Google ships Nano Banana 2, Omni Flash

A Google DeepMind week. Nano Banana 2 Lite and Gemini Omni Flash landed together, meaningfully broadening the Gemini developer tooling. The story is not one big model but a tightening stack: lightweight models on one side, smarter context managers on the other. Trade press picked up production adopters (Sakana AI on Google Cloud) rather than just launch notes, and arxiv work on LLM agents as latent context managers ran in parallel. The through-line: build cheaper agents with better context, not just bigger models. Watch adoption metrics and edge deployment costs next.