Notes from a founder building a dozen ventures alone, with agents.
Earned Autonomy · Part 1
9 min read
Agents got good at producing work. Deciding it's done is the hard part now. What the field learned about loops, judges, and evals, and how I'm changing Faber.
Read more →
Earned Autonomy · Part 2
10 min read
Agents grade their own work generously. Why a separate judge with evidence and tools beats self-review, what it should see, and what it must never be told.
Read more →
Earned Autonomy · Part 3
9 min read
An agent stops when the work looks done. A four-layer definition of done, written before the build and enforced by code, is how looks done becomes is done.
Read more →
Earned Autonomy · Part 4
9 min read
The engineer quoted for running thousands of agents still supervises about ten sessions. What scale takes: code that owns the loop and a check that's cheap.
Read more →
Earned Autonomy · Part 5
10 min read
Every part of an agent harness is a bet on what the model can't do yet. Which parts go stale at each model upgrade, which never do, and how I'm pruning mine.
Read more →
Earned Autonomy · Part 6
9 min read
A validator decides whether one agent run passes. It can't tell you whether the pipeline got better. How to build evals that measure the judge, not the work.
Read more →
Earned Autonomy · Part 7
9 min read
Run a 90%-reliable agent ten times and the odds that every run succeeds are about 35%. The arithmetic of trusting agents nobody watches, and what to measure.
Read more →
Earned Autonomy · Part 8
35 min read
The full research note behind the Earned Autonomy series: how code owns the loop, a separate judge decides done, and evals show whether any of it helps.
Read more →
Studio of One · Part 1
7 min read
One founder, no employees, a dozen ventures at different stages. What alone actually means in an agentic studio, what runs without me, and what still doesn't.
Read more →
Studio of One · Part 2
7 min read
Bolt, Lovable, and vibe coding get you a demo fast. Shipping a real system takes agentic practices: specs, checks, guards, and review. Here are the ones I use.
Read more →
Studio of One · Part 3
7 min read
A week of shipping in a studio of one: the decisions, the pull requests, what Claude Code did, what I did, what got cut, and how I knew it was safe to merge.
Read more →
Studio of One · Part 4
6 min read
Why I built Make First for my kids: a one-semester path from an idea to a first sale to a stranger, built with AI tools instead of code, with gates on purpose.
Read more →
Studio of One · Part 5
7 min read
First Idea, First Build, First User, First Dollar, First $1K, First $10K. Why Make First ends every stage at a First with evidence, and what each gate tests.
Read more →
Studio of One · Part 6
7 min read
A dozen ventures, one founder, finite hours. The rules I use to decide what gets built, what gets parked, and what gets killed, with real calls from this year.
Read more →
Studio of One · Part 7
6 min read
Design notes from Make First's private beta: why the AI cofounder checks, coaches, and unblocks a teen founder but never writes the product or the posts.
Read more →
Studio of One · Part 8
7 min read
The Foundry is a free, invite-only community for solo founders past $100K. Why it costs nothing, why we take no equity, and how the studio makes money.
Read more →
Building for the Agentic Age · Part 5
10 min read
Every platform promises speed and delivers dependency. Where no-code and hosted stacks stop, when they're still the right call, and how I own what matters.
Read more →
Building for the Agentic Age · Part 4
6 min read
I retracted the claim that ten people can build what took a hundred. The number I have evidence for is one. The real math of a studio of one, gaps included.
Read more →
Building for the Agentic Age · Part 3
9 min read
Codex, Faber, Forge, and Helm: what each tool does today, what's still coming, why I built them to run my own ventures, and why they're Apache-2.0 open source.
Read more →
Building for the Agentic Age · Part 2
10 min read
AI stopped being a tool and became infrastructure. What that changes for a founder building alone: structure, the hire you skip, moats, and where the hours go.
Read more →
Building for the Agentic Age · Part 1
10 min read
The business that gives you the most freedom is the one that needs you least. Why I design every venture for my own absence, and the three levels I run.
Read more →
Two doors
Know a teen who should build their first real business? Make First is taking applications. Already past $100K on your own? The Foundry is forming.