Based on recent podcast episodes, here's my attempt to summarize the Cambridge, MA view of the economics of AI, particularly Acemoglu's version. I find it unconvincing, but it's useful to lay out the components. 1. The ideology of technologists and capital providers directs too much capital and top talent toward AGI-pilled companies. 2. These companies devote too much effort to improving general models and too little to applications and diffusion, including for pro-worker AI. Pursuing AGI is not necessarily the best route to near- and medium-term productivity growth. 3. Pro-worker AI can create new tasks for workers and can make human expertise more valuable. It is feasible and can be deliberately developed. This requires knowing enough ex ante about which technological directions will increase labor demand. 4. Changing beliefs and correcting incentives that favor automation would redirect innovation toward this alternative. The result would be greater shared prosperity and healthier politics, because workers' continued economic importance helps sustain their political power.