We use cookies to measure how the site is used and how our ads perform (Google Analytics and Ads, Microsoft Clarity, and the Meta, OpenAI and X ad pixels). “Essentials only” turns off advertising cookies everywhere, and analytics too in the EEA, UK and Switzerland. Privacy Policy
Tool Retrieval is in preview in the Gemini Interactions API: add a tool_search tool, set defer_loading on the long tail, and the server loads parameter schemas on demand. What that changes for tokens, prefix caching, MCP catalogs and debugging — and where the saving does not show up.
Job listings spotted by Bloomberg show Apple building an autonomous agent that browses, codes, controls apps and coordinates sub-agents — the last major platform holder to enter the computer-use race.
NVIDIA integrates GPT-6 Astra and other frontier AI agents with its Omniverse and Isaac Sim platforms, allowing engineers to automate physical AI and robotics simulations. Coupled with the Halos safety architecture, the toolchain accelerates real-world deployment.
Ollama v0.40.2 upgrades models pulled on earlier versions to llama.cpp in the background the first time they run, and keeps the originals as backups. What that means for disk budgets, safe downgrades, canarying local fleets and the documented jq cleanup loop.
Databricks shows how Lakebase can give every coding agent and pull request an isolated Postgres branch, changing how teams delegate schema changes, tests, previews and data-heavy fixes.
Amazon Bedrock AgentCore payments gives agents managed wallets, x402 signing and spending ceilings enforced by AWS infrastructure. Incarna cut a two-to-three-month build to 200 lines and three days, and LangChain's Restock shows the build-it-yourself path with Stripe Link.
Anthropic's Claude Code 2.1.295 makes broken guard hooks block the action, adds OSC 7501 terminal status, pins models per gateway upstream, and fixes MCP reconnect loops. Here is what to change this week.
Mellum2.1 keeps its 12B mixture-of-experts architecture while expanding reinforcement learning for repository work. JetBrains reports coding gains, but deployment artifacts and speed claims need separate checks.
Google is rolling its SynthID Detector out globally so anyone can check images, video and audio for AI watermarks from Google, OpenAI, Nvidia and Kakao — with real limits on what it can and can't catch.
Automater Lite 0.1.29, published 2026-10-08 in the official releases feed, makes Automater Core the only engine on Windows, macOS and Linux, opens conversations straight in Codex, Grok Bot and GitHub Copilot, states that Grok Bot conversations open without flicker, and asks for permissions only when you choose Allow.
Google Cloud's MCP Toolbox for Databases Java SDK reached v1.0 on October 6, 2026, hardening the beta that followed the earlier MCP Toolbox server v1.0 milestone. Here's what changed and how to wire it into a Spring Boot or Quarkus agent today.
The Qwen Code 0.25.1-preview.0 pre-release ships an experimental Kubernetes CSI runtime with durable worker ACK, M5b durable local Runtime tool outcomes, Stage H5 channel contracts, and the W2 session-directory change; Stage H4/H5/H6 slice designs and the cross-node EventTransport are documentation only.
Claude Code v2.1.293 ships Claude Haiku 5.5 as the default Haiku model with 1M context and new pricing tiers, fixes a bug where Claude re-did finished work after context compaction, patches an HTTP MCP memory leak, and adds agentType to subagent status lines.
GitHub resolved two linked degradations on Oct. 7 affecting Git Operations, Pull Requests, Issues, Actions and Webhooks. Here is the verified timeline, scope, and what CI and agent-fleet operators should do while root-cause analysis is pending.
Anthropic's status feed reported usage-data errors on the platform.claude.com Usage and Rate limits pages and the Admin API usage report twice inside one UTC day, with a mitigation rolling out and roughly 30 minutes of lag on recent usage. Here is what fleet operators should reconcile.
OpenAI's 09:28 UTC update on October 7, 2026 released an updated ChatGPT desktop app for Linux and macOS after build 26.1002.51308 blocked new ChatGPT Work threads. The Windows fix was still rolling out, with Codex in ChatGPT Desktop in partial outage.
Automater's manifest fetched as .2 at 03:55 UTC on October 7, 2026; a historical capture showed .6 at 05:38 UTC. The attached .6 screenshot is a separate 06:47:21.658 UTC recapture. The public News index offers topic filtering and RSS; its editorial method describes the review gate.
Automater Lite 0.1.26 shipped October 6, 2026 with Claude Code and Codex connections under Settings > AI tools, Codex cloud-session archiving, Mini dock modes and pop-out cards, and a free leaderboard. Corrected against the current releases feed, changelog and marketing build manifest.
Google DeepMind's EmbeddingGemma 2 maps text, code, images, audio, and video into one 740M-parameter Apache 2.0 embedding space that runs on a phone. What shipped, what is still promised, and why open weights matter for vector stores.
GitHub's initial report quantifies a four-hour Actions and hosted-runner outage: at peak, 47% of hosted-runner workflow runs failed and 71.9% of hosted jobs missed the five-minute start mark.
Anthropic ships Claude Code v2.1.292 with per-sub-agent effort budgets, a prompt.autocomplete hook for mods and plugins, --marketplace flag on plugin install, prompt caching for mods, workflow agent visibility in hooks, and critical security fixes for sandbox escapes and UNC path bypasses.