The four tools referenced below are open source at . Each checks one boundary and keeps the final decision with the human operator.
How I Run an AI Software Factory: Distributed Agents, Boundary Checks, and Evidence Trails
- ▸ The four boundaries of an AI run
- ↳ 1. Input boundary — the instruction, before the agent sees it
- ↳ 2. Execution-authorization boundary — the command, before it runs
- ↳ 3. Execution-contract boundary — the result, before it is trusted
- ↳ 4. Output boundary — the public copy, before it leaves
- ▸ Why deterministic and local, not "ask another model"
- ▸ What this deliberately does not do
- ▸ The larger shape
SOCIAL SHARE CARD GENERATOR