Literally when I got to advanced and beyond just picking the longer and more complicated looking answer was the right one. I think this test is extremely flawed.
I’ve been building this since early May with active captains through the last month. I randomly noticed traffic from hacker news a few days ago, turns out somebody had linked tinywind in the comments there. Love to see other people in the space!
Have you seen Fable randomly jump from 50% session limit to 100%? That happened to me a couple hours ago. It was preceded by a bunch of errors about failing to submit a bunch of screenshots.
I haven't noticed that, but I did notice that on a single turn of maybe a few sentences, the cache hit was somehow roughly 500K. Either that's a bug, or there are some truly massive thinking blocks or Claude Code harness system injections behind the scenes.