Articles
Risk scoring on pull requests: what agents can do about it
Risk scoring on code review focuses human attention. Agents can reduce how much is needed: three loops built on top of Unblocked's risk labels.
Measuring AI productivity: Why your numbers are flat and your rework is up
AI usage is up ~65% but PR throughput rose only 7.76%. Why productivity dashboards can't tell motion from delivery, and four metrics you can baseline next week with gh, jq, and git.
REPL is dead, long live REPL
Developers have always moved up abstraction layers. Agents are forcing us up another one, towards intent, and we've never had a probabilistic layer until now.