Today’s reading list covers a couple of Google DeepMind releases — a new Gemini Omni 1.1 Flash developer update with added creative controls and generative video, plus a first pilot of double-blind AI evaluation methods — and an interesting new open-weights Qwen model that previews the Qwen4 architecture. From individual writers, there’s a look at a prompt-injection attack against Claude Code’s now-default auto mode.
From the labs
- Gemini Omni 1.1 Flash lets you build with more control — Google DeepMind
- Qwen3.8-Flash-Next — Qwen
- Piloting the world’s first double-blind AI evaluations — Google DeepMind
Worth your time
- Breaking Claude Code Opus 5 Auto Mode — Simon Willison