
OpenAI Codex Now Reads and Writes Across Multiple Project Folders at Once
Codex local projects can now span multiple folders, letting the agent read and write across them while keeping one Git root in charge.
20 stories

Codex local projects can now span multiple folders, letting the agent read and write across them while keeping one Git root in charge.

ChatGPT Voice lands on macOS and Windows desktops, letting you control your computer and direct multiple AI agents using only your voice.

Claude's voice mode now runs on Opus and Sonnet, can call your connected tools mid-conversation, and speaks 18+ languages on every plan

AMD and Cerebras pair Helios rack-scale GPUs with the Wafer-Scale Engine to deliver the fastest production inference on trillion-parameter models at 5x greater efficiency

Runway's new Media Router lets developers set quality, cost, or latency preferences once and automatically routes each request to the best video, image, or audio model.

ChatGPT Health graduates from beta to full U.S. rollout, letting users connect Apple Health and medical records directly into their AI conversations — powered by the new GPT-5.6 Sol model

Google's Agent2Agent protocol hits v1.0 under Linux Foundation governance, with signed Agent Cards, multi-tenancy, and cross-framework interoperability now stable.
New MIT CSAIL work shows generalization can live in the harness: train short, generalize 8-32x longer. Here's the test to run on your agent first.

Black Forest Labs launches FLUX 3, a single model that generates video, audio, images, and drives real robots on Audi's production floor

Serena gives AI coding agents IDE-level semantic code understanding via LSP and MCP, slashing token waste and refactoring errors on large codebases

AT&T and GSMA release OTel 2.0, a 31B Gemma-4 fine-tune trained on 440B telecom tokens, topping the Open Telco AI leaderboard.

Lightricks drops a 22B open-weights audio-video model that renders a 10-second 720p clip in under seven seconds and keeps character identity across multiple shots.

Prime Intellect unifies 365,000+ validated agentic RL tasks across SWE, terminal, and search behind a single API and image registry

Google's SymptomAI outperformed independent clinicians on differential diagnosis in a 13,917-person national study, using Gemini Flash 2.0 to conduct real-world symptom interviews end-to-end.

Browser Use's new /game-mode lets an AI agent build, test, and iterate on playable web games — 2D, 3D, and multiplayer — powered by Kimi K3 or Claude Fable 5.

Google's Willow processor now uses reinforcement learning to self-calibrate mid-computation, achieving a 3.5x stability improvement and record-low logical error rates published in Nature.

Exa launches semantic search over 350M scientific papers, outperforming Google Scholar with 86.4% recall at sub-400ms latency

Cursor Router automatically picks the cheapest model that can handle each coding task, delivering frontier quality at up to 60% lower cost for Teams and Enterprise.

Claude Security's terminal plugin for Claude Code lets you scan staged changes or your full codebase for vulnerabilities without leaving your editor

Manus now generates a structured, editable plan before executing any task, letting you catch misalignments before a single line of code is written.