Put rules in code, not prose: what actually makes an AI instruction a Rule
gerardsans · x · 2026-11-18
- Replying to SparkTsaiX's Behavior Rule Architecture series (02/06), gerardsans argues that any rule an AI must always follow should live in code, not plain text a model can forget or dismiss.
- The article distinguishes normative strength: "Please avoid…" vs "You should not…" vs "You must not…" point the same way but aren't the same rule; moving an instruction into a file makes it reusable but doesn't automatically make it a Rule.
- Practical takeaway for agent engineering: hard constraints belong in code/validation layers; prompts carry soft preferences only.
More from coding & agent
- Microsoft ships MAI Code 2 and Code 2 Flash coding models into GitHub Copilot — marlene_zw · 2026-11-18
- Microsoft launches Agent 365 Business Analytics to track agent spend and ROI — donasarkar · 2026-11-18
- YC-backed Ample launches one-stop cloud for agents, hits 5,000 users two weeks after launch — ycombinator · 2026-11-18
- OmniLoRA: train any model on any GPU inside your ComfyUI graph — pente5 · 2026-11-18
- findatasets ships /feedback endpoint so AI agents can file bug reports, following Coinbase's lead — virattt · 2026-11-18
- Dev shares a tiered file-permission policy to stop agents stalling on approvals — gaganghotra_ · 2026-11-18