Daily News · 1 min read

Anthropic AI Updates: July 21, 2026

1. Anthropic Opens AI for Science Grants for Rare Disease Research

Anthropic. Anthropic launched an AI for Science grants program offering rare disease researchers up to $50,000 in Claude credits over six months, plus access to Claude Opus and Claude Science and possible exemptions from biosafety classifiers for relevant projects. Two tracks are open: basic research scientists partnering with clinicians and patient organizations, and early-stage biotechs accelerating therapeutics. Anthropic positions Claude for synthesizing scattered clinical literature, detecting disease patterns, and automating regulatory paperwork, with applications closing August 2, 2026. Source

2. Rakuten Runs Overnight Autonomous Agents With Claude Fable 5

Anthropic. In a new Working at the frontier post, Rakuten’s Yusuke Kaji describes deploying Claude Fable 5 agents that run unattended for extended periods, completing tasks roughly 10x faster than manual processes across product, sales, marketing, and finance. Kaji credits the model’s self-reflection and self-verification behaviors, including self-correction when conditions change and first-principles re-validation against original goals, for enabling agents to work overnight without human course-correction. It reads as an enterprise signal that the unit of delegated work is shifting from granular task execution to whole-task decision-making. Source