Built a telco anomaly detection system in 3 days: 'the code doesn't matter, the cookbook is the product'
l4rz · x · 2026-11-09
The author describes building and shipping an anomaly detection system for a mobile operator's packet core in just three days. Within two weeks it became a quasi-automated NOC: pinpointing root causes of "subscriber X can't get service Y in country Z", drafting emails to the partner's NOC, and tracking tangential attacks like credit-card and email scanning.
His key claim: for this kind of AI system the code doesn't merely 'not matter' — it's irrelevant. Next deployment should discard the code entirely and re-instantiate from scratch with a better model, a more suitable language (Rust or Erlang), and appropriate DB engines for general and time-series workloads. The real product is a cookbook: general approaches, standards references, rules and patterns. A pointed challenge to the notion that code is the product in the AI era.
More from AGI Musings
- Dev reframes LLMs as "universal translators" translating between representations — brandon_xyzw · 2026-11-09
- The bitter lesson isn't 'naive wins' — it's whatever scales best with compute — pzakin · 2026-11-09
- Owain Evans: now is the right time to switch into AI safety work — OwainEvans_UK · 2026-11-09
- Henry Farrell: the "machine gods" metaphor shrinks AI policy debate and should be resisted — sebkrier · 2026-11-09
- Thomas Dietterich: map what LLMs can assess reliably, design UX for the rest — JessicaHullman · 2026-11-09
- Bitter lesson redux: methods that generalize beat compute brute force, Sutton weighs in — ryanallred · 2026-11-09