Opus 4.8 did "plan a month of content" 81× faster than a typical person: 22 s against 30 min, 93% accurate on the task's rubric. A recorded run, replayed. Measured by howfastai.com.
Opus 4.8 did "plan a month of content" 81× faster than a typical person: 22 s against 30 min, 93% accurate on the task's rubric. A recorded run, replayed. Measured by howfastai.com.