MIT researchers see AI becoming default for theoretical science beyond math

The story isn't that AI can do math now (that's settled). It's that institutions are treating this as a template for wholesale methodology replacement in other domains, which means the bottleneck shifts from 'can AI solve this?' to 'can we hire people fast enough to integrate it and who do we need?'

This follows directly from the math crisis documented over the past week. OpenAI's October 9-13 releases created a proof-of-concept that AI systems can generate research-grade output faster than humans can validate it (Platformer, Oct 9). The Decoder's coverage of Shing-Tung Yau's eight-institution coalition (Oct 13) showed that despite Western mathematician resistance, adoption momentum is real and organized. Now MIT is signaling that other fields see the same playbook working and are moving to replicate it. The question isn't whether this spreads; it's how fast and whether institutions that delay adoption fall behind on talent acquisition and funding access.

Track whether economics and physics departments announce new hiring for 'AI integration roles' or 'computational validation' positions in the next two quarters. If those roles outnumber traditional postdoc openings at top institutions by Q2 2027, it confirms this is a labor market restructuring, not just a tool adoption.