Richard Craib: WallStreetBets, Numerai, and the Future of Stock Trading | Lex Fridman Podcast #159

WallStreetBets coordinated millions of mostly anonymous retail traders to engineer a short squeeze in heavily shorted stocks like GameStop, showing that meme-driven, bottom-up movements can meaningfully damage or discipline large hedge funds.

Craib distinguishes diversified, risk-managing shorting (e.g., shorting oil to go longer Tesla) from ‘evil shorting,’ where large funds take oversized positions, coordinate narratives, and try to crush specific companies purely for profit.

The fund gives away anonymized, high-quality equity data, lets anyone build models locally, and only receives predictions; contributors stake Numeraire (NMR) on their models, earning more if they perform and getting burned if they don’t, which naturally surfaces serious, high-quality models.

Craib argues that staking—putting capital behind a claim in an unbreakable smart contract—could replace flimsy signals like karma or screenshots on platforms such as Reddit, making promises and predictions more trustworthy and harder to fake.

Unlike typical ML benchmarks, market data is noisy, constantly changing, and heavily exploited by others; Numerai participants must neutralize exposure to obvious risk factors and design robust ensembles rather than overfitting historical quirks.

Craib believes a narrow AI will eventually ‘win all the money in the stock market,’ and Numerai is structured to be the platform on which that kind of model emerges, aggregating global talent and data into a single, ever-improving trading system.

He advises founders not to treat startups as short experiments or quick flips; instead, they should openly state bold, even uncomfortable goals (like managing all the world’s money) to attract the right people and avoid building mediocre, forgettable companies.