gpt-5.1 is a model. It is not an application, like ChatGPT. I didn't say that personal requests were 0% of ChatGPT usage.
If we are talking about a new model release I want to talk about models, not applications.
The number of input tokens that OpenAI models are processing accross all delivery methods (OpenAI's own APIs, Azure) dwarf the number of input tokens that are coming from people asking the ChatGPT app for personal advice. It isn't close.
I'm starting to think that I have a weak minds eye. I can not easily mentally visualize ideas, but I do feel them.
When I am thinking about something new there are no words, images, etc. just feeling my way through ideas. It feels kind of like acting on instinct. For example if I am wrestling with an idea I get an impression of "resistance". Feeling my way through the path of least resistance from impression to impression.
I do say words, or see flashes of images but only for ideas that I have already felt out. Words are kind of a breadcrumb trail so that I can retrace the exact train-of-thought that I had taken before and images like mile-markers.
If we are talking about a new model release I want to talk about models, not applications.
The number of input tokens that OpenAI models are processing accross all delivery methods (OpenAI's own APIs, Azure) dwarf the number of input tokens that are coming from people asking the ChatGPT app for personal advice. It isn't close.