Profiling web-edit with Python 3.15: flame graphs and source heatmaps
Profiling web-edit with Python 3.15: flame graphs and source heatmaps
Willison asked Codex to use Python 3.15’s new sampler on the unmodified web-edit app, producing a flame graph and source heatmap. In a busy editing workload, line-ending preservation appeared in 64.8% of 15,296 recorded CPU stacks. Sampling all threads mattered: most stacks were in a worker thread, which a main-thread-only profile would have missed. Willison then fed the report to Claude Opus 5.5, crediting it with performance changes in his October 13 web-edit 0.1a1 release.
The first agent produced an observation and the second used it as a repair input; neither short account gives before-and-after latency, an independent review or evidence of a real credential edit. A fraction of sampled CPU stacks identifies a hotspot, not the size of a speedup. This is follow-up engineering on a personal tool, not a claim about maintained customer value.