Articles
Short reads on one question each, answered from the same 24 prompts and every answer on this site.
- Does max effort help? What 144 pairs showSix Claude models answered the same 24 prompts at low and at max effort. Where the extra thinking raised the score, where it did not, and what it cost.
- Why some Claude answers run out of roomEvery answer has an output limit, and thinking counts toward it. Which answers reached it before they were finished, and what that did to their scores.
- Seven Claude models, one Minesweeper promptThe same request for a Minesweeper game in one HTML file, put to seven Claude models. The game each one built, and the checks it passed and missed.