a browser	Next.js	Calculator, dashboard, generator
Save data between sessions	Supabase	User accounts, product lists
Run without you pressing a button	Trigger.dev	Daily reports, automated emails
Make something smart	Claude API	Generation, analysis, scoring

You don't memorize this. When you start a build, Claude picks the right tool — if your CLAUDE.md mentions the stack.

The only rule to remember

Start with
the simplest
option.

If an HTML file can do the job, don't build a Next.js app.
If a Google Sheet can track your leads, don't build a CRM.

You can always upgrade later. You can't easily downgrade.

Zoom out

What you now have.

An AI agent that can build real software — Claude Code
Permanent project context so it always knows your setup — CLAUDE.md
Repeatable playbooks for consistent results — Skills

Together, these three things turn a powerful tool into a reliable business system.

What this unlocks for founders

The barrier
is gone.

Ecom founders build custom flows, automations, and tools — without waiting on developers or paying agency retainers.
Product builders go from idea to working prototype in a day — testing fast, iterating faster.
Sales operators automate sequences, dashboards, and reporting — built exactly to their workflow, not a template.

The thing that used to be called "you need to hire a developer" — is gone.

NOW

Why now?

The window
to move first
is open.

This technology is six months old for most people. The learning curve is real but short.

The people who start now will have six months of practice by the time the rest of the market catches up. The contractors, developers, and agencies who rely on the old model are already feeling the shift.

The question was never "is this the future?"

The question is: are you early, or are you late?

Your first step

You don't need to
understand every
part of this today.

You need to open Claude Code, describe one thing you want to build, and let it run.

Everything else — the CLAUDE.md, the Skills, the workflows — comes from using it. From doing the thing.

Start small. Start today. The system builds itself as you learn it.

Before Session 2 · homework

Pick one thing
to build for real.

Something for your business. Not a toy. Something you'd actually use.

Then write the spec — the three sections you now know:

What does it do? — 5–10 specific bullets
What does "done" look like? — the moment you'd say "this works"
What must it NOT do? — the walls, the guardrails

Bonus: fill in the company-context/ templates so Session 2 already knows your brand.

Bring it to Session 2. We'll build it live.

What we covered today

Six things you now know.

Why agentic AI is a different category — not just a better chatbot
The difference between Claude.ai, Cowork, and Claude Code
How Claude Code works — as your contractor
What CLAUDE.md is and why it exists (the goldfish problem)
What Skills are and why they matter for consistency
Why the window to move first is open right now

Remember this · the system

Three pieces.
One system.

CLAUDE CODE

The agent that builds.

+

CLAUDE.MD

The brain that remembers you between sessions.

+

SKILLS

The playbooks that repeat consistently.

= A system that runs.

Remember this · four rules

Four rules
to live by.

01

Brief like a contractor

Quality in, quality out. The brief is the work.

02

Build what users touch first

No infrastructure before the workflow works.

03

Constraints beat instructions

The "must NOT" is the secret weapon of every spec.

04

/clear is your best friend

Fresh whiteboard for every new task. CLAUDE.md reloads automatically.

SHIFT

Remember this · the whole shift

Hours
by hand.

Minutes
by brief.

That's the whole shift.

You already have
the ideas.

Now you have
the contractor
to build them.

Let's get to work.

See you in Session 2.

FOUNDATIONS

Not long ago,if you had an idea —

It's alreadyhappening.

Six things.

Let'sbegin.

What isAgentic AI?

You've used AI before.

Imagine you ask an assistant:

Conversational

Agentic

This is whatagentic AI unlocks.

You now know:

What's what?

Claude.ai

Cowork

Claude Code

All three, side by side:

You now know:

What this cando for you.

A landing pagefor your new product.

Researching50 prospects.

Five descriptionsin your brand voice.

Three competitors,fully mapped.

Your Mondayweekly report.

If you do itby hand today,you can automateit tomorrow.

How itactually works

Imagine you'rebuilding a house.

Don't worry aboutthe word "terminal."

Six steps.Same every time.

A skilled chefin your kitchen.

Three windows.One view.

You now know:

The Mental Model

Two things:

Claude Code isnot Google.

Claude Code isa contractor.

Contractors don'tread minds.

Imagine you hire a designer.Two ways to brief them:

Bad brief

Good brief

Same thingat a restaurant.

"Bring mesomething good."

"Spaghetticarbonara."

You now know:

The Boat With Wings Problem

Two things:

You tell Claude tobuild you a car.

A boat with wingsand wheels —

Two traps killedthis build.

Trap 1 — Scope creep

Trap 2 — Infrastructure first

Build the thingyour user touchesfirst.

Same problemat dinner.

You now know:

The Spec —How to Brief Claude

Three things:

Don't be intimidated.

Tonight's dinner — two ways to ask:

Bad order

Good order

The build:

Bad spec

Good spec

Wildly differentoutput.

You already havethe most important skill.

Reference the appsyou already love:

You now know:

CLAUDE.md —Giving Claude a Brain

Three things:

Every sessionis day one.

Like hiring a newcontractor everysingle day.

CLAUDE.md

You and your friendboth make pasta tonight.

You wing it.Your friend doesn't.

No recipe.Pure vibes.

Try anotherangle.

Imagine you'rehouse-hunting.

You with me?One more,just for you.

Before the hammerstarts swinging.

Same goal,same people,different result.

Not long ago,
if you had an idea —

It's already
happening.

Let's
begin.

What is
Agentic AI?

This is what
agentic AI unlocks.

What this can
do for you.

A landing page
for your new product.

Researching
50 prospects.

Five descriptions
in your brand voice.

Three competitors,
fully mapped.

Your Monday
weekly report.

If you do it
by hand today,
you can automate
it tomorrow.

How it
actually works

Imagine you're
building a house.

Don't worry about
the word "terminal."

Six steps.
Same every time.

A skilled chef
in your kitchen.

Three windows.
One view.

Claude Code is
not Google.

Claude Code is
a contractor.

Contractors don't
read minds.

Imagine you hire a designer.
Two ways to brief them:

Same thing
at a restaurant.

"Bring me
something good."

"Spaghetti
carbonara."

You tell Claude to
build you a car.

A boat with wings
and wheels —

Two traps killed
this build.

Build the thing
your user touches
first.

Same problem
at dinner.

The Spec —
How to Brief Claude

Wildly different
output.

You already have
the most important skill.

Reference the apps
you already love:

CLAUDE.md —
Giving Claude a Brain

Every session
is day one.

Like hiring a new
contractor every
single day.

You and your friend
both make pasta tonight.

You wing it.
Your friend doesn't.

No recipe.
Pure vibes.

Try another
angle.

Imagine you're
house-hunting.

You with me?
One more,
just for you.

Before the hammer
starts swinging.

Same goal,
same people,
different result.

The more specific,
the better the output.

Split it into
context files.

Don't write the perfect
CLAUDE.md on day one.

Context Management —
The Invisible Skill

Imagine Claude's brain
is a whiteboard.

Not because
Claude got dumber.

This is 80%
of debugging.

Don't let Claude
just dive in.

Skills —
Repeatable Playbooks

Even with a great
CLAUDE.md —

Saved, reusable
instructions.

A great restaurant
has a recipe book.

You don't need to
answer this in detail.

Start with
the simplest
option.

The barrier
is gone.

The window
to move first
is open.

You don't need to
understand every
part of this today.

Pick one thing
to build for real.

Three pieces.
One system.

Four rules
to live by.

Hours
by hand.

Minutes
by brief.

You already have
the ideas.

Now you have
the contractor
to build them.