Two figures sit together on a rooftop at night under a string of warm bulbs, a laptop glowing between them, a city skyline behind.

AI does the busywork. You keep the final say.

I sit down with you and build your AI setup on your own machine, in your own accounts. Then I write down what it may never touch, and what it has to ask you about first.

Request a session See how it works
By request · 2 places a month One to oneYour machine, your keysYou keep everything
How it works

Most setups fail quietly. That is the expensive one.

A crash gets fixed on Tuesday. A job that exits clean having done nothing gets found in March. The difference is whether anything was written down.

01

It knows its edges

Written down, not implied: what it may read, what it must never touch, and what it has to ask you about first. Those three answers are your standards document.

02

It stops before it matters

Anything outward — a send, a delete, a charge — waits for you. Everything else just runs. That gate is the difference between useful and frightening.

03

It leaves a receipt

Every run says what it did, what it skipped and why. You can tell success from silence, which is the one thing most setups genuinely cannot do.

weekly-support-digestrun 0041

A real run, paused at its gate. Approve or hold it — the page is live.

Experience

I run this on myself first.

I am not a consultancy. I am one person who has been running a production agent system every day for months — memory, scheduled jobs, approval gates and a verification layer that catches its own mistakes.

Everything I would build for you is already running against my own work, and it has caught me more than once. That is the whole qualification, and it is the only one I will claim.

See the packages
  • DailyA system that runs without meScheduled jobs, a morning brief, and gates that refuse work when something is unverified.
  • 4Questions become your standardsWhat it may touch, what it never touches, what needs your approval, what it costs if wrong.
  • 2–3Sittings, then it is yoursNo retainer, no lock-in, no monthly fee. One follow-up, then the engagement ends.
  • 0Of your credentials I holdYou sign in, you keep the keys, you pay your AI provider directly.
What people built

Four people, four different jobs, the same two sittings.

None of them were engineers. One had never opened a code editor. They each paid the same $950 and walked away with something running on their own machine.

Starting from zero
“I’d never opened VS Code. Two days later I have a system that remembers what I’m working on and keeps my tasks in order, and I understand how it works.”

Caitlin G.Had never written a line of code. Keeps building on it herself.

CG · $950 · 2 SITTINGS · PAID

Email and calendar
“Most of my day was email and calendar. Now it’s a fraction of the work. It’s not a demo I play with; it’s the thing I actually use every day.”

Jeffrey P.Works at a travel agency. Uses it daily.

JP · $950 · 2 SITTINGS · PAID

Agents that take phone calls
“He got me set up from scratch. From there I built agents that take phone calls. I went from not knowing where to start to running something real.”

Paul M.Turned it into a service he now sells.

PM · $950 · 2 SITTINGS · PAID

His own project, not a tech project
“I came in with my own thing to work on. He set up the foundation and we pointed it at that. That’s the part that made it stick.”

Jacob S.3D printing. Came in with the work, not the tooling.

JS · $950 · 2 SITTINGS · PAID

Four engagements, four unrelated domains, $3,800 paid, no refunds, five stars from all four. Quotes and ratings used with each client’s written permission. None of them claims a revenue result, and neither do I.

Pricing

Pick by where you are.

Two are for someone who has not started. The third is for someone already running agents who wants to know what is wrong with them.

Starter
$950 fixed
2 sessions · ~2 weeks · 1 use case
  • Editor and Claude Code running on your machine
  • The tour — only the parts you will use
  • Plan, research and execution kept apart
  • Your standards, written down
  • One use case, built and working
Book Starter
START HERE
Standard
$1,500 fixed
3 sessions · ~3 weeks · 2 use cases
  • Everything in Starter
  • A second use case — $550, not $950
  • Architecture is built, so job two is mostly wiring
  • Both share one set of standards
  • Scope agreed in writing before we start
Book Standard
Audit
$500 review
90 minutes · credited · already running agents
  • No setup — you already have the tools
  • Where your data can actually go
  • Whether anything grades its own work
  • What your jobs do when they fail quietly
  • Written findings, ranked. Full review from $2,500
Book an Audit
Not sure which one? Say so in the form below and tell me what you are trying to automate. I will tell you which package fits, or that none of them does.

One real finding, monthly

Something my own gates caught in my own system, with the receipt. No pitch, one click to leave.

Questions

Before you ask.

One job you do repeatedly, with a clear start and finish. “Every Monday, turn last week’s support email into a tagged summary” is one. “Help me run my business better” is not — that is a wish, and I will say so in the first session rather than at the end.
No. The first session assumes you have never opened a code editor. If you write code already, we skip the tour and spend the time building instead.
You do, directly to the provider, in your own name. I never hold your credentials, your card or your API keys. If I disappear tomorrow, nothing of yours stops working.
You get one 30-minute follow-up inside 30 days, included. After that it is an hourly session. I do not sell retainers, because I will not promise a response time I cannot keep.
No. If the price does not fit, I cut the scope to fit it and tell you exactly what is no longer covered. That is a better deal than a cheaper version of the same promise.