Essays and write-ups · RSS
Writing
- 12 min readFun with Core AI, 9 years laterApple’s Core AI against Google’s LiteRT, ML Drift and ADK: what each one is, where each one hurts, and which one I’d pick.
- 6 min readAn agent for Android performance regressionsperfettoagent diagnoses Android performance regressions from two Perfetto traces and a commit range, under one rule: no citation, no claim.
- 11 min readRate-distortion curves for LLM agentsMeasuring LLM agents on quality, cost and latency at once, and why most of the findings are about the measurements rather than the models.