Writing6 notes

Working notes.

Agents, serving engines, and the occasional side quest. Mostly written up when something surprised me, or when a result I believed turned out to be measuring the wrong thing.

August 5, 20265 min read

A Loop Is Only as Good as Its Sensor

An agent that grades its own work is a control loop whose sensor sits inside the plant. Every time I fixed my evaluation, it started rewarding a different degenerate agent.

Agent EvaluationControl SystemsMeasurement
July 25, 202611 min read

The 31% Page-Size Slowdown That Wasn't

An SGLang result survived every rerun and vanished only when I timed the right interval. Nsight and source tracing revealed when KV-cache pages actually matter.

LLM SystemsSGLangGPU Profiling
June 24, 20265 min read

The Honest Machine

A generated paper can look finished and be wrong in the two places nobody checks. The useful engineering is the part that fails the build.

AI for ScienceAutonomous ResearchResearch Integrity
March 15, 20264 min read

The Detector Was the Easy Half

A Chrome extension that reads your pulse through the webcam and decides when to interrupt you. The signal took a month. The interruption is a different kind of problem, and I turned it off within a day.

rPPGChrome ExtensionReactSignal Processing
March 1, 20265 min read

The Page Moves While You Think

A recommender scores your interest by how long you watch. A browser agent's watch time is however long the model takes to decide. So the agent expresses preferences by thinking.

Web AgentsBenchmarksTemporal ReasoningResearch
February 20, 20263 min read

The Ghost Stories Had the Best Data

Song-dynasty supernatural tales were filed for centuries as minor entertainment. They hold the richest geographic and social detail in the corpus, and the reason is structural.

NLPDigital HumanitiesClassical ChineseFlask