Thread

Show HN: An AI agent that trades inside limits you set, starting on paper

16 points · 8 comments · henryzhangpku

dpoloncsak · 1785948892 · focus · HN ↗
Do you see any edge using LLMs that more traditional ML doesn't have? In my limited experience, small models trade off noise, whereas frontier models are too slow to be usable for something like HFT. I find you really want something trained on market data, not generalist next-token predictors.
1. georgemcbay · 1785974610 · focus · HN ↗
  > Do you see any edge using LLMs that more traditional ML doesn't have?
  Depends upon whether you pay for early access to Trump's Truth social posts. Which, yes, is really a thing, because of course it is.
  <a href="https://www.pbs.org/newshour/show/paid-early-access-to-trumps-posts-raises-insider-trading-questions" rel="nofollow">https://www.pbs.org/newshour/show/paid-early-access-to-trump...
  1. dpoloncsak · 1786020235 · focus · HN ↗
    Sure, but that's an edge anyone can get for paying.
    I'm asking how specifically does the Large Language Model contribute to generating alpha over traditional ML used in the field, such as regression models

Unofficial Hacker News client; not affiliated with Y Combinator.