Ahmed E. Hassan

A clever routing strategy lets a tiny 3B code model outperform a massive 480B model on routine code completion tasks, slashing accelerator usage by 58%.

AI coding agents are surprisingly bad at logging, requiring humans to silently fix 72.5% of their logging mistakes.

Turns out, almost all AI agent tool descriptions are "smelly," and while fixing them improves performance, it also introduces a tricky efficiency trade-off that can be solved by carefully choosing which components to include.

Discover how AI coding agents are *actually* being used in real-world software projects with a new dataset of nearly one million agent-authored pull requests.