Ask HN: What is still hard to research in investment analysis?
I've been doing it for a while for myself, somewhat isolated from industry standards. I would like your help to understand if there's a broader hunger for the techniques I've developed, a market.
LLMs have made it possible gather big-picture qualitative insights from scratch by analysing thousands public statements from companies: New/Upgraded Products, New/Upgraded Infrastructure, Strategic Direction and Commitments, R&D Initiatives, M&A and other Partnerships…
By finding common patterns and grouping them, by extracting structured features, categorising, filtering, aggregating, by relating them to each other, by enriching via automated follow-up research in other sources... One can get novel metrics about the concrete moats and plans of a company or a whole market, that are otherwise not available. Of course, this has been possible for a while, but it required training specialised Deep Learning models for every type of feature one wanted to extract. LLMs allow for much faster exploration without training, which enables modes of analysis that were simply infeasible before.
I've been wondering though: What concrete pain-points can be addressed like this for professional analysts? What categories of research still take hundreds of human-hours, even taking full advantage of all the modern tools and sources?