AI

Anthropic Says Claude Beat Best Listed Solutions on 72 MIPLIB Problems
The company also reported 45 optimality proofs and says it has submitted its solutions to MIPLIB, whose public records have not yet confirmed the tally.

Markey Demands AI-Free Google Search Option for Schools
Google says schools will be able to enable Search and Gemini separately, but AI Overviews remain built into Search.

Shield AI Says Hivemind Flew Multiple Air Force Drones as a Team
An earlier test let an F-22 pilot direct a single Fury, while the broader CCA program remains in operational experimentation.

Anthropic Launches Fable 5.2 and Mythos 5.2 at Lower Token Prices
The models share an underlying system but use different safeguards, with Mythos access restricted to verified organizations.

Meta Launches Muse Ember With Pricing From $0.10 Per Million Input Tokens
Available through Meta Model API and Muse Code, the lightweight multimodal model handles repeated agent tasks while larger Muse models plan the work.

Claude Code Approval Files Expose 1,379 Secrets on GitHub
Saved command approvals can retain passwords and tokens. Anthropic’s Claude Code team is exploring safeguards, including built-in secret scanning.

Google Adds Confidential and Long-Context Systems to TPU Lineup
TPU 8i Max combines two 8i chips in one package, while Google's product page lists all four eighth-generation systems as 'coming soon.'

Nvidia Board Opens Huang Investigation as CEO Admits Infidelity
Independent directors will oversee the review with outside counsel's help. The board backed Huang and said it had no reason to believe national security or company secrets were compromised.

Dropbox Plans $350 Million Tigris Deal to Expand Into AI Storage
DBX Agent Cloud starts with standalone S3-compatible object storage, with filesystem and content-processing capabilities planned next.