Meta hotfixed its Muse macOS app after researcher Patrick Wardle showed a local zero-day that redirected cloud transcription to an attacker endpoint and let Muse take photos and write files without alerts. Meta says practical risk was low because the exploit needed malware already running under the user’s account.
Databricks made its Genie One MCP server generally available, exposing governed business data and Genie Ontology to any MCP agent through Unity Gateway. Teams can pull the same trusted answers and citations from Claude, Cursor, ChatGPT, or custom agents without per tool connectors…
OpenAI expanded GPT 6 with Sol (coding/complex work) and Luna (high volume clerical tasks), cutting API cost 50% vs the 5.6 Sol/Luna line via caching and inference gains. Sol is live in ChatGPT Work, Codex, and the API; Luna also rolls to Free and Go, with broader ChatGPT rollout through the day…
Anthropic released Claude Opus 5.5 on Tuesday with Fable level coding and knowledge work performance at $4/$20 per million tokens (20% below Opus 5). Sonnet 5.5 and Haiku 5.5 are due in the coming weeks; the model carries Fable class cybersecurity and biology safeguards…
Google released AX, a declarative orchestrator (Task, Workspace, Gateway, Model) for agent fleets on Kubernetes via Agent Substrate. Its HN launch hit 626 points as builders argued the “easier agents” pitch still needs a cluster, ko, and a registry…
AWS released Strands Harness (Apache 2.0) — a ready made general purpose agent for Python/TypeScript that runs locally or in any Linux cloud, with shell/file/web tools, context compaction, memory, and subagents. On matching Claude/GPT models it claims 28% lower token cost vs other harnesses across six benchmarks, and with Fable 5 it scored higher than Claude Code on Terminal Bench 2.1 at 77% lower cost…
Spotted something we missed? Start a thread.