Hacker News AIT2·Agents
72 pts
The Claude Code team reveals that roughly 90% of its code is written and reviewed by Claude, with verification loops that produce evidence as the key enabler. Internal testing showed a separate grader checking output against a rubric improved structured file generation success by up to 10 points, with the biggest gains on the hardest tasks. Cheaper verification lets engineers run dozens or hundreds of Claudes in parallel and shifts their role toward defining done and reviewing evidence.