Google commits $10B to Anthropic

Bloomberg

Google and Amazon are each committing billions to Anthropic with performance-based structures, highlighting the intensifying competition among tech giants to secure relationships with leading AI labs and differentiate their cloud services.

Reuters

China forced Meta to reverse its $2 billion Manus acquisition over national security concerns regarding advanced AI assets, highlighting escalating geopolitical friction in the AI sector.

WIRED

David Silver's new startup Ineffable Intelligence has raised $1.1 billion to build superintelligence through reinforcement learning instead of LLMs, arguing that current AI approaches are fundamentally limited by their reliance on human-generated training data rather than independent learning and discovery.

arXiv

A new test-time scaling method for coding agents uses structured trajectory summaries and recursive voting to improve performance, boosting Claude-4.5-Opus by 7-12 percentage points on coding benchmarks. The approach reframes agent scaling as a problem of representing and reusing experience rather than simply generating more attempts.

Anthropic

Anthropic tested AI agents autonomously negotiating real transactions between employees, closing $4,000 in deals. More advanced models secured objectively better outcomes, but users with weaker agents didn't realize they were disadvantaged—highlighting potential for undetected inequality in AI-mediated commerce.