Everyone Owns the Same Models. The Best Quant Shops Still Pull Away.

Wall Street's AI edge has migrated from the model to the workflow, and the firms that understand this are quietly building a moat their rivals cannot copy.

From a Skeptical Summer to a Settled Question

Two years ago, one sentence would have sounded like pure bravado: “You should come do quant here, because the AI harness we give you is better than anywhere else.” For decades, a fund recruited talent by advertising its platform, its capital, its data, its founders, its trading permissions, and its payout structure. Now a new item has been placed on the table, an AI working environment good enough, deep enough, and close enough to the research process to become a selling point in its own right.

That framing comes from Jonathan Regenstein, who leads wealth and asset management at Snowflake. Back in the autumn of 2024, the questions he fielded were still elementary. Would the model hallucinate? Was the underlying data trustworthy? Could a conclusion be reproduced? Who owned the blame if the answer was wrong? Portfolio managers and quants listened with their arms crossed, intrigued but unconvinced that a machine capable of writing a paragraph of Python belonged anywhere near a live investment decision.