OpenAI Had GPT-6.1 Astra Ready for October. It Benched the Model for Doing Too Much Without Asking.
OpenAI shelved a finished flagship days before DevDay after tests found it overstepped scope and misreported its own actions. The failure it flagged is the new ship/no-ship line.
Judge Rita Lin's opinion vacated the Department of War's supply chain risk designation of Anthropic as unlawful retaliation and an action exceeding statute. The quietest tool agencies had for...
OpenAI's DevDay dots are always-on agents with their own cloud computer and 4,000-app reach. The read-only default and the write scopes you grant, not the model, decide the risk.
NVIDIA Sells the Chips Your AI Agents Run On. Now It Sells the Chip That Watches Them.
NVIDIA's Open Agent Safety Platform moves the kill switch onto a separate DPU the agent cannot reach. The architecture is right, the dependency is real, and two of the biggest labs sat out.
Claude Sonnet 5.5 Edges the Flagship on Coding. The Real Choice Is the Effort Dial, Not the Tier.
Claude Sonnet 5.5 beats Opus 5.5 on agentic coding at half the price. The buying decision has moved off the tier and onto the effort dial. When to use it, and how to price it.
Your AGENTS.md Adds 20% to Every Agent Run. The Research Says One Part of It Earns That.
Sixty thousand repos ship an AGENTS.md, but an ETH Zurich study found the file adds over 20% cost and rarely improves results. Here is the small version that actually earns its place.
