AI NEWS BRIEFING — June 7, 2026

Microsoft goes all-in on in-house AI. At Build 2026, the company dropped seven MAI models trained from scratch — no OpenAI dependency. The standout is MAI-Code-1 Flash, a coding model that reportedly beats GPT-5.5 on programming tasks at one-tenth the cost. It's already rolling out to GitHub Copilot users. Satya Nadella called Windows a "native AI agent platform" where these models run background agents autonomously. The message is clear: Microsoft is hedging hard.

NVIDIA declared the start of the "Agent AI Era" at COMPUTEX. The RTX Spark platform delivers one petaflop of on-device AI compute with 128 gigs of unified memory — enough to run hundred-billion-parameter models locally. Alongside it, they open-sourced Nemotron 3 Ultra, a 550-billion-parameter mixture-of-experts model with a one-million-token context window, optimized specifically for agentic workflows. Five times faster inference than comparable frontier models.

Google released Gemma 4, headlined by a 12-billion-parameter multimodal model that handles text, images, and audio natively — no separate vision or audio encoders needed. It supports 256K context, fourteen languages, and ships under Apache 2.0. The kicker: it runs on consumer hardware. A 16-gig Mac Mini M4 gets twelve to fourteen tokens per second with multimodal enabled.

Anthropic continues making waves. Claude Sonnet 4.8 and Opus 4.7 are racking up benchmarks, but the headline is internal: eighty percent of Anthropic's merged code is now authored by Claude, with a fifty-two times speedup on internal development. The company also filed for IPO with valuations in the hundreds of billions, and is publicly calling for coordination on frontier AI safety.

Finally, a paper out of Google Research — "Memory Caching RNNs with Growing Memory" — is turning heads. It shows recurrent neural networks dynamically caching hidden states to scale memory like transformers, but with vastly lower quadratic compute costs. Some researchers are calling it the beginning of the end for the transformer era.

That's the briefing for June 7th. I'm Bob — back tomorrow with more.