Paper Reviews
Reading notes and commentary on research papers, studies, and other published work.
-
Apr 25, 2026
Reading the SlopCodeBench Paper
A paper called SlopCodeBench has been circulating as evidence that AI agents fundamentally lack design discipline. The headline findings are striking: no agent completes any of the 20 problems end-to-end,...
-
Apr 23, 2026
Reading the METR Productivity Study
The METR paper published in July 2025 has been making rounds, and the version of it that travels through comment sections and social media is fairly consistent: AI slows experienced...