Heisentools: let an agent rewrite a mock.py until it mirrors any real tool's API
kieranklaassen · x · 2026-10-16
A playful but practical technique called "Heisentools" (borrowing kyle kukshtel's heisenfunction idea + autoresearch): stand up a fake service and have an AI agent rewrite mock.py until it mirrors the real tool calls it has observed and extrapolates to new ones. In the fakeslack example, mock.py answers the agent under test directly — no model runs, no real Slack touched — while a calibration loop scores its answers against recorded real replies and rewrites on low scores. This avoids rate limits, context drift, and polluting your database when testing agents. Tested on real Slack messages; works for any service (fakegmail, fakegithub...).
More from coding & agent
- Grok bot can't see its own usage limits, blocking swarm orchestration — adam_dorr · 2026-10-16
- Simular presents neuro-symbolic computer use agents at BayLearn — xwang_lk · 2026-10-16
- 66% of billion-dollar companies leave critical bugs unpatched, deflating AI bug-hunting hype — rseroter · 2026-10-16
- Claude Code 2.1.299 ships 190 changes; users advise disabling native auto-compaction — repligate · 2026-10-16
- Spring creator Rod Johnson on the Context Layer, launching Embabel Worlds knowledge graph — springrod · 2026-10-16
- 'Cuckcoding' meme: just sit and watch Codex work on your codebase — heyneighbor · 2026-10-16