1 / 7
Open Source
Be My Butler

End AI coding
self-verification bias

10 Agents · 12 Steps · Cross-Model

Multi-agent orchestration for Claude Code
Cross-model blind verification + Analytics pipeline

01 The Problem

AI coding is fast, but
dangerous

🪞

Self-Verification Bias

The same model writes and reviews
— blind spots never disappear

🌀

Context Explosion

In long conversations AI loses direction
and forgets critical decisions

🎯

Edge Cases Missed

Blinded by speed,
boundary tests get skipped

🔧

Design Tunnel Vision

Fixated on the first idea
— better alternatives unexplored

02 The Pipeline

12-step pipeline
starts with one command

/BMB
PLAN
Setup
Brainstorm
Council
BUILD
Arch
Execute
Frontend
VERIFY
Test
Verify
REFINE
Simplify
⑩½Analyst
Learn

Bird's Law analytics · Every step auto-orchestrated

03 Architecture

Agent Orchestration

Lead Consultant Arch Executor WORKTREE ISOLATION · FILE-BASED HANDOFFS Frontend Tester BLIND WALL Cross-Model Verify Analyst · Bird's Law

10 agents · Lead orchestrates via tmux panes · Analyst retrospective per session

04 Killer Feature

Cross-Model
Blind Verification

Claude
Write + Test
BLIND
Codex / Gemini
Independent Review
🔀
Divergent Framing

Intentionally reframed specs sent to verifier.
Same code, different perspective.

🛡️
Assumption Leak Detection

Judges code correctness independently,
not the original model's assumptions.

05 For Everyone

One command,
two wings

🏗️

For Experts

  • Cross-model blind verification
  • Council Debate design optimization
  • 3-Tier auto-learning system
  • Production-grade code quality
Higher code quality
/BMB
🚀

For Beginners

  • No coding jargon needed
  • Just describe your idea
  • Students, non-devs welcome
  • Ask questions anytime
Fearless start

Other AI coding tools
optimize for speed.
BMB optimizes for accuracy.

10Agents
12Pipeline Steps
Learning Cycles
Get Started on GitHub

MIT License — Use it however you want.