Agent builder says he runs with all permission checks off and relies on reproducible Mac setup
blelbach · x · 2026-07-29
The author says they disable permission checks and auto-review for agents, and explains why in three points:
- They treat the machine as an employee’s device, not a personal one, so constant approvals are seen as a productivity tax.
- The real safeguard is a fully reproducible setup: if things go wrong, the machine can be wiped and rebuilt from scratch.
- The linked public dotfiles repo shows the exact Mac setup managed with nix-darwin and home-manager, so a fresh machine can be configured the same way every time.
More from coding & agent
- A production inference checklist spans vLLM, SGLang, quantization, and load testing — ZeYanjie · 2026-07-29
- Claude Opus 5 allegedly built a Sea of Thieves clone with generated assets and game systems — ChrisGPT · 2026-07-29
- Agent loops can quietly turn your eval set into training data — Vegetable-Rub-8241 · 2026-07-29
- Hermes Agent speeds up voice chats by streaming TTS after the first clause — Teknium · 2026-07-29
- GitHub repo curates 100+ libraries for LLM apps, RAG, agents, and testing — tom_doerr · 2026-07-29
- Claude and Codex open-source programs reached an estimated $77,685 in API value — rudrank · 2026-07-29