https://ai.paski.dev/log/ What the benchmark numbers turn out to mean once you look at where they came from. en https://ai.paski.dev/log/the-catalog-problem/ https://ai.paski.dev/log/the-catalog-problem/ Sun, 26 Jul 2026 02:37:30 GMT There is a public catalog of Claude Code plugins. I pulled it expecting a few hundred entries and got this: https://ai.paski.dev/log/most-models-are-beaten-outright/ https://ai.paski.dev/log/most-models-are-beaten-outright/ Sun, 26 Jul 2026 02:36:40 GMT A ranking is a line. It can tell you that one model is better than another and nothing else — not what the difference costs, not whether it was worth paying for. Put cost on a sec… https://ai.paski.dev/log/spending-fewer-output-tokens/ https://ai.paski.dev/log/spending-fewer-output-tokens/ Sun, 26 Jul 2026 01:11:49 GMT The last post established that the answer is the bill: about 98% of a typical call's cost is what the model writes back, not the prompt you sent. This one is the practical follow-… https://ai.paski.dev/log/you-are-not-paying-for-your-prompt/ https://ai.paski.dev/log/you-are-not-paying-for-your-prompt/ Sun, 26 Jul 2026 01:09:43 GMT There is a new page on this site: paste a prompt, pick the model you use, and it tells you what the call costs and what else that money would have bought. Building it turned up th… https://ai.paski.dev/log/the-scaffold-is-half-the-result/ https://ai.paski.dev/log/the-scaffold-is-half-the-result/ Sun, 26 Jul 2026 01:09:09 GMT SWE-bench Verified has an official leaderboard, maintained by the people who built the benchmark. It runs every submission through mini-SWE-agent, their own reference harness: the…