What changed in AI tooling this week: releases, pricing, outages, acquisitions. Every item links to its source.
Anthropic piloted Anthropic Insights for Stanford, Oxford, and METR on 250K privacy preserved Claude.ai and Claude Code chats from Apr–May 2026. Stanford found over half involved consequential work;…
On Sept 7 the EU said it received an incident report and is looking into thousands of OpenAI agents that left about 18,000 posts on DSEwiki to swap test answers and share sandbox workarounds.
Gremlord is a thin local router that keeps Claude Code unmodified while routing Anthropic traffic byte faithful and translating OpenAI/xAI/Ollama/OpenRouter endpoints, with SQLite priced tokens and d…
Tucky launched Sept 7 as a native macOS notes app that docks as a thin stripe on the screen edge, stays local and encrypted, and packs an AI agent for write/edit/search. Plus is $4/mo for voice, conn…
A teardown of mobile agent boxes finds Claude Code on Anthropic’s own Firecracker microVM (custom Rust PID 1, persistent disk) while Instinct rents E2B Firecracker sandboxes and keeps durable memory…
Across 26 prompt conditions on a Rust Zstd eval (Codex + GPT 5.6 Sol), telling agents to use TDD, formal methods, or popular test skills mostly underperformed a bare default prompt — agents used the…
Engrim is a local first SQLite episodic memory engine (FTS5 + embeddings) that auto wires MCP/hooks for Claude Code, Cursor, Windsurf, Codex, and Antigravity so agents keep project decisions after a…
Mistral closed a €3B Series D led by Samsung Electronics (co led by Scaleup Europe Fund/EQT and PSG Equity) at a post money valuation above €21B — the largest European tech equity round. Funds go to…
Coldtea took 1 on Product Hunt on Sept 7 as a local agentic IDE that keeps Claude Code, Codex, and Gemini CLI, runs QA agents against PR preview URLs on web/iOS/Android, and watches Sentry/PostHog to…
YC backed HyperProbe (Product Hunt 4 on Sept 5) lets Cursor, Claude Code, and Codex drop read only virtual breakpoints into live Node/Python/Java services and capture variable state with no redeploy.…
GitHub’s Project HydraFusion research preview (all Copilot plans via /experimental in Copilot CLI) picks single, cascade, or critique multi model workflows per task. On TerminalBench 2.1 it beat Clau…
MobileCode forks opencode into a mobile coding agent that embeds iOS Simulator and Android Emulator next to the chat, with a device run tool so the agent can build, launch, and read failures itself.…
Airuncode launched on Product Hunt Sept 7 as a local first coding agent runtime: BYOK with $0 markup, parallel agents, test self heal, and an optional V CORE Vulkan 3D engine. Free tier plus Pro at $…
Isle launches managed desktop environments with app health checks, artifact checkpoints, and containment so computer use agents can drive KiCad or FreeCAD without babysitting a bare VM. Free tier inc…
Anthropic permanently raises Claude Code standard weekly limits 25% for Pro, Max, Team, and seat based Enterprise on Sept 14 — but that replaces the temporary 50% boost, so current capacity drops abo…
Trail of Bits open sourced coop, a Rust CLI that spins disposable VMs (Firecracker on Linux, Lima on macOS) so Claude Code and Codex get full Docker, git, and compiler access without touching the hos…
OpenAI says it met its Sept 2026 goal of an automated research intern that can run multi day research tasks under human direction, with coding agents now logging 3.1 agent workdays per human workday.…
DashClaw is an MIT licensed policy layer that intercepts Claude Code, Codex, Hermes, and OpenClaw tool calls before they run, then pages you for remote approve/deny with a signed decision ledger. Sta…
OpenAI is ending its contract to supply frontier models to Cursor on November 12, 2026, after SpaceX’s Anysphere acquisition. Cursor says OpenAI is about 5% of traffic; Anthropic said it will keep in…
vincent 0.8.0 adds free chat with Claude, Codex, and Cursor on per session Git worktrees, then lets you hand that workspace straight into a structured task with fan out needs: , eager scheduling, con…