Apertis blog
Guides, updates and field notes from the team building Apertis.
- Zero alerts did not prove our leaked-key detector was runningSecurityPer-key spend detection, automatic throttling, a one-step revoke and GitHub secret-scanning intake, and why a silent detector had to prove it was alive.
- 900 KB of one repeated letter took almost six minutes to countPerformancetiktoken-go is quadratic on one long piece of text: how we measured it, two fixes that did not hold, the 512-byte bounded count, and what it does to billed tokens.
- Our gateway spent its latency budget before calling the modelPerformanceSerial database and Redis round trips on our relay hot path, three changes that removed many of them, and what the measurements do and do not show.
- We moved CI three times in two monthsCIOur CI moved from GitHub-hosted runners to a Mac mini, a GCE fleet and per-job microVMs in two months: the walls, the measured costs and what we got wrong.10 min