Alexandr RichAlexandr Rich

The person behind the experiments

Hey, I’m
Alexandr Rich.

I’ve spent 14 years building software, and more than 10 of them leading engineering teams.

Then AI agents started showing up on those teams. Not as tools a developer drives, but as workers that pick up tasks, change code and open pull requests on their own.

The gap was obvious. We had real processes for people: access, review, approvals, audit. For agents we had almost nothing. They were being managed with prompts and hope.

So I’m building Bookbag: Codebook to organize agents into software factories, and Madebook to govern their work where it merges. And I run experiments in public on what agents actually do, failures included.

Alexandr Rich

A prompt is not a control system.

Bullish on agents

I’m not against AI agents. I think they will do a large share of software work. That is exactly why they need a system around them.

Enforce, don’t trust

Never accept an agent’s word that the work is done. Build the check that proves it, and put it where the agent can’t skip it.

Show the evidence

Run the experiment, measure it, publish the failures. Separate a promising result from a claim nobody has tested.

Watch the experimentsSee how Bookbag works