Windows 11 makes Execution Containers generally available
Windows 11 makes Execution Containers generally available so organizations can bound agent file and network access. Intune controls and Claude Code support are still ahead.
Windows 11 makes Execution Containers generally available so organizations can bound agent file and network access. Intune controls and Claude Code support are still ahead.
Claude Code 2.1.294 fixes instruction-style prompt and agent hooks that allowed actions they were meant to block, and tightens Stop and SubagentStop judgments.
Max 5x includes $100 and Max 20x includes $200 in monthly Claude Platform credits. Team seat credits are pooled and capped at $500, separate from Claude and Claude Code limits.
Codex CLI 0.161.0 adds /mcp login and local voice-device selection. Daybreak stays opt-in on the CLI.
Haiku 5.5 is Anthropic's cheaper small model, about 75% less to run than Haiku 4.5, and Claude Code 2.1.293 makes it the default Haiku.
Claude Code 2.1.293 adds subagent type reporting and deferred-tool control, fixes an HTTP MCP memory leak, and reverts a cloud wakeup change.
Claude’s Workspace beta adds a Docs, Sheets, and Slides sidebar that reads the open file and can edit it with approval.
Anthropic’s expanded Cyber Verification Program gives vetted teams three tiers of access to Opus 5.5, Sonnet 5.5, and Mythos 5.1.
Claude Code 2.1.292 adds Agent-tool effort, marketplace install, and fixes for UNC reads, sandbox copies, and managed-policy cache tampering.
Atlassian and OpenAI expand their deal so GPT-6 models power Rovo, while ChatGPT and Codex can use Teamwork Graph context.
OpenAI trains computer-use agents with Ironclad on 11 contracting tasks. Astra beats GPT-5.6 Sol on a research rubric; reported times are simulated.
Dell announces a semantic layer, knowledge graph, and Knowledge Agents for AI, but those context features are scheduled for 1H 2027.
OpenAI will watermark eligible ChatGPT and Codex text in the EU, while API watermarking stays opt-in and detector access stays limited.
A 28-day ship-or-reset pledge for Codex and Work users opened with a faster default for GPT-6 models and a free Auto-review path for ChatGPT sign-in.
Reflection’s Beam is a 501B open-weight MoE with 23B active parameters. Weights are promised later this month; scores are company-reported.
Cohere’s North 2 adds cross-session memory and spend caps. The platform is model-agnostic and can run in cloud, on-prem, or air-gapped deployments.
Claude Code 2.1.290 adds hook fields for API-run tool calls, a subagent permission check, and the org approval ceiling, plus session attach by partial name.
Oct 3 grok.com notes put Grok 4.7 on Expert and paid Fast, and open Grok Bot threads to SuperGrok Plus as well as Heavy.
27 Sep–3 Oct 2026
Uber's MCP Gateway hosts 800-plus servers and over 5,000 tools, registers them disabled by default, and lets agents discover tools without loading every schema.
Claude Code 2.1.289 holds nested deny rules over user mods, closes a symlink Read gap, and reverts a VS Code auth change tied to extra sign-outs.
T3 Code Usage now splits estimated API-equivalent cost by token type and speed across Codex, Claude Code, and other local agent histories.
When Cursor Rollouts catches a regression it opens an issue, and Fix issue starts a cloud agent. It still does not merge or roll back on its own.
T3 Code’s Orchestrator V2 nightly adds Pi and OpenCode 2, a thread MCP, and delegate_task so one agent can run children on other harnesses.
SEMIBASE reports Lean certificates for all 15,973 semigroups of order 6 — a finite basis, or for four a proof that none exists — accepted only after a kernel audit, not model self-approval.