Anthropic describes Claude-assisted work across six drug-development partnerships and its own programs, while emphasizing that the research remains early and mostly preclinical.
Why it matters: For agent builders, the examples show Claude being used alongside scientists, lab systems and operational workflows, with humans still involved in testing and review. Anthropic presents these projects as early case studies, not evidence that the candidate medicines will succeed in people.
Simon Willison says October 2026 marks an inflection point: AI models can now produce games he finds genuinely fun with little creative input.
Why it matters: For agent builders, the reported shift is from generating game-shaped code to supporting iteration on mechanics and playability, areas Willison says previously demanded substantial human craft. His examples also show that playtesting and prompting remain important, and that his model comparisons are personal impressions rather than an exhaustive evaluation.