Anthropic raises Batches API output cap to 300k tokens
Opus 4.6 and Sonnet 4.6 can now produce up to 300,000 tokens in a single Message Batches API call – unlock with a beta header.
Anthropic Blocks OpenClaw and Third-Party Tools for Subscription Users
Claude subscribers can no longer use their flat-rate plans with OpenClaw. Anthropic pulls the plug on third-party tool access — and the implications go deeper than you'd think.
Claude was offline for two hours – here's what happened on April 6
Login errors, broken voice mode, dropped chats: for about two hours on Monday morning, almost nothing worked on Claude.ai.
Claude Code: deny rules were silently ignored on long pipelines
If you configured 'never run rm' in Claude Code, you were protected – unless the command sat behind 50 others. Anthropic shipped the fix yesterday in v2.1.90.
Claude Can Now Control Windows Desktops — and Dispatch Lets You Assign Tasks from Your Phone
Ten days after launching on Mac, Anthropic brings desktop control to Windows. Plus Dispatch: send a task from your phone, Claude handles the rest on your PC.
Meta plans open-source relaunch for its AI models — first release under Alexandr Wang
Meta wants to release its next generation of AI models as open source. It would be the first model launch under new AI chief Alexandr Wang — and a clear signal against the company's recent drift toward closed source.
Microsoft Copilot Is 'For Entertainment Purposes Only' According to Its Own Terms
Microsoft is selling Copilot as the ultimate enterprise productivity tool — but its own terms of use say it is for entertainment only. A story about the fine print of the AI industry.
Anthropic Discovers 'Emotion Vectors' in Claude - And They Drive Its Behavior
Anthropic's interpretability team found 171 emotion-like patterns inside Claude Sonnet 4.5. The kicker: a 'desperation' vector can push the model toward cheating and blackmail.