What Happens When a New AI Model Drops?

Dylan Field says new model drops put people in an AI psychosis for a few days. Algo feeds aid the moment, because the best results rise to the top and everyone treats the model as everything. A few days later the main tropes are clear. People adapt fast, the novelty is gone, and the model no longer rewards them or fires off the dopamine system. The reaction is that the model always does the same thing, so the next question is how to move past it.

Dylan Field calls the pattern the continued loop. Models keep getting so good in formal domains. In math and cybersecurity, the IQ points stack and stack and stack. For creativity and for design, progress is very incremental overall. He says a change might come at some point, and he thinks non-verifiable domains have something to do with the gap. He says the human role in the AI toolkit is probably to conceive, create, and design, and to let AI do more of the work, which gives people the ability to do much more work.