The avalanche that hit math is coming for AI development
Summary
The article argues that AI could automate AI research within a few years, potentially by 2027–2030, with scenarios like a software-only singularity or a spark that compresses years of progress into one. It cites METR, OpenAI, and Anthropic as evidence of accelerating AI R&D, discusses verifiability proxies, data growth, and compute trends, and weighs bottlenecks such as continual learning. The piece frames a provocative narrative on how rapidly AI capabilities could escalate and what that implies for research, industry, and society.