Humanmaxxing
Menu

Product · Automated self-improvementDraft

Your AI engineering setup reviews its own work.

Every week, a retrospective agent reads what actually shipped — merged changes, review findings, tickets, and the delivery scorecard on each pull request — and finds the patterns that waste human time. Every finding arrives as a specific proposed change to your rules or tooling. Nothing changes until you approve it.

Book a demoRun /retro yourself (open source)

How a week becomes a rule

Machine · runs without you

Collect

Each week the retro gathers everything your workflows already produce. No new instrumentation.

  • Merged PRs, commits, issues opened and closed
  • Review findings and per-PR delivery scorecards
  • Per-session retros written after each PR

machine runs human reviews

What you get

Weekly retro

Reads the week’s merged work, reviews, tickets and scorecards and reports what shipped, what broke and what keeps recurring.

Per-session retro

After every PR, a lightweight agent writes a short, secret-scrubbed retrospective of the session that only you see.

Review-findings log

Every review writes its findings per branch, so the retro can see which agents are pulling their weight and which only add noise.

Audit skills

On-demand sweeps for silent errors, end-to-end and API coverage, and a five-pass full audit, each reporting a class of problem rather than an instance.

Advisor compliance watch

The orchestrator monitors the team for rule violations during the run and challenges claims of done before they reach you.

Guard budget

New checks and hooks are earned by recurrence and measured against a monthly budget, so the setup gets sharper, not heavier.

Questions

What is open source and what is hosted?

The weekly retrospective skill is in the repo and runs by hand today. The hosted service schedules it per organization, keeps the history so trends are real, adds the per-session retros, and ties findings to the analytics.

Does it read my code or prompts?

The weekly retro reads git history, PR metadata and the notes your workflows already write. The per-session retro reads a scrubbed extract of the session, never raw prompts or secrets, and the result is private to you.

Can it change rules automatically?

No, by design. The retro is read-only. It proposes; a human disposes. That is the point of the framework.