How should we process AI's math breakthroughs? Bharath Ramsundar's field guide

rbhar90 · x · 2026-09-24

DeepForest author Bharath Ramsundar published a long-form essay on what recent AI math breakthroughs mean for mathematics and beyond.

Key points:

Mathematics may be undergoing its biggest shift in centuries, but the impact hasn't yet rippled into broader society; the author, a former math major who writes proofs professionally, processes his own emotional reaction.
He traces the lineage: DeepMind's RL run from Atari through Go and StarCraft, OpenAI's DOTA, and why game-solving approaches stalled on real-world problems without explicit rules.
In the quoted thread, rbhar90 is skeptical that AI auto-accelerates science: AI training is increasingly physical—power, datacenters, infrastructure. AI as a 'mathematical telescope' may optimize chip design, rack layouts and energy protocols, but physical hard constraints remain, just as LP, convex optimization and deep learning shifted disciplines without upending society.

The essay offers scientists and engineers a framework for the current mix of excitement and confusion.

More from AGI Musings