Building with AI, for real.

You leave with a setup, not a vibe.

Where AI gets hands.

A chatbot answers you. An AI in your terminal can touch your files, run programs, and ship things while you watch.

What I use

warp.dev

A modern terminal built for working with AI. This is my home screen all day.

What friends here use

superset.sh

Same idea, different flavor. The terminal is a preference, not the magic.

claude.com or warp.dev? Both. Different jobs.

claude.com (chat)

Thinking. Writing, research, planning, strategy, questions. A brilliant conversation partner with no hands. Also on your phone.

Claude Code in Warp

Doing. The same brain, but with hands: it reads your files, runs commands, builds and ships. Warp is just the terminal it lives in.

The rule of thumb

Need an answer or a draft? Chat. Need something to exist afterwards: a file, a page, a system? Terminal. Most people never make this jump. Tonight you do.

claude.com is not the lesser one. It is the thinking room.

Claude, ChatGPT, Gemini: engines, not religions.

Use different engines for different jobs. The workflow is the asset. The model is just the engine you swap.

Claude Code does the doing.

I describe the outcome in plain language. It reads my files, writes the code, runs the commands, and checks its own work.

You, in plain language

→

Claude Code

→

Files

Terminal

APIs

Browser

Piece 3 · Skills Claude Code

Write it once. Run it forever.

A skill is a playbook written as a file: how I write proposals, how I research a market, how I build a dashboard. Nine years of knowing things, condensed into files my AI executes.

• /dashboard turns any data into a live page.
• /brain-dump turns a messy voice note into a build plan.
• /api-docs fetches real documentation before any integration.
• When I find a better way, I edit the file. Every future run improves. That is compounding most people never start.

One folder. Everything in it.

ai-command-center/

→

skills

→

projects

→

handoffs (memory)

→

docs

The folder most people miss: handoffs. Every session ends with a note to the next one. Your AI stops forgetting who you are every morning. You get this exact structure in the starter kit tonight.

I talk to my computer. Seriously.

The biggest bottleneck in using AI is not the AI. It is how fast you can get what is in your head into the machine. Typing makes you abbreviate. Speaking gives ten times the context and nuance.

What I use

superwhisper. One-time purchase, runs locally on your Mac. Press a hotkey, speak, and the text appears wherever your cursor is, including the terminal.

Free alternative

Whisper, the open source model it is built on. More technical to set up, completely free.

Every brain dump, every build request, every skill I run starts with me talking. In my setup this is not optional.

Your voice in. Working systems out.

Voice (superwhisper)

→

Terminal

→

Claude Code + skills + MCPs

→

Data

→

Shipped

→

Always on

→

Your phone

Everything else tonight lives somewhere on this map. When you feel lost later, come back to this slide. It is drawn out on your resources page too.

I was slowing down after quitting everything. I doubled down on AI for one reason: it matched the values my first businesses ran on. Joy. Excitement. Curiosity.

• Scraped repositories, studied them to the core: the architecture, the setup, the why.
• Then executed on hundreds of ideas I had carried for years.
• And condensed everything I knew into skills. The folder I just showed you was born in that room.

The info space was toxic. I spent over $200K just recovering pages that were taken down weekly. So no, messaging strangers on the internet was not casual for me.

• Zoom calls, screens shared, my setup against theirs. I wanted to know how much I really knew. Turns out, a fair bit.
• Private groups with the AI guys you all follow.
• Then one of them, after seeing my builds, asked me to become CTO of his company. I agreed.

The waiter who said yes.

• First month here we ate at the Hilton almost every day. Not for the food. For the staff. That is where I met Hadi.
• One night I told the whole staff: "You could be hired for $1,000 a month by a US company if you learn this. It will not be easy. But you could." I gave everyone my Instagram.
• One person messaged. I sent him three videos, over twelve hours each. The long road, on purpose.
• Then I disappeared: three infections at once. Weeks gone. I honestly wondered if he had quit.

• People wanted to pay me to teach them. I did not feel good enough. My mentor asked two questions: "Do you feel excited? Do you get them results?"
• So I consulted hundreds for free for a year. You win on results. Then you charge.
• At nineteen: a personal brand agency, built with my two best clients. Shoutouts, funnels, ads.

• At seventeen I went deep into Human Design. Years later I built an AI tool for it. My coach, twenty years in, got more results with it in one week.
• In 2025 came an awakening experience I did not plan. It took a week to recover.
• Then I sold everything. Pages, coaching, content, agency. People I loved depended on it. I knew it would flop without me. I left anyway.

Which lands us back in December, where story one started. Now you have the whole map.

There is no AI expert.

Everything you build falls into three pillars.

01 · Problem solving

Every manual, repetitive, or draining task in your life and business: reporting, scheduling, research, follow-up. Each one becomes a skill or an automation.

02 · Capability learning

Discovering what AI can actually do: websites, automations, browser control, voice agents, whole products. Most people know five percent of this.

03 · Opportunity leverage

Systems that learn from the outside world: scrape channels, monitor markets, extract competitor intel. AI finds what you would miss.

Your homework before any code: write down every action in your business that needs manual work or manual thinking. That list is your build queue.

Where almost everyone gets stuck.

01 · Dopamine

Watching AI generate feels like progress. A demo is not a system.

Antidote: one question every Friday. Did it remove work from my week?

02 · Lost in the sauce

New model, new tool, new framework, daily. A quarter can vanish into switching.

Antidote: pick one stack, go deep, review what is new once a month.

03 · Losing focus

Focus is the one thing you cannot outsource to AI.

Antidote: one business, one bottleneck, one build. Finished beats impressive.

Roadmap first. ROI decides.

Playing with AI is addicting. You are not here to play: you are here to leverage your time, skills and abilities. So before you build anything, write a roadmap and rank every idea by one question:

• Proven high-ROI first builds: dashboards on data you have, automations that were painful in make.com, market research sweeps, copywriting engines, funnel duplication.
• If a build scores on none of the three: it goes to the bottom of the list, no matter how cool it looks.

Six words, the real definitions.

Agent

An AI model running in a loop with tools and a goal: it plans, acts on files, browsers and APIs, reads the results, and repeats until the task is done.

Skill

An instruction file the AI loads on demand: your process, inputs and rules, written once and run as a repeatable command.

MCP

Model Context Protocol: an open standard that connects AI to external software. A tool exposes its actions and data, and the AI calls them directly.

Agent team

Multiple agents in parallel, each with its own role and context, coordinated by an orchestrator that splits the work and merges results.

Harness

The structure around agents: rules, retries, validation and checkpoints that control how they run and catch failures.

Symlink

A file-system link that makes one folder appear in multiple places: one skills library, available to every tool on your machine.

My business runs through Discord.

One of you wrote on the intake form: "Jarvis seems cool." Good news. This is Jarvis.

1 · Message in Discord

2 · Gateway routes it

3 · Claude Code + skills do the work

4 · Result lands back in Discord

# jarvis

A

Alexander

/task build me this week's client report and send it over

J

Jarvis BOT

On it. Pulling the numbers, writing the report, deploying the page.

J

Jarvis BOT

Done in 4 minutes.

Weekly report ready
Live dashboard updated · summary written · link below ↓

Recreation of the real flow · I run this from my phone

The automation recipe.

1 · A schedule fires

2 · A skill runs

3 · Data lands in Airtable

4 · Ping hits my Discord

Any repeating, rule-based chore fits this exact recipe. The honest catch: the machine has to be running. Next slide.

# daily-reports

S

Skool Watcher BOT

Good morning ☀

Daily community report
Members: up again overnight · engagement climbing
Full numbers landed in Airtable → row 214

Recreation · this arrives every morning while I sleep

VPS or Mac mini? Decide in ten seconds.

The rule: never buy hardware for automations you have not built yet.

A live dashboard, in minutes.

Two of your intake forms asked for exactly this. It is the best first build there is: the data already exists, and nothing in your business can break.

Let AI read your finances.

Bank accounts (business + personal)

→

Fintable syncs them

→

Airtable

→

AI reads it

• Once your money has an address, AI does real work: a live finance dashboard, weekly spending reports, runway math.
• Same data, next step: client invoices generated for you.
• Your AI gets read access through the sync: it never touches the bank itself. I run my entire finance OS this way.

The Delegation Ladder. Every task climbs it.

1Do it · you, by hand, like always

2Document it · write the process down once: that is a skill

3Hand it to AI · the skill runs it while you watch

4Schedule it · it runs without you. It stopped being your task

Take any chore from this week and ask: which rung is it on, and what moves it one up?

MCPs: how AI plugs into your tools.

An MCP is a plugin that gives your AI native access to a service. No API code, no copy-pasting between tabs. You connect it once, and your AI can just use the tool.

You speak

→

Claude Code

→

MCP

→

Notion · Search · Browser

→

Result back to you

• Start with three: Notion (your second brain), Perplexity (live web research from the terminal), Playwright (your AI gets a browser: log in, scrape, fill forms).
• You will not install these by hand: the setup wizard in tonight's kit connects them for you.

Every tool has one job.

GitHub

Saves every version of your work. Your undo button for everything.

Vercel

Puts your work on a live URL in one command.

Airtable

A spreadsheet your AI can read and write. Your data's first home.

Supabase

The grown-up database, when Airtable starts creaking.

GoHighLevel

Funnels, CRM, email and SMS in one box. AI can drive it.

Close

A clean CRM for pipeline and follow-up, pure sales.

Zapier · Make · n8n

Visual glue between tools. Fine for simple pipes.

Trigger.dev

Runs code on a clock in the cloud, so your laptop can sleep.

You do not need all of these. You need to know which job each one does. Full plain-language rundown is on your resources page.

The first prompt is never a prompt.

Two directions. Then you stack them.

Back to front

Trial-and-error in the terminal until it works. Then one sentence: "condense everything we just did into a skill." The mess becomes a playbook.

Front to back

Write the playbook first, like a job description. Then refine it by running it. Best when you already know the process cold.

• Skills stack. One main skill calls others as subskills: my content skill calls my research skill, which calls my scraping skill.
• You design the behavior: interview me before starting, or always read from my data first. You write it that way once.

I do not follow AI news. My machine does.

Schedule fires

→

Scrape rising repos

→

AI reads the code

→

Compares to MY goals doc

→

Suggests full builds

• Same recipe with YouTube transcripts, mostly from smaller channels: that is where the real alpha hides.
• Do not chase the feed. Build a system that filters the world against your goals and hands you only what matters.

Never let AI guess an API.

Building something with Calendly, your CRM, a payment tool? The model's memory of their API is probably stale. Guessed endpoints fail silently and eat your afternoon.

In the starter kit this is one command: /api-docs calendly. And the kit ships with a library of pre-collected docs for the most common tools, so for those you skip the step entirely.

I read every one of your answers.

This room runs marketing agencies: renovations, brokers, contractors. And your answers rhyme.

Your first build is on this slide.

Start here. One evening each.

Win 1

A dashboard on data you already have: ad accounts, CRM, sheets. You watched it take minutes.

Win 2

One home for your data. Airtable first. Everything AI does gets easier once your data has an address.

Win 3

Automate the weekly client report. The exact chore two of you named on the form. Recipe: schedule, skill, Airtable, ping.

Three apps before you even touch AI.

Akiflow

One inbox for everything. Tasks, calendar, and every "can you quickly..." in a single keyboard-first command bar.

Beeper

All your chats in one app. WhatsApp, Instagram, iMessage, Telegram together, and your AI can plug into it later.

Arc

A browser built like a workspace. Profiles, spaces and tab hygiene that make the rest of this setup feel calm.

None of these are AI. All of them clear the desk your AI is about to work on.

Here is what you are about to get.

The takeaways

This whole talk as short pages, three minutes each.

These slides

The full deck, clickable, whenever you want it.

One-command install

Mac or Windows: Claude Code, my folder structure, my skills, in one paste.

The AI interview

A wizard connects your tools, then Claude interviews you and personalizes everything.

API docs library

Pre-collected docs for 45 tools, so your AI never guesses.

Who to follow

The creators worth your feed, with hand-picked videos.

The QR code is on the next slide. Cameras ready.

No funnel. No discovery call.

• After tonight, ask me anything on WhatsApp: +31 6 22 58 46 48 (or through the group). Free, no catch.
• Want to go deeper? Message me and we will look at your business together: what to build, how to build it, and how to stay off the dopamine treadmill.
• Custom builds together are a separate conversation with their own scope.