Blog

Each video gets a written walkthrough. Same content, different format — handy if you want to scan, copy commands, or read on a phone with sound off.

YouTube launch series

• What is an LLM, really? (plain English, no math) 3m 00s · shipping soon
  An LLM is your phone's autocomplete. That's it.
• The $15/yr Website Stack (no subscription trap) 3m 09s · shipping soon
  I figured out the best way to build websites.
• Why local LLMs beat API bills — the one-afternoon setup 3m 00s · shipping soon
  Most people in my feed are paying $30–$200 a month for AI. The same thing on your hardware costs $0 per request.
• Your first custom CRM in 2 afternoons (exact steps) 20m 00s · shipping soon
  This CRM took me two afternoons. It runs my whole roofing business now.
• The one mental model for how AI bots talk to each other 3m 00s · shipping soon
  Bots are restaurants. Once you see it, you can't unsee it.
• Folder structure for AI agents — the rule that saves your sanity 3m 00s · shipping soon
  Your folder structure is why your agents are confused.
• Claude Code vs. Cursor vs. n8n — plain English, which does what 3m 00s · shipping soon
  Claude Code, Cursor, n8n. Which one do I use? Trick question — they're on different layers.
• What a 'prompt' actually is — vs. system prompt, vs. context window 3m 00s · shipping soon
  Prompt, system prompt, context window — three words people mix up. Knowing the difference saves you hours.
• Planning bots — the category nobody talks about 3m 00s · shipping soon
  Most folks use one bot for everything. That's why their AI projects spin.
• Architect the project BEFORE writing code so Claude doesn't spin 3m 00s · shipping soon
  Asking Claude to code before architecting is like pouring foundation without a blueprint.
• When Claude is worth the $ vs. when a local model does the job 3m 00s · shipping soon
  Claude or local? Wrong question. Both. Here's how to pick.
• Removing 90% of your API spend — the audit framework 3m 00s · shipping soon
  You can cut 90% of your API spend. Most folks just have no idea where the money is going.
• Reading your own logs — why observability beats features 3m 00s · shipping soon
  Most people add features. Better people watch the system.

Short demo

• Local Model Dojo in 90 seconds (install → scan → benchmark) 1m 30s · shipping soon
  Your AI bill is too high.

Education series · Week

• Week 1 — Setting up your AI dev environment 15m 00s · shipping soon
  This is what week one looks like for most people. This is what it's going to look like for you.
• Week 2 — Planning with AI (architecture before code) 15m 00s · shipping soon
  Last year I shipped a working app. Then I deleted it. Here's why.
• Week 3 — Your first real website (HTML+CSS, no framework) 15m 00s · shipping soon
  Three pages. Live in 12 minutes. Fifteen dollars a year.
• Week 4 — Making it dynamic (drop-in JavaScript) 15m 00s · shipping soon
  Form submits. Email arrives. Zero React. Zero servers.
• Week 5 — Converting to WordPress (when it makes sense) 15m 00s · shipping soon
  WordPress isn't the enemy. The enemy is using the wrong tool.
• Week 6 — Real backend (FastAPI + SQLite) 15m 00s · shipping soon
  One Python file. A real API. Free docs.
• Week 7 — Stripe payments (accepting money online) 15m 00s · shipping soon
  Stripe is the cleanest API I deal with — by the end of this video you take a real test charge and watch it land in your dashboard.
• Week 8 — Automation flows (n8n + your own site) 15m 00s · shipping soon
  TODO: write the hook line here
• Week 9 — Using local LLMs to replace API spend 15m 00s · shipping soon
  $211 to $19 in one routing rule — same workflows, no quality loss.
• Week 10 — Production deploy (domain, DNS, HTTPS) 15m 00s · shipping soon
  Three subdomains, one domain, $15/year + $5/mo — the localhost-to-production flip in 15 minutes.
• Week 11 — Observability (what to monitor, what to ignore) 15m 00s · shipping soon
  Three signals — front door, code errors, email delivery — catch 95% of production failures. Everything else is noise.
• Week 12 — Selling it (marketing plan in a day) 15m 00s · shipping soon
  Five conversations beats five thousand views — the LMD launch playbook applied to your product, in one day.

Workflow module · Lesson

• Lesson 1 — Intake call (6 questions + auto-transcription) 20m 00s · shipping soon
  If you screw up the intake, you screw up the build — here's the 10-minute call I run every time.
• Lesson 2 — Obsidian folder discipline (one folder per customer) 20m 00s · shipping soon
  Anyone can build a website — almost no freelancer can run five customers without dropping a ball. The folder is the moat.
• Lesson 3 — Claude Code Plan Mode (full plan before code) 20m 00s · shipping soon
  Plan first. Code second. Always — the discipline that kills 60% of mistakes before a single line of HTML gets written.
• Lesson 4 — Live Server preview (the customer demo) 20m 00s · shipping soon
  The free VS Code extension that turns the customer Zoom call into a magic show — and bumped my close rate fifty points.
• Lesson 5 — Real-time edits on Zoom (the wow moment) 20m 00s · shipping soon
  Customer says 'make the headline bigger and red' — 30 seconds later it is, on their screen, and that's how the deal closes.
• Lesson 6 — Pricing the speed (why 45 min ethically charges $1k) 20m 00s · shipping soon
  Customer is paying for the years that made today possible — not the 45 minutes. Don't punish them for hiring an expert.