Agent work is being judged by study tests, boundary checks, and local receipts
The day’s clearest signal is accountability around agents. Machine Studying asks whether agents can learn a new corpus before an exam.
Topic summary
Open Weights
Coding Agent Operating Controls
Agent teams now have concrete checks for three recurring failure points: cheaper coding agents crossing module boundaries, coding sessions creating unclear spend, and agents entering unfamiliar corpora without proven…