Two background jobs that marked themselves done too early
Two background jobs stopped early and recorded themselves as complete, and once recorded, nothing in the system could reopen them.
Scott Mallinson
26notes from production work
I build software end to end. I'm Scott Mallinson, a software engineer at Cognizant in Barcelona, Spain. I write about API design, type systems, guardrails for AI features, and the maintenance that keeps a codebase navigable.
- Role
- Software engineer
- At
- Cognizant
- Based
- Barcelona
- Notes
- 26 published
Contents Page 1 of 5
The model that wouldn't take instructions
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.
The guard that only worked because something else was broken
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.
A passing test that never saw the page
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.
Four boundaries an AI assistant's guardrails have to cover
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.
What a promotion actually asserts
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.