Stop vibe coding. Start engineering agents.
Agentic engineering is a 10x multiplier on the judgment you already bring — or a 10x mistake-amplifier without the scaffolding. I help you land on the right side of that. On your real code, one module at a time.
Same tools. Different discipline.
Vibe coding and agentic engineering look identical from across the room — same editor, same model, same keystrokes. They are not the same thing. The gap is the entire job.
Prompt. Glance. Accept. Move on.
- close No spec — the goal lives only in your head
- close No eval — “done” means it ran once
- close Review is a vibe check, if it happens
- close Great for prototypes — collapses on real users
Result: mistakes, shipped fast.
Brief. Eval. Orchestrate. Gate.
- check Briefs are contracts — goal, constraints, done
- check Evals define done — the agent can’t fake it
- check Output is checked against a review pass
- check Survives real users — because it was built to
Result: compounding leverage.
What I install.
Six pieces of scaffolding that turn the multiplier real. We build them live, on the project you actually care about — not toy examples.
Briefs & evals
Stop describing tasks. Write contracts — goal, constraints, and an eval the agent can’t fake done on.
The review pass
A short standard for what your team considers ready to ship. Encode it as a check the agent runs against its own output.
Design discipline
A DESIGN.md the agent reads on every task — so the output looks intentional, not generated.
The second brain
Stop re-explaining your stack. Build the context layer — CLAUDE.md, playbooks, memory — that compounds.
Orchestration
Run parallel agent streams and keep them coherent. The discipline that scales per-person, not per-hire.
Review gates & shipping
Where mediocre gets caught — and the habit of never claiming done without proof.
Who this is for.
Most AI consulting is priced for Fortune 500 budgets. This is for people who want to actually get good at the work.
You can build. You can prompt. You suspect you’re leaving most of the leverage on the table.
You are — and it’s not a talent gap, it’s a scaffolding gap. We close it on your codebase, one module at a time, until orchestrating agents is your default mode.
You want to ship something real in a weekend — not drown in mediocre output at hour 30.
Five hackathons of hard-won lessons, compressed into a pre-event session: what to set up before the clock starts, how to run parallel streams, what actually wins.
You rolled out AI coding. The gains aren’t compounding.
Because the transition is a re-org, not a tool install. We build your team’s review pass, eval infrastructure, and orchestration habits — together.
The curriculum is open.
The full 8-module path — and every artifact, the review-pass skill, the DESIGN.md template, the second-brain starter — is free and public. Fork it. Run it. Argue with it.
Mentoring is for when you want a guide through it — on your real work, at your pace.
The curriculum
Vibe coding → agentic engineering, with exercises you run on real work.
The review pass
A 10-section review pass — fix-or-justify — the agent runs on its own output.
Second-brain starter
CLAUDE.md skeleton, playbook stubs, a one-fact-per-file memory structure.
How it works.
No SOW. No pitch deck. The first session is on me.
Free 30-min session
Where you are, where you’re stuck, what you’re building. We decide if this is a fit — you leave with something useful either way.
Pick a track
One module deep, the whole curriculum over a few weeks, or a focused team rollout. Scoped to your real work.
Work on your code
We install the scaffolding live — briefs, evals, the review pass, your second brain — on the project that matters.
You own it
Everything stays in your repos. No lock-in. The whole point is that you don’t need me after.
Want a guide through the jump?
Free 30-minute session. Tell me what you’re building and where you’re stuck.