Rant.
Their poor communication is exemplary in the industry. You can't even ask the old models about new models. The old models think that 4o is 4.0 (cute, team, you're so cool /s), and think that it's not possible to do multimodal. It's as if model tuning does not exist. I had a model speaking to me telling it cannot do speech. It was saying this out loud. I cannot speak, it said out loud. I get that the model is not the view/UX, but still. The models get other updates; they should be given at least the basic ability to know a bit of their context including upcoming features.
And if not, it would be great if OpenAI could tell us some basics on the blog about how to get the new features. Unspoken, the message is "wait." But it would be better if this was stated explicitly. Instead we wonder: do I need to update the app? Is it going to be a separate app? Is it a web-only feature for now, and I need to look there? Do I need to log out and back in? Is it mobile only maybe? (obviously unlikely for Canvas). Did I miss it in the UI? Is there a setting I need to turn on?
This branching combinatorically exploding set of possibilities is potentially in the minds of millions of their users, if they take the time to think about it, wasting their time. It brings to mind how Steve Jobs was said to have pointed out that if Apple can save a second per user, that adds up to lifetimes. But instead of saying just a simple "wait" OpenAI has us in this state of anxiety for sometimes weeks wondering if we missed a step, or what is going on. It's a poor reflection on their level of consideration, and lack of consideration does not bode well for them possibly being midwives for the birthing of an AGI.