About

Why I built act101.

Two things people keep asking: why I'm the one building this, and why you'd reach for a procedural tool when your agent can already write code. Here are both answers.

The builder

Decades building software that has to be correct.

I've spent a career building production software across financial services, geospatial, automotive, medical devices, and biotech — domains where a defect ships in a trading system, a navigation unit, a vehicle, or a life-critical device. I've led engineering teams and owned architecture through every class of problem the field presents.

That's why act101 isn't a guess at what engineers need. I know the challenges engineers and architects face because I've faced them in domains that don't tolerate error — and I know the canonical tooling those challenges demand. act101 is that tooling, rebuilt for agents.

Domains shipped into
Financial services Geospatial Automotive Medical devices Biotech
Why not just let the agent do it?

An LLM editing your code is a guess. act101 is a transform.

Codex can refactor. So can any agent. The question is what you're trusting.

Let the agent freestyle it

Non-deterministic, expensive, opaque. It re-reads files to guess at callers, burns tokens per turn, and leaves no clean way back. Change a method signature and it grinds through file after file one at a time — and you hope it caught them all.

100s of turns

Call act101

Procedural, structurally-grounded, cheap, reversible. Change that signature across all 25 affected files in a single tool invocation — every reference updated, with a receipt that proves it, and rollback if you don't like it.

1 invocation

Same job. One is a probability. The other is a proof.

The story

act101 is the verification layer I built for myself.

Before I'd let an agent merge a million-plus lines into a binary I was going to ship under license, I had to be able to prove every change. Generation was never the bottleneck — agents generate all day. The bottleneck is prove: writing down what a change actually did, with evidence, against the contract it was supposed to preserve.

So I built that layer. Every act101 workflow has the same shape — analyze → act → attest — and they're the workflows I ran on my own codebase, every day, for ninety days. The product is the recursion: what I shipped is the toolchain I needed to ship it.

The bottleneck of agentic software is not generation. Agents generate. The bottleneck is proof.

Try it yourself

One native binary, every grammar, zero setup. Builder Edition is free forever.

Install — 60 seconds

Stay in the loop

Ship notes and new capabilities, occasionally. No spam; unsubscribe anytime.