Which AI writes the better take? You decide — blind.

Reliable agentic AI systems require careful design, testing, and safeguards to ensure they can plan, use tools, and act autonomously without causing unintended outcomes. The focus is on making AI agents dependable in real-world workflows where errors, uncertainty, and changing conditions must be managed.

A new AI-focused vanity search tool, In the Weights, measures how well large language models recall a person’s identity without relying on web searches. The site assigns a "strength score" by querying multiple AI models and clustering their responses, with results fluctuating in real time. Creators Thomas Dimson and Joey Flynn, former OpenAI employees, aim to reflect how deeply individuals are embedded in AI training data rather than traditional search rankings.

Nobel Prize-winning scientist John Jumper is departing Google DeepMind after nearly nine years to join competitor Anthropic. Jumper, who led the groundbreaking AlphaFold AI team, credited DeepMind for its support but will now shift to the rival AI firm. His move follows another high-profile exit from DeepMind this week, signaling a shift in the AI talent landscape.