How the AI bubble will pop

https://novacut.ai/

Shashank and Mark break down a packed few weeks in AI: new open-source and local models, MCP and tool  use, Chinese labs closing the frontier-model gap, Google's full-stack advantage, and the exploding cost  of data-center buildouts.

They debate whether AI spending keeps compounding as models get cheaper and demand rises, or whether  smaller, local, good-enough models eventually puncture today's valuations.