P-Zero Research's new TasteVal benchmark measures research taste: how well a model designs experiments and reads their results on eight real AI-development tasks, scored by how much compute it needs compared with human experts (the dashed 1x line). Progress was slow from GPT-4 in 2023 through 2025, then the slope steepened: since December 2025, scores have doubled every 3.0 months. Claude Opus 5.5 now sits at 2.30x the expert baseline, ahead of Fable 5.1 and GPT-6 Astra, at roughly 1/30 of the experts' cost per attempt. Research taste is the skill that decides how quickly AI could start improving AI, and this curve is one reason Jack Clark now expects that in 2027 (wrote about that yesterday as well).

P.T. finally has a proper native PC port thanks to developer @smokisevdalisi, and even Kojima approved it The fan-made project has reached version 1.0.1 and runs the original PS4 game natively through a custom C++/Vulkan renderer, no PS4 emulator required. >DLSS 4.5, FSR 3.1/4.1 and XeSS >DLSS/FSR Frame Generation >Ray-traced shadows, AO and reflections >Ultra graphics settings >Higher-resolution textures >Photo Mode and Free Camera >Mods >Experimental third-person mode >Experimental VR support The developer tested it on an RTX 4060 laptop and was getting 100+ FPS at 1080p Ultra You still need your own copy of the original P.T. files, the port doesn’t include Konami’s game data.

P-Video-Edit and P-Video-Edit Draft are in the same performance band on Design Arena. Draft scores 1170 Elo vs 1163 for Standard, while cutting average generation time from 64.1s to 28.5s. That puts P-Video-Edit Draft on the Pareto frontier for preference × speed, as well as price × speed. Benchmark by @DesignArena. 👉 Free Playground: 📚 Documentation: 🧩 API: 🥇 P-Bench: