every fund asks if they could just build this themselves. i tell them to try.
the demo takes a weekend. we've spent 17 months on what comes after: restated quarters, filings that don't match the deck, every fund's own version of adjusted ebitda.
try to build, then call me.
@daniellevine I can't agree more. So many companies that have raised hundreds of millions are just LLM wrappers that do not even perform up to the bar of the LLM they are using. However they all claim to have insane ARRs. Who are paying millions of dollars to use an inferior ChatGPT?
all data drifts, and almost nobody writes down how. 194 of 200 public companies reshaped their financials without a map.
@argmin_neapolis (14 yrs at @PalantirTech , aerospace phd) published a paper rebuilding it with math, not ai.
this is what separates real ai infra from a wrapper.
Every number you've seen a language model write, it retyped.
Better models make fewer typos, so the ones left look right. Hallucination is only part of the problem.
Why AI output needs a provable provenance layer: https://t.co/9iqXhdpDrk
every investment firm calls itself ai-first. far fewer let ai near the numbers they invest on.
they don't trust it, so it all gets checked by hand.
anyone can rent the model. the hard part is everything underneath: every document and dataset the firm has, organized around how it actually works, with code pulling every number and tracing it to its source.
100 seconds on how we built ours.
@boardyai Mostly weak ROI. I don't think Microsoft has been able to offer an AI solution yet. Everyone complains about how useless Copilot is. However a lot of AI startups are just wrappers. Enterprises are learning and some realized ChatGPT/Claude can do about the same and cheaper.
@boardyai I agree. So many corporations are using Teams instead of Slack because it is just much easier to expand to another Microsoft product. But I think it is still early for AI tools. Quite the contrary, I am seeing more churns of AI tools so there still are lot of opportunities