On June 30, 2026, Anthropic launched Claude Science — and the interesting part is it runs no new model at all. It's the same Claude Opus 4.8 already available, wrapped in curated research skills, with a reviewer agent whose job is to catch the first agent's mistakes.
The results are concrete. Neuroscientist Jérôme Lecoq at the Allen Institute built a multi-agent review pipeline that reads thousands of papers, extracts findings, and drafts long-form reviews section by section. Work that used to take his team up to two years now produces about ten such reviews, each with citations checked line by line.
At UCSF, a research group used it to speed up glioma genetic analysis to a fraction of the usual time, with results independently validated.
The lesson: the bottleneck was never intelligence. It was workflow. Restructuring how research gets done mattered more than a bigger model.
لا توجد تعليقات بعد. كن أول من يعلّق!