Alexandr RichAlexandr Rich

Alexandr Rich · Founder, Bookbag.ai

AI agents are joining the team. I'm building the system to run them.

Engineering teams are adopting AI agents faster than they are building the controls around them. I run experiments on what actually happens when agents do real work, and I'm building Bookbag so companies can brief, route, supervise and audit them.

Watch the experimentsSee how Bookbag works

Founder of Bookbag.ai, the company behind Codebook and Madebook. 14 years building software, more than 10 leading engineering teams.

Alexandr Richmanage the agents.
Brief → work → check → approveEnforced, not just instructed.

A prompt is not a control system.

If a rule only lives in an agent's instructions, it's a suggestion.

The shift

Agents are becoming coworkers.
Nobody built the management layer.

Coding assistants waited for a person. Agents don't. They pick up work, change code, open pull requests and run in parallel, often across many repositories at once.

Teams have real processes for people: access, review, approvals, audit. For agents, most are running on prompts and hope.

That gap is what I build for, test and write about.

The questions every engineering leader is about to ask
  1. 01Which agent is working on what?
  2. 02What is it allowed to touch?
  3. 03Who approved this change?
  4. 04Why did it make that decision?
  5. 05What happens when two agents conflict?
  6. 06Can we prove what happened six months from now?

For engineering leaders whose teams already run

  • Claude Code
  • Codex
  • Cursor
  • GitHub Copilot
  • Devin
  • Kiro
  • GitLab CI/CD
  • CircleCI

The experiments

What actually happens when agents do real work?

Real agents on real repositories, measured. Each experiment becomes a video, and every result lands here, failures included.

Visit the Build Lab
  1. 01

    Can an AI developer pass QA without a human fixing its work?

    A coding agent ships a feature. A strict QA agent reviews it. Nobody steps in. How much survives, and why does the rest fail?

    In planning
  2. 02

    Five agents, one feature: where does coordination break?

    Five agents on one codebase at the same time. Who overwrites whom, what conflicts, and what it takes to merge.

    In planning
  3. 03

    Does an agent follow your architecture when nothing enforces it?

    The rules are in the prompt. Nothing checks them. Does the work pass every test and still break the architecture?

    In planning
  4. 04

    Instructions aren't permissions. What does an agent do with access it shouldn't use?

    The prompt says don't. The tools say it can. Which one wins, and how often?

    In planning
  5. 05

    Can you audit what an AI agent did six months later?

    Pick a merged change from an agent and try to reconstruct who asked, who checked, and why it was allowed through.

    In planning
  6. 06

    Can an AI software team maintain the same app for 30 days?

    Shipping is day one. Bug reports, dependency updates and real users are the other 29.

    In planning

Build along now: Can an AI remember the app it built yesterday?

What I'm building

Bookbag.ai

Run AI development. Keep control.

Coding agents are getting good at writing software. Bookbag is the layer around them: what they know, what they may touch, how work moves between them, who approves it, and the record of what happened.

  1. Brief
  2. Assign
  3. Work
  4. Check
  5. Approve
  6. Record
Start freeTalk to usHow it works

One Bookbag account for Codebook and Madebook.

Agent factories

Codebook

Lay out the work as assembly lines. Put an agent, a person, or an agent gated by a person on every station. Routing, retries, budgets and permissions are enforced by the platform, not trusted to the prompt.

codebook.workCoding governance

Madebook

Sits above every coding agent your team runs. Briefs each one from what is currently true, then posts one required check on every pull request. Agents that don't report don't merge.

madebook.ai

Free resources

Checklists for running agents.

Small, practical files for teams putting agents to work. Use them with whatever stack you have.

Get all the free resources

AI Agent Rollout Checklist

For engineering leaders moving from coding assistants to agents that act on their own: inventory, permissions, gates, audit and ownership.

Download free

Agent Project Brief

The one-page brief to hand an agent before it writes a line of code: the outcome, the constraints, what it may touch and what done means.

Download free

AI Release Checklist

The last pass before agent-written changes reach users: checks, rollback, costs, monitoring and who is watching after it ships.

Download free

Follow along

Three ways to keep up.