BlackRock's Sept 22 paper, written with the Bitcoin Policy Institute's research, says AI models favored Bitcoin for long-term value, stablecoins for payments.

Headlines are calling it massive for crypto.$ETH

Here's what they're skipping though.

BlackRock's own paper says it plainly, these findings reflect simulated model responses, not observed agent behavior. Real agents aren't out there transacting and choosing Bitcoin.$BTC Models answered a hypothetical question in a controlled setting.

Bitcoin won 79.1% of the time in that original study. But a newer round of the same research told a different story, Bitcoin's share actually falling, with fiat leading at 41.4% and Bitcoin trailing at just 13.8%.

I don't think that makes BlackRock's broader thesis wrong, tbh. Agents probably will need some kind of machine-native payment rails eventually.

Just noting the asterisk matters as much as the headline does here.