The alpha chase just flipped inside out.

The Summary

  • BlackRock and Balyasny are pointing AI agents at their internal data, mining past research, analyst communications, and historical investment decisions for signals competitors can't access.
  • Alternative data sources (credit card receipts, satellite imagery, foot traffic) became commoditized once everyone bought the same feeds.
  • With LLMs scraping all public data, the new edge is proprietary: your own firm's institutional memory.

The Signal

BlackRock's Jacob Bowers called it directly at the Future Alpha conference: AI is "great at structuring unstructured data," and "some of the best unstructured data you have is internal." Translation: every analyst memo, every internal Slack thread about a sector thesis, every research report that never made it to clients, those are now training data for agents trying to predict what moves next.

This is the natural end of the alternative data arms race. For a decade, asset managers competed by buying weirder data. Satellite images of Walmart parking lots to predict retail earnings. Aggregated Venmo transactions to gauge consumer spending. Shipping container movements tracked by IoT sensors. The problem: once Goldman buys the feed, so does Citadel, and then it's priced in. The edge lasted maybe two quarters.

LLMs accelerated that commoditization to weeks. Any publicly scrapable data is now everyone's data. So the only data moat left is what nobody else can access: the accumulated judgment calls, failed theses, and pattern recognition sitting in your own analysts' heads and inboxes. BlackRock manages $14 trillion. That institutional memory, properly searchable by agents, is a dataset no competitor can replicate.

The shift is architectural. Instead of agents reading the entire internet for common knowledge, they're reading your company's brain for uncommon knowledge. The question isn't "what does the market think," it's "what did our team learn three years ago about this exact situation that we forgot."

The Implication

If you're building in finance, this is your blueprint. The value isn't in scraping Twitter or buying satellite feeds anymore. It's in making your proprietary knowledge queryable. If you're an analyst, your old reports just became training data. If you're anywhere else, ask what institutional knowledge your company is sitting on that agents could turn into competitive advantage. The firms that win Web4 won't be the ones with the best public data pipelines. They'll be the ones whose agents learn fastest from their own history.

Source: Business Insider Tech