LLMs Turn All of This Into a Markdown File

Nicolas Bustamante, founder of Fintool (an Anthropic-backed competitor to Bloomberg and FactSet), writing about the vertical software selloff:

LLMs collapse all proprietary interfaces into one Chat.

Years of engineering versus one week of writing. That's the shift.

Competition doesn't increase linearly — it explodes combinatorially. You don't go from 3 incumbents to 4. You go from 3 to 300.

There's a useful framework buried in here. Bustamante breaks vertical software defensibility into ten discrete moats and stress-tests each one against LLM capabilities. Five hold, five don't. That's a better lens than the "vertical software is dead" takes that have dominated the selloff coverage. His distinction between revenue decline and multiple compression — that stocks can crater 60% while revenue stays flat, because the market is repricing the durability of the business, not its current cash flows — is the single most clarifying observation I've read about this selloff.

But. Bustamante is not a disinterested analyst. He's the guy building the thing that's doing the disrupting. His company's valuation depends on every word of this thesis being correct. This isn't hidden — he says it openly — but it means you should read the piece the way you'd read a brief: as the strongest possible case for one side.

The "80% of capability at 20% of the price" line is doing load-bearing work that it can't support. In financial services, the last 20% is where the liability lives. A hedge fund doesn't need a fast answer. It needs an auditable, reproducible, defensible answer that holds up when a regulator asks how the number was derived. The distance between "impressive demo" and "I'll stake my fiduciary obligations on this at 3am" is where vertical software companies actually earn their margins.

"Competition goes from 3 to 300" sounds devastating for incumbents until you remember how enterprise procurement actually works. Three hundred options doesn't empower buyers — it paralyses them. In conservative industries, a fragmented market of AI-native startups can reinforce incumbent positioning. Nobody has ever been fired for buying Bloomberg. That instinct doesn't dissolve because someone demoed something slick.

The piece generalises from the one category where LLMs are strongest — search-and-access layers over public or semi-public data — to all vertical software. That's the segment Bustamante knows, and he's probably right about it. Financial data terminals built on licensed exchange data and legal research platforms built on public case law are getting repriced for real. But most vertical software doesn't look like that. Most of it is tangled up in regulatory compliance, transaction processing, and institutional trust in ways that make "just replace it with an agent" a fantasy for the foreseeable future.

The most quotable line in the piece: "LLMs turn all of this into a markdown file." It's the kind of sentence that sounds more true the less production software you've shipped.