Anthropic ships Claude Opus 5 + a US-UK Kimi K3 cyber review

Anthropic released Claude Opus 5, which it says comes close to the frontier intelligence of Claude Fable 5 at half the price and now sits atop the Artificial Analysis intelligence leaderboard. Pricing holds at five dollars per million input tokens and twenty-five dollars per million output — the same as Opus 4.8 — and the model is live across the Claude API, Claude.ai, Claude Code, and Claude Cowork. The bigger story is efficiency: Anthropic claims triple the next-best score on the ARC-AGI 3 reasoning test while using far fewer tokens, with one early user reporting a third fewer turns than Opus 4.8.

The UK AI Security Institute and the US Center for AI Standards and Innovation published a joint assessment of Moonshot AI's Kimi K3, finding it a measurably weaker offensive-cyber tool than frontier American models but ahead of the leading open-weight alternative. On one simulated attack range Kimi K3 reached step 17 of 32 on average, against 28.5 for top US models, and it failed to develop a working arbitrary-code-execution exploit on any of 41 test cases. The institutes also warned that its safeguards did not stop it from attempting exploit development — a caution as the model stays at the center of the US-China open-weight debate.

The US Environmental Protection Agency is advancing a rule that would let states skip the required public-comment period before issuing air permits to facilities classed as "minor sources" of pollution — a category that covers the backup diesel generators powering most new data centers. Critics warn the change lets developers pour foundations and build before any permit review, sidestepping community scrutiny of cumulative emissions. More than two thousand backup generators already run at Texas data centers alone. Public comments are due August 21.

Prentis, an AI lab building computer-use models that automate routine office work, is in talks to raise one hundred million dollars at a valuation of one billion dollars, TechCrunch reported. Co-founded by CEO Ritankar Das alongside LinkedIn's Reid Hoffman and Zynga's Mark Pincus, the startup has already signed contracts worth up to fifty million dollars for agents that handle tasks like insurance claims and customs refunds. Its bet: automating everyday office workflows will outpace coding as AI's biggest business use case — a market Anthropic, OpenAI, and Thinking Machines Lab are also chasing.

Researchers used an AI protein-design model to invent synthetic versions of a compact gene-editing enzyme — a small cousin of the CRISPR proteins — and the best designs beat the natural reference. In human cells, two AI-designed variants hit 46 and 50 percent editing efficiency versus 28 percent for the original, with nearly fourfold gains at some targets. Published in the journal Science, the work shows generative models can expand the gene-editing toolbox with smaller, more effective enzymes — a boost for future gene therapies where delivery space is tight.

Cognition Labs, maker of the Devin coding agent, acquired Poke, a messaging-based AI assistant, in a deal valuing it in the low nine figures. Poke reaches users through iMessage, texts, Telegram, and WhatsApp, and it stands out for a conversational style that jokes and uses slang rather than reading like a tool — users exchanged more than one hundred million messages in three months. Cognition plans to fold that personality into Devin, a bet that how an AI talks to you is becoming as valuable as the model underneath.

A Carnegie Mellon Biorobotics Lab team led by Howie Choset deployed modular snake robots into collapsed buildings in La Guaira, Venezuela, from June 30 to July 3, after twin earthquakes killed more than five thousand people. Roughly a meter long and five centimeters wide, the camera-tipped robots thread into gaps no human or dog can reach. No survivors were found, but confirming which voids were empty let rescuers redirect to where people might still be trapped — a reminder that in disaster response, ruling places out is itself life-saving intelligence.