| 2026-09-06 | 1) Replaced Claude Opus 5 with Gemini 3.8 Flash, because I was unsatisfied with the quality of Opus' results. Maybe this is related to the fact that Opus scores lower than Gemini on the scientific benchmark GPQA Diamond (https://artificialanalysis.ai/evaluations/gpqa-diamond). Also, I tried Claude Fable, but it has an in-built protection against bioterrorism that works by disallowing it to answer biological questions. 2) Added "About" and "Change log" pages. |