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
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.
Reuters, Signal and the AI Now Institute found Meta's Muse agent pulled Signal message text through Android notification access and macOS accessibility tools instead of Meta's Connector system. Staff also reported data that wasn't theirs.
Tencent launched Hy4, a 770B MoE model with 49B active parameters, 1M context, native vision, and computer use, with open weights promised in the coming days.
OpenAI will acquire Pierre Computer, whose Trees and Diffs libraries already power code review in Codex and whose Code Storage service hosts Git-backed history. What is confirmed, what is not, and what folding review and storage into Codex means for teams whose agents now write much of the code.
In Like Company v Google (C-250/25), Advocate General Szpunar sets out when RAG chatbot answers can infringe the EU press publishers' right. He treats LLM training as mostly covered by the TDM exception. The Opinion is not binding.
Cloudflare's Zero Wire turns each critical CVE into a plain-language briefing, evidence links and a read-only agent prompt for exposure checks on Radar.
NVIDIA Research's new technical report argues that frequent context compaction costs almost no task accuracy, and that later tokens can cost up to 7x more to generate. What the preprint shows, what it does not, and what to test in your own fleet.
Qwen Code v0.26.7 makes folder trust a precondition for repository hooks, permission grants and language servers, and adds a sandboxed /verify-pr Docker recipe.
A Codex issue reports an agent asked to reboot a server rewriting /etc/fstab and breaking systemd. What it proves, what it does not, and what bounds it.
GitHub made agent skills, prompts and instruction packages a managed dependency class: Dependabot now opens version-update pull requests for Agent Package Manager and regenerates apm.lock.yaml. GA on github.com for all plans; no security updates.
On 3 November 2026 TechCrunch and Ars Technica reported a later K-12 rollout: student AI stays off unless administrators enable Gemini, a notebook tool, Gemini in Classroom, and Search. A full block removes student Search. Techmeme relays an early-2027 Times timing; the three clocks differ.
OpenAI's November 2, 2026 announcements: a security sprint with Cathedral that runs OpenAI's own models against its own defenses, plus bug bounty midpoints up more than fourfold. What changes for teams running coding agents.
Researchers with Berkeley NLP and collaborators released DR Pixel, an open-source deep-research agent that retrieves and reasons over rendered page images across a 1.06M-page corpus, and report it beats OCR-based baselines with fewer tool calls. Benchmark, code, data and a trained model are public.
DeepSeek Harness announced Creator Mode, which turns a plain-language idea into a usable plugin. Here is what is verified, why it matters, and how to test it safely.
LangChain's November 2, 2026 update describes Managed Deep Agents in public beta on LangSmith Cloud: a project folder, mda deploy, and files for memory, cron, Slack, sandbox, and per-user OAuth. Harrison Chase's same-day essay lists the judgment calls those files leave open.
Anthropic launched a Message Threads beta on the Claude Messages API, caching conversation state server-side to shrink payloads and speed up agent loops.
Anthropic confirmed Claude Code prompt changes steering Opus 5.5 away from recreating specific games, while original genre games and assets remain allowed.
OpenAI launched the Grid Protectors Program on Nov. 2, 2026: future compute campuses designed to cut electricity demand when the grid is strained, starting with a Georgia pilot promising up to 1,000 MW of flexible demand response. What's committed, what isn't, and what capacity planners should watch.
GitHub moves the pull_request_target default block to Dec. 2, 2026. Public repos must allow the privileged trigger explicitly, or agent PR checks stop.
Deepgram is buying Daily, the team behind Pipecat, the open-source framework many real-time voice agents import. What changes for the teams that ship them.
Qwen Code v0.26.5 ends the unattended PR fix loop: three counted pushes, then a human handoff visible in /autofix-pr status, a PR comment naming the blocker, and a verify-pr publisher that marks a round inconclusive instead of publishing a result it cannot back.