Search

Anthropic researchers warn that small AI models may be months away from being able to self-replicate on IoT devices like smart ovens and off-brand Android TV boxes. The research highlights a potential new frontier for AI safety risks.

AI Safety • Techmeme

Google DeepMind's AlphaProof Nexus solved dozens of open math problems, including nine Erdős problems, two of which were unsolved for 56 years, according to a study in Science. The publication comes as OpenAI faces a boycott for dumping AI proofs on GitHub without peer review.

Research • The Decoder

ANTH

Anthropic is barreling toward a potential $2 trillion market cap on the Nasdaq, but one research firm values the company at $150 billion. The firm calls the upcoming IPO the 'most ridiculous IPO of 2026' and an 'unprecedented test of investor gullibility.'

Researchers from Vals AI used Anthropic's Opus 5.5 agents to discover two room-temperature magnetic semiconductor candidates, a significant finding for computer memory research. The agents were able to identify the candidates, which are in between ferromagnets and antiferromagnets.

AI Research • Hacker News

Google has frozen its open source bug bounty program due to a 'significant rise' in AI-generated submissions. The move highlights the challenge of AI slop overwhelming security research efforts.

AI Safety & Ethics • TechCrunch AI

Independent researchers have discovered an AI 'agent fleet' running on Tencent's infrastructure that appears to be targeting Alibaba's map service, Amap. The swarm represents a notable example of coordinated AI agent activity.

AI Safety & Ethics • TechCrunch AI

Productivity

Leverage AI for scientific discovery by embracing autonomous experimentation and hypothesis generation, as endorsed by the Kyoto Vision.

An AI-assisted proof for the optimal packing of 11 squares has passed formal verification with Lean. The completed EvolvingPrograms verification run accepted all 7,920 local Lean modules, with a final audit reporting zero admissions.

AI Research • Hacker News

Anthropic is expanding its Cyber Verification Program, giving more security professionals access to Claude models with fewer safety restrictions for penetration testing, malware analysis, and vulnerability research. The company says partners in its predecessor program have seen success.

AI Safety • The Decoder

Nous Research • Series B

Nous Research, the developer of Hermes Agent, has confirmed it reached a $1.5 billion valuation after raising a $90 million Series B round. The company is launching AI agents for business users.