Warp is opening more than source code. The terminal company put its client under AGPL, moved product planning into public GitHub issues, and says nearly 1 million active developers can now steer Oz-powered agent builds in the open.
Why it matters: faster models stop feeling fast if orchestration overhead eats the gain. OpenAI says WebSocket mode made agent workflows up to 40% faster end to end, while lifting effective inference speed from about 65 to nearly 1,000 tokens per second.
The funding headline matters because it prices a new layer of the AI stack: infrastructure for agents that need live access to the web. Parallel says it raised a $100 million Series B at a $2 billion valuation, bringing total funding to $230 million just five months after its last round.
NVIDIA is targeting the cost bottleneck in multimodal agents, not just the demo factor. Nemotron 3 Nano Omni claims up to 9x higher throughput, a 256K context window, and six leaderboard wins for document, video, and audio understanding.
Cursor is pushing coding agents out of the editor and into infrastructure. Its new SDK exposes the same runtime and harness behind Cursor itself, targeting CI/CD jobs, cloud execution, and embedded agent workflows inside other products.
The important shift here is distribution, not one more model endpoint. OpenAI says GPT-5.5, Codex, and Bedrock Managed Agents are entering limited preview on AWS, giving enterprises a way to keep identity, security, and procurement inside Amazon's stack.
HN liked the hack, but the real excitement was infrastructure. Cua’s background macOS driver keeps the cursor, focus, and Space in place while an agent works inside another app.
Why it matters: most agent stacks look solid in a notebook and fragile in production. Mistral says Workflows is already running critical processes at six named organizations and adds the pause-resume, observability, and fault tolerance teams usually have to build themselves.
The improvement sounds small until you remember where agent products lose trust: waiting. GitHub says its Copilot cloud agent now starts more than 20% faster, building on a 50% startup improvement shipped in March.
This matters because Beijing just showed that cross-border AI acquisitions can be reversed even after the people and product start moving. TechCrunch reported on April 27 that China blocked Meta’s $2 billion Manus deal, ordered the transaction unwound, and left one of the agent race’s splashiest bets in limbo.
This matters because Copilot is no longer priced like a lightweight autocomplete tool. Starting June 1, 2026, GitHub will convert every Copilot plan to token-based AI Credits, end the fallback model safety net, and make code review consume GitHub Actions minutes too.
This matters because Xiaomi just put a frontier-scale model family behind permissive terms instead of a closed API gate. The MiMo-V2.5 release promises a 1M-token context window, MIT licensing for commercial use and fine-tuning, and a Pro variant Xiaomi says leads open models on GDPVal-AA and ClawEval.