> It's remarkable that academia hasn't produced rigorous studies with statistically significant samples, with hundreds of developers across languages and skill

> It's remarkable that academia hasn't produced rigorous studies with statistically significant samples, with hundreds of developers across languages and skill levels, to properly assess GenAI impact on productivity. It's much more remarkable, to me, that the AI vendors have been unable to publish a study on the effects of agents on developer velocity. I mean, if it was really a 10x (or even 2x) boost to productivity, why aren't they seeing it? The returns on each generation have been diminishing compared to the cost. Shouldn't it be the other way around if AI was such a productivity boost?