The Model Context Protocol underlying Claude Code, Copilot, Cursor and Windsurf finalizes a sweeping stateless rewrite this week, tearing out the session handshake that has anchored it since launch.
GitHub's autonomous background coding agent can now be assigned straight from a Linear issue and works independently in its own cloud environment.
A failed GitHub Actions check can now be diagnosed and patched from a phone with a single tap, GitHub said July 23.
Cognition acquired texting AI agent Poke to bring a more casual, human interaction style into its Devin coding agent.
Cognition's cloud coding agent can now execute directly on a company's own Mac minis, GPU servers, or Kubernetes clusters.
Anthropic let Claude's voice mode run on Opus and Sonnet instead of only Haiku, and wired it into Canva and Notion.
Runway launched a Media Router that automatically selects the best video, image, or audio model for a job, including rivals' models.
Alibaba's research lab pushed its live, full-duplex video model to higher resolution without slowing its response time.
Netflix disclosed it paid $587 million cash for Ben Affleck's AI filmmaking startup and used generative AI on about 300 shows and films this year.
AI video platform Higgsfield is raising up to $500 million as its annualized revenue quadruples toward half a billion dollars.
Chinese video-model maker ShengShu unveiled Vidu S1, a foundation model that generates and reacts to video in real time from continuous voice input.
Synthesia moved beyond generating training videos into live AI roleplay coaching that scores employees on their performance.
Adobe agreed to acquire AI upscaling and restoration specialist Topaz Labs, planning to fold its models into Firefly and Creative Cloud.
CapCut introduced a new top-end Ultra subscription with far higher AI credit limits, native 4K output, and faster generation.