Four weeks. A shipped tool. A Skills library you own.
A habit that sticks.
A live, hands-on cohort for PMs, operators, consultants, and indie builders ready to ship the real thing. Next cohort starts May 12.
A Skill is a saved playbook your AI loads automatically — your rules, examples, and process in one file you can reuse across projects. Stop re-explaining the same thing every session.
From Chatbot to Builder
Four weeks of live, hands-on sessions. You bring a real problem you've been promising to fix. You leave with a tool that's live, your own Skills you'll reuse, and the habit to keep building without me.
- One 90-min working session (Tue 6pm ET) + one 60-min office hour (Thu 6pm ET), all recorded
- Private Discord, weekly ship checkpoint, async "I'm stuck" support
- Spec-first workflow with CLAUDE.md, plan mode, and the one-change rule
- Graduation: your first shipped tool that gets used, a Skills library you'll actually reuse, and a next-project brief
The course updates every cycle
AI tools move fast. Other courses go stale. This one rebuilds the examples around what's working now. Recent updates:
- Apr 28, 2026 · Added the AGENTS.md + CLAUDE.md harness flow as the Week 0 starter.
- Apr 21, 2026 · Added a Skill-based context reset pattern using
/clearfor long sessions. - Apr 14, 2026 · Added a Codex / Cursor fallback workflow for when Claude Pro limits bite mid-build.
- Apr 07, 2026 · Added lightweight evals so you can confirm an AI tool actually does what you asked.
- Mar 31, 2026 · Added the seven-question scope filter that picks the lane on Day 1.
What we build, week by week
-
Week 00 Pre-flight · before May 12
Get the stack ready before we start
Setup is where beginners lose momentum, so we separate it from the real work. One short video, one checklist, one readiness gate. Show up to Week 1 with the tools actually working.
- Claude Code + desktop app
- GitHub + Vercel
- Secrets 101
- CLAUDE.md + Skills scaffold
-
Week 01 May 12 – May 18
Pick, spec, and ship v1
The project you pick this week is the one you ship. We protect scope on purpose. By Sunday you have a real URL and a spec Claude can read before every session.
- Four project lanes
- CLAUDE.md + your first Skill
- Plan mode + one-change rule
- Deploy v1 to Vercel
-
Week 02 May 19 – May 25
Debug without spiraling
The ugly middle stays in. You learn to read errors without panicking, debug with screenshots instead of descriptions, and use /clear with your Skill to reload only what matters. Also: the Codex or Cursor fallback for when Claude Pro limits bite.
- Error triage
- Screenshot debugging
- /clear + Skill-based context reset
- Codex or Cursor fallback
-
Week 03 May 26 – Jun 1
Make it look real
Move from "it works" to "someone else can use this." Basic product taste, UI copy rewrites, and a silent watch session with one real user. Run lightweight evals to confirm the tool does what you asked, then pick three fixes and ship them.
- Hierarchy, spacing, contrast
- UI copy pass (with a copy Skill)
- Watch one real user + lightweight evals
- Polish ship
-
Week 04 Jun 2 – Jun 8
Build the habit
Finish strong and leave with momentum. Public build story, your own Skills you'll reuse on the next project, a named map of what's next (subagents, MCP), and a next-project brief you already approve of. Demo day closes the cohort.
- Build-story writeup
- Your own Skills set
- Subagents + MCP (named, not taught)
- Demo day
The right fit, and the wrong one
Right fit
- You're a PM, operator, consultant, or indie builder
- You've used Claude, or shipped something simple in Lovable, Bolt, or basic Claude Code, and you're ready for the real thing
- You have an idea you'd build if you weren't stuck in "I need a developer" mode
- You can commit 2 to 3 hours of live time per week for four weeks
Not the right fit
- You're looking for a tour of every AI tool, framework comparison, or model architecture primer
- You've never used Claude Code/Codex or a no-code/low-code AI build tool. We recommend starting with Lovable or similar
- You want a passive, watch-and-learn video library
- You need enterprise compliance, SSO, or team-wide rollout support. Let's talk consulting instead
FAQ
Do I need to know how to code?
No, but you need to have used Claude Code/Codex or a no-code/low-code AI build tool at least once. If you haven't, start with Lovable or similar, then come back.
Anthropic launched a free Academy with Claude Code courses. Why pay $497?
Anthropic teaches the tool. The cohort is what makes you finish. A Tuesday working session, a Discord where "I'm stuck" gets answered inside a day, a refund through week one if you don't ship. Plenty of people learn fine from the free path. This is for the ones who've watched the videos and still haven't built the thing.
The AI space moves fast. Will what I learn still apply in three months?
The principles, yes. The specifics, by design. The cohort runs on what's working this month: scoped outcome-based asks (the new pattern for Opus 4.7's literal instruction-following), matching model thinking effort to task complexity to save tokens and time, lightweight evals to know your tool actually works, and building Skills you'll reuse instead of prompt libraries you'll abandon. Anthropic ships behavior changes faster than the docs catch up. The cohort closes the gap.
What's the time commitment?
One 90-minute working session (Tuesday 6pm ET), one 60-minute office hour (Thursday 6pm ET), plus about 2 hours of your own work between sessions. Around 5 hours a week for four weeks.
What if I can't make a live session?
Every session is recorded and posted within 24 hours. The Discord stays active for async questions, and the "I'm stuck" template gets you a real answer inside of a day.
Will I hit Claude Pro usage limits?
Probably, late in the cohort. Week 2 covers the Codex or Cursor fallback so you're never blocked. Claude Pro is the minimum. If you're building heavily, the Max plan is worth it, not required.
What stack will I use?
Claude Code as the spine. GitHub for version control. Vercel for deployment. Codex or Cursor introduced in Week 2 as a fallback. All free or near-free tiers work for the month.
Is there a refund policy?
Full refund through the end of week one, no questions asked. After that, a 50% refund through week two. The founding cohort is limited, so I want you fully in or fully out.
Can my company pay?
Yes. Invoice with a W-9 and a line-item receipt. Most students expense this as professional development.
Will there be another cohort if I miss this one?
Yes, monthly. Future cohorts run $697+. The founding-cohort price and direct access are first-cohort only.
From people I've worked with
1:1 session"Vincent pushed me to prioritize building over overthinking and simplified the daunting world of AI coding agents. If you need someone to help you cut through the noise and actually ship your ideas, Vincent is the person to talk to."
Consulting client"Vince has a rare ability to translate technical ideas into business decisions: how to reduce human bottlenecks, where oversight matters most, and how to think about AI adoption in a way that's structured and responsible."
Not sure you're ready? Start with the free agent harness, scope filter, and an optional 60-second Vince Loom take on your idea. Use it on a real project, then come back. Get the free harness →
24 seats left. Future cohorts run $697+.
Founding-cohort pricing and direct access to Vince are locked for this cohort only.