Notes on building, learning, and thinking.
An analysis of information aggregation failures in thin-market conditions, and what calibration research tells us about fixing them. The gap between theoretical efficiency and observed accuracy reveals structural problems worth examining.
Read more →When inference cost matters more than benchmark scores, quantized 7B models outperform their larger siblings in production.
Lessons from designing xity-sdk: schema decisions, rate limiting patterns, and why GraphQL was the wrong choice initially.
Philip Tetlock's framework for calibrated prediction and what it means for everyday decision-making under uncertainty.
Partial indexes, materialized views, and BRIN indexes: the three tools that solve 90% of my query performance problems.
A practical walkthrough of Raft consensus for application developers who need coordination but not a PhD in distributed systems.
The best technical writing starts with empathy: what does the reader already know, and what single idea will unlock the rest?
Donella Meadows reveals how feedback loops, stocks, and flows govern everything from ecosystems to organizations. Essential mental models.