Criticizing OpenAI and Google for giving new models to a select few first
The author argues that Astra and Gemini 4 Argon were only opened to a small group, so momentum will be spent by public launch, and advises labs to either launch publicly at announcement or wait until the model is ready.
Why 'release' a model when normal people can't try it? OpenAI made this mistake with Astra, and now Google is making the same mistake with Gemini 4 Argon. The problem with such moves is that when it finally releases to the general public, the "momentum" is usually much lower, or even dead. Fable 5.5 could launch before then or around the same time - and it will probably overshadow everything else. So, a little advice for any lab that wants to release a model: either release it to the general public when announcing it OR delay the release until it's ready.