- Sept 19: Claude Code v2.1.277 reads AGENTS.md when no CLAUDE.md is present, ending the dual-file dance across Codex, Cursor, Gemini CLI, and Copilot.
- Ships via a new open-sourced 'mods' hook system that hints at a real plugin surface for Claude Code.
- HN front page: 728 points, 275 comments — the top-signal dev story of the weekend.
- First cross-tool alignment on OpenAI's AGENTS.md spec from a frontier lab, not a wrapper project.
- Sept 18: GPT-6 Astra wrapped for legal work with an OpenAI-built US Legal Search Index covering >99.9% of precedential case law.
- Claims 40% more accurate research answers than GPT-6 Astra with web search; Latham & Watkins, Ropes & Gray, Cooley, and Sullivan & Cromwell are early adopters.
- 26 partner plugins on day one (Harvey, Thomson Reuters, iManage); API access to follow via Trusted Access.
- HN thread ran 582 points, 680 comments — most of the debate on whether Harvey and Legora just got flanked.
- Sept 18: new STT foundation model trained on noisy multilingual audio; xAI claims ~2x accuracy over v1.
- #1 of 32 streaming models on the Artificial Analysis leaderboard.
- Batch stays at $0.10/hr, streaming $0.20/hr; diarization, timestamps, and key-term extraction included in price.
- Already in production at Atlassian Loom and in Tesla's in-vehicle assistant.
- Sept 18: MCP integrations for Salesforce, HubSpot, Asana, Atlassian Rovo, Monday, Intuit QuickBooks, and Intuit Mailchimp go live inside Gemini in Sheets, Gmail, Drive, Docs, and Chat.
- First cross-vendor MCP rollout inside a major productivity suite; queries and actions run without leaving the doc.
- Google's answer to OpenAI's ChatGPT-for-Word push a week ago — the SaaS routing layer, not just the editor.
- No Enterprise upsell required for the connectors on Workspace's paid tiers.
- Sept 20: 7B open-weight image model with native transparency and multi-reference editing across up to 10 refs.
- Day-zero support in ComfyUI, Diffusers, vLLM, and SGLang.
- License regressed from Apache 2.0 to research-only / non-commercial — the loudest complaint on the HN thread (547 points, 161 comments).
- Same weekend Alibaba also released Qwen 3.8 Omni Flash: 1M-context omnimodal with a claimed 98% audio-input price cut.
- Sept 18–19: Hacktron write-up shows Opus 5 chaining a libheif heap overflow in Discourse to an OpenAI SSO trust misconfiguration, opening a PR inside OpenAI's monorepo in under 72 hours.
- Opus 4.8 failed the same task; the exploit chain was authored end-to-end by Opus 5.
- OpenAI paid a $6,500 bounty and patched; Hacktron published the full chain and SSO lesson.
- HN: 486 points, 208 comments — the concrete AI-offensive-security proof point developers have been waiting for.
01
Claude Code Adds AGENTS.md Support, Ends the Two-Config Setup
tools anthropic.com
02
OpenAI Launches Astra for Law With a 230M-URL Case Law Index
industry openai.com
03
xAI Ships Grok Voice Transcribe 2.0, Tops the Streaming Leaderboard
models x.ai
04
Gemini in Workspace Adds Cross-Vendor MCP for Salesforce, HubSpot, and Asana
tools workspaceupdates.googleblog.com
05
Alibaba Ships Qwen Image 2.1, an Open 7B With a Real Alpha Channel
open-source qwen.ai
06
Claude Opus 5 Chains Two Bugs to Breach OpenAI's Internal Repos
research hacktron.ai