The debates

Mollick argues Fable 5.2 is close to a jagged general intelligence

Ethan Mollick, a professor of entrepreneurship who writes One Useful Thing, defined general AI as a system as good as most experts at most intellectual work done on a computer, and concluded that Fable 5.2 comes close, with gaps in taste and judgement. Greg Brockman, president of OpenAI, wrote that frontier models with heavy test-time compute are becoming the engine of scientific discovery.

Filed under Large language models, Reasoning and inference

Mollick’s case rests on his own trials. Without instructions, the model turned a small USB keyboard into a console with six games. Asked to launch a business, it screened 22 ideas and drafted one built on public agendas from 201 city and county governments, which Mollick stopped before launch. It also wrote an empirical paper he judged ahead of a second-year doctoral student but not publishable. Where it fell short was taste and judgement, and it did not notice those problems itself. On X he described an experiment in which about 25,000 agents building 3D worlds developed an internal bureaucracy that slowed them down.

Brockman’s claim is broader and is a statement of his view rather than a measured result. The week’s evidence cuts both ways: Anthropic’s own benchmarks put Fable 5.2 first on most tasks it reported, while the company itself still recommends Opus 5.5 as the starting point for most coding and knowledge work.