What's this about?
dotey's observation about the release pattern: 3.6 Flash in late July, 3.7 Flash in mid-August, 3.8 Flash today — roughly three weeks apart each time. Meanwhile Gemini 3.5 Pro, which was due in June, still has not appeared. He notes the context: DeepMind had a leadership shake-up in early August, with Demis Hassabis moving out of day-to-day management to become DeepMind chairman and Alphabet chief scientist, chief scientist Jeff Dean leaving to start a company, and new SVP Koray Kavukcuoglu taking over Gemini development. His question is whether the Flash cadence is partly about visibility.
Why would a lab ship the cheap tier and not the expensive one?
Several readings, and the honest answer is that outsiders cannot tell. Flash models are cheaper to train and iterate, so a faster cadence is natural. A frontier Pro release is a much bigger commitment with more to go wrong. It is also possible the Pro line is simply not ready, or is waiting on infrastructure, or has been re-scoped. What is observable is the pattern; the explanation is inference.
Why is it worth noticing?
- Release cadence is one of the few genuine signals outsiders get about a lab's internal state.
- Mollick's early-access verdict fits the picture: "a very good Flash model, but not equivalent to a frontier model."
- Worth holding loosely — it is a pattern plus a plausible story, not evidence.