On Slowness
The case for doing less, more deliberately — and why the pressure to move fast is often a collective hallucination.
Writing on software, systems, and how things actually work.
A lightweight tool for tracing and visualizing token-by-token generation across model runs. Useful for understanding where attention concentrates during reasoning chains.
Watches production data pipelines for structural changes and surfaces diffs in a clean terminal UI before they propagate downstream.