Early-stage founders
You're the product thinker. You've got a small engineering team or none at all. You want ideas to reach production without a translation layer eating your runway.
fit · high+ JB/05AI ⁄ NOWEST. 2024AGENTIC ⁄ 0101 ⁄ Thesis 02 ⁄ Before ⁄ After 03 ⁄ Principles 04 ⁄ Engagement 05 ⁄ Who Type A Type B Type C
Agentic systems for teams that want to ship like a studio, not a committee.
The bridge I’m building.
The gap between product, design, and development has always been a translation problem. Ideas get written down, handed off, misread, rebuilt, compromised. Each step leaks a little of the original intent.
Agentic systems — done well — collapse that gap. The people with the sharpest instinct about the product don’t have to wait for a technical intermediary to turn their thinking into software. They describe what they want, correct what’s wrong, and the agent handles the technicalities.
For the last year I’ve been building these workflows — design → brainstorm → plan → implementation → test — inside real teams. The outcome isn’t “AI doing the work.” It’s non-technical people contributing to a technical solution, and technical people moving at a pace the market now demands.
If you’re running a company where product ideas die on the way to engineering, or engineering is the bottleneck on shipping, this is what I do.
How the team works now.
Before
A translation pipeline.
- PM writes a ticket. Designer drafts a flow. Engineer interprets both into code.
- Three rounds of review to align intent with implementation.
- Non-technical ideas stall at the engineering intake.
- Every small change requires a specialist.
- The team’s velocity is bounded by the narrowest funnel in the pipe.
After
A direct contribution loop.
- Anyone in the company can describe, try, and refine a feature in the real codebase.
- The agent handles the technicalities — types, tests, patterns, compliance.
- Engineers spend time on architecture, ambiguity, and judgment — not mechanical work.
- Product decisions happen next to the product, not three meetings away.
- The team’s velocity tracks the sharpness of its ideas, not the size of its eng team.
How I design agentic systems.
- P 1Democratize the input.The best idea in the room rarely belongs to the person with commit access. Give everyone a path from thought to production.
- P 2Keep the craft bar.Agents don't replace taste. They replace typing. The system has to make the good thing the default thing.
- P 3Specify the end state, not the keystrokes.Users describe outcomes. The workflow handles decomposition, tooling, and the thousand small technicalities in between.
- P 4Make correction the primary UI.Nobody gets the prompt right the first time. The loop — try, see it, redirect — is the product.
- P 5Earn trust with every commit.Tests, types, review gates, reversible actions. Non-technical contributors can ship only if the system refuses to let them break things.
- P 6Stay legible.Agentic output nobody on the team can read is future technical debt. Favor legible code, legible decisions, legible paper trails.
AI workflow consulting, in practice.
We start with a real delivery bottleneck: a prototype that cannot reach production, repetitive implementation work, or a team that needs a safe way to contribute through agents.
The engagement covers workflow design, agent tooling, and coaching. We work through your actual product and codebase, with review, tests, and handoff practices your team can keep using.
Typical engagements run 6–12 weeks, with scope agreed around the team’s constraints. The goal is a repeatable path from an idea to a reviewed, deployed change.
Who this is for.
Product & design leads
You're tired of watching crisp ideas become compromised tickets. You want your team to prototype, test, and ship without begging engineering for every iteration.
fit · highEng orgs at the bottleneck
Your engineers are strong but can’t scale with demand. You want to lift the whole team — giving juniors seniority, seniors leverage, and non-eng contributors a real seat.
fit · strong