operator sprint

Become the operator who got AI right.

You're already doing the hard part — running the whole thing yourself. The question isn't whether to use AI. It's how to get the leverage everyone's promising without losing a month to tools that don't fit how you work. So I work it with you, in two steps.

Book 20 minutes See Bitter pricing

Founder-led and one operator at a time. Fixed fee, scoped on the call — not engineering hours.

step 01

one to two weeks

AI strategy sprint

I watch how you actually work — where your time goes, what's repetitive, what's high-stakes, where context lives and where it leaks.

You come out with a concrete AI-transformation plan: the two or three highest-leverage things to change, in priority order, with the why written down so it's yours to keep. The deliverable is a plan you could hand to anyone — including future you.

  • A read on how your work really runs
  • The 2–3 highest-leverage changes, prioritized
  • The reasoning written down, not a slide deck
  • A plan that's yours to keep

step 02

a short embedded build

Forward-deployed engineer sprint

Then I don't just hand you the plan — I embed and build the top item with you, inside your environment, on your real workflow.

We ship something that runs after I leave. You're not buying engineering hours; you're buying the thing actually existing in your business when the sprint ends.

  • I build inside your real workflow
  • The #1 item from the audit, shipped
  • It keeps running after I'm gone
  • Left in a workspace you own and can improve

where the work lives

The sprint runs inside Bitter.

I don't hand you a plan and disappear. The build runs in a hosted Bitter workspace — a terminal, your files, and an agent surface — so the work has a home you keep and can keep improving after I'm gone. You don't manage infrastructure or have to be the one who SSHes in.

why me

Twelve years as an early engineer at a venture-backed startup, then I left to do exactly this: build the receipt-bearing loops around AI work so each run compounds.

I build around two ideas — the Bitter Lesson (general methods that scale with compute win; don't overfit to one tool) and Amdahl's Law (the human is the serial bottleneck, so the job is designing the environment and the loops). In practice that gets operators from "I have an AI prototype" to "I have a real system that runs."

Book 20 minutes