"Here are a few theses: After 2.5 years of insane hype, there’s no evidence that current AI is making the design process faster. Good design comes from a broader process, not a one-off conversation – meaning the one-off chat paradigm is unlikely to generate good design for non-designers. AI architecture means it will continue to be worse at designs that are 'out of the training data' – the bold, the novel, designs with very tight constraints."