25notes from production work
I build software end to end. I'm Scott Mallinson, a software engineer at Travelport in Barcelona, Spain. I write about API design, type systems, guardrails for AI features, and the maintenance that keeps a codebase navigable.
Chief's daily brief runs on a small local model. Getting it to behave meant moving a bullet-count limit out of the prompt and into code, then measuring what a repeat penalty setting actually cost in accuracy.
Two defects were cancelling each other out. Fixing the real one on its own would have woken a dormant guard with the wrong logic and made the page worse.
The tests for a desktop app's auto-scrolling passed the whole time the scrolling was broken: jsdom has no layout engine, so every height it reports is zero.
Validating user input is the most familiar guardrail and the smallest part of the problem. The model's own output, what it hands to tools, and what gets rendered afterwards all need the same fail-closed treatment.
For an AI feature with guardrails, releasing it carefully requires two disciplines: coordinating multi-service promotions with evidence at each step, and keeping the validation environment faithful enough that human feedback can be trusted.
When you add an AI layer to a retrieval feature, data decisions that developers used to make by instinct become explicit contract requirements. Live demos are often the first place you find out which ones are missing.