10 years building vertical software: are we cooked? | Hacker News Reader

10 years building vertical software: are we cooked?

twitter.com

49 pointsnbstme51 comments

> Now consider what the same analyst does with an LLM agent: "Show me all software companies with over $1B market cap, P/E under 30, and revenue growing over 20% year over year. Build a DCF model for the top 5. Run sensitivity analysis on discount rate and terminal growth."

While I think LLMs can improve the interface and help users learn/generate domain specific languages, I don’t see how a professional can trust an llm to get a technical request like this correct without verification. Wouldn’t a financial professional trust the Bloomberg llm agent that translates their request into a set of Bloomberg commands more?

Yes, I balked at this point as well. Moreover, how do you accommodate new analysis concepts? What are you training the model on?

Real financial analysts already have DCF spreadsheets where they can just plug in numbers for any company. An LLM can help with fine tuning or catching errors but it's not a game changer.

I think there's a lot of overstatement about LLM capabilities throughout this piece, but I think it's generally directionally correct. There's an attitude of "LLMs are just going to directly perform business logic" or "data extraction and ingestion" or "calculations". The reality is that deterministic human-mediated code is going to do all that stuff (and AI is going to drastically amplify human leverage in building that code), and LLM agents will call into it as tools.

It's like the people who talk about how LLMs can't count the r's in "raspberry" and don't seem to understand that GPT5 can reliably e.g. work out a transformed probability distribution function from a given PDF by integration and derivation --- in part because frontier models are smarter but more importantly because they're all presumably just calling into CAS tooling.

It’s worth asking: what do Wall Street traders know about building software companies? Almost nothing. Anyone who has attempted to start a startup knows that the software is always the easy part. Building the business is hard. The notion that we’re going to undo 100K+ years of specialization just so that companies can run mediocre, buggy versions of SaaS tools just to save a few bucks is crazy to me.

SaaS stocks are currently the buying opportunity of a lifetime.

Also, has nobody thought about operations? I pay SaaS companies so I don't need to think about operations. It is insane to me to think that people think a restaurant company is going to suddenly want to get into managing the operations of its vibe coded accounting app. Absolutely not.

Yes, I’m very aligned with “minimal dependencies, live off the land, roll your own” to avoid lock-in and vendor bullshit. Shit I’d rather rent baremetal from Hetzner than VPS from EC2/GCP/etc if I have enough workload to justify it.

But for my startup I still use a ton of SaaS services for things that I could probably do just fine myself. (Clerk/StackAuth, Supabase/PlanetScale, Cloudflare STUN/TURN, Clickhouse, Vercel, Calendly, Google Workspace, ngrok, Tailscale).

Spiritually, I hate using these. Any one of these would be dead-simple to replace. But my time is genuinely better spent on my startup’s particular value-add. Maybe I’ll replace these some day when we can hire someone to manage internal replacement services - some of which are as easy as “a postgres database” or “wireguard on some VPS instances”. But it’s just not worth my time right now when I’m focused on building revenue.

Even if they all cost $300/mo in total, and we’re bootstrapped, it’s a lot easier to cut back on UberEats or shiny nerdy toys than it is to replace all of these SaaS offerings. I recognize there’s a lot of ”I don’t know what I don’t know” and I’m liable to subtly misconfigure something in a potentially disastrous way.

There's enough unhappiness with commercial SaaS EHRs that I expect as few health system CIOs will decide to operate their own vibe coded replacements. This won't work out well for most of them but I think it's going to happen.

Or how you'll get all your data out and imported into your custom system.

Anecdotally, my partner does experience design for fintech systems akin to Bloomberg for a particular niche. They do the standard lifecycle of user research, design iteration in figma, handoffs to devs and all that. The tech department at this company has also been building OpenAI integrations for more than 2 years and are neck deep in LLM technology doing exactly what this thread describes. My partner is still doing exactly the same work she's been doing the entire time and getting the same level of adoption for all the bespoke UI development while the chat interface is just kinda there. I'm sure it's getting usage for some tasks, but it's supplemental.

We spent the last 20 years building a whole set of specialised little tools on the web to do specialised little tasks for businesses.

Now an LLM can do most of that, easier, and effectively for free. The first pass on a new problem is not googling to see if there's a SaaS for that, it's prompting an LLM to see if it can do it, or if it can build a tool that can do it.

Case in point: in my job we have to data-enter invoices. I have dealt with or in this industry for 30-ish years. I worked on various projects trying to get computers to read invoices, to various degrees of success. It's a hard problem; there's no standard format, or layout. Every company does its invoices differently. Some are Excel files, some are PDFs, some are Word docs, etc.

This entire problem vanished this year. You get an LLM to read the invoice. It does this more accurately than humans do. Job done.

There are entire SaaS businesses that read invoices that are now obsolete and have no moat.

There has been a standard X12 EDI format for invoices for decades. It's kind of a hassle to work with but it can at least be reliably parsed. A lot of huge businesses like Walmart use it successfully, and even require their suppliers to submit all invoices that way.

I don't object to using LLMs to parse PDFs but over the long run it's going to be less efficient and reliable than other options.

Yes, there has been a standard format for invoices for decades, but it was only ever used if both companies were using a ERM system (and as you say, large enough purchasers could force their suppliers to). We have to deal with small business who don't use the standard format, which is the vast majority of them.

Please go ahead and try parsing non-standard invoices without an LLM. I spent 20+ years on and off dealing with this problem. It's not as simple as it looks. And then LLMs came along and made it simple.

I think that’s different. You have a problem: invoice management. LLMs have made that cheaper and you should expect disruption. But you’re not building your own invoice scanner. You’re using another, cheaper product on the market.

However, the hypothesis in the SaaS market is that LLMs have made software have zero value and therefore the SaaS companies will be less profitable. That’s like if wood was suddenly free, expecting home builders to go out of business. If anything, home builders are going to do better, because they can apply their expertise while deploying capital elsewhere. We should expect software companies to be more profitable, not less.

Of course, there are exceptions. Sometimes AI replaces the product itself, e.g. image generation models vs. contractors on fiverr.

> SaaS stocks are currently the buying opportunity of a lifetime.

Is anything down far enough to look like the opportunity of a lifetime right now?

There were some deals out there over the last few years. I still kick myself on a daily basis for not taking Carvana at $5 in 2023.

The bottom of the market fell out because of political turmoil. The U.S just revised job numbers down by a million for the first time in decades. The author’s thesis about the future of software businesses could be right or wrong but it is not the cause of the recent market collapse.

Show me even a single trader who would prefer a slow-ass chat interface to 36 monitors full of stock tickers and charts.

Heck; show me anyone who is not more productive with a boring old UI once they learn how to use it. Maybe with a chat interface as a how-to assistant.

There's probably legit uses in finance for LLMs, in exploring potential scenarios and strategies, spit-balling ideas, summarizing financial news. Sure. Great. Maybe even as a trading agent, questionable, but I'd buy that as maybe happening.

But replacing Bloomberg terminals with "Chat with NYSE", that is no exaggeration one of the most out of touch ideas I think I have ever heard in my life!

Is trading even a real thing? Is there really a job title "trader"? Entities really think they can outperform DCA SPY?

I guess you better go tell them they don't exist.

Yes. Traders are involved in all kinds of deals that aren't like index funds.

Who do you think makes the price of SPY change?

I think using the BB terminals as the big example here was probably not the best move; there are other huge vertical software platforms where this argument is more of a layup.

(Though: you can do a lot with IB, right? Same deal.)

My experience is that in most of small companies, the revenue extraction cycle from big SaaS suites has gone far enough that people are moving to self-hosted (or rather colocated with a local-ish provider that installs and provides unofficial support) software, like they used to do before the great plague.

Considering how Zipf's law works, there might be a huge discrepancy coming on as we see the products deteriorating from all the AI and H1B spaghetti code to the point where LibreOffice appears quite competent by comparison. Most of the people I worked with just want to sell lamps or furniture or trumpets or whatever they do, and the inventions of modern SaaS make this a lot harder to do. Once enough small businesses stop paying their 5-user subscriptions, I think this whole thing will pivot heavily into the favor of those that just maintained their product well and didn't ensloppify it in the mantime.

I build a website generator from Word files in a folder for a coop I'm in. Four years back I had to move it to using libre office because Word broke how links in files work. As of last year the whole coop is using libre office because it works better than ms office which has corrupted old important files multiple times. I don't see a reason why they can't move to an xfce Debian. Having a consultant come in once every 5 years to update everyones computers is a lot cheaper than all the ms licenes you have to pay for now and easier to use without constant breaking changes.

I can follow the arguments, and I find many of them plausible. But LLMs are still unreliable and require attention and verification. Ultimately, it's an economic question: the cost of training the model and the computing power required to produce accurate results.

The strongest argument is the one about the interface. LLMs will definitely have a large impact. But under the hood, I still expect to see a lot of formally verified code, written by engineers with domain knowledge, with support by AI.

I find this line of thinking so hard to follow. I happen to have worked at FactSet and when I was there we were stuck supporting chromium 18 because it was embedded in the version of the workstation that a partner had signed a contract which prevented them from needing to upgrade the base install. SASS is of course to some extent about functionality. Maybe it will make sense for larger customers to duplicate functionality themselves using AI. I very much doubt we’re anywhere near being able to prompt Claude and get FactSet out but fine we can grant that. The reason SASS is so valuable though is because it allows firms to pass off the costs and risk of maintenance behind SLAs. Why would you want to bring that risk in-house when you can essentially buy insurance against software errors by paying a SASS company?

I think this is more interesting as a rubric than as a prediction. I agree with some of it and not with others; I don't know if we're "cooked" or not; I do like how they've broken vertical software's moats down though.

1. I don't buy that chat interfaces will replacing existing user interfaces. I'm in particular a little bit familiar with Bloomberg's user culture. I don't know that I buy that it's going to be replaced with LLM chat prompts. But software agents are going to make faithfully reproducing those existing user interfaces much easier, so: half credit?

2. Half credit again on LLMs vaporizing the "business logic" moat, because the vertical-specific rules that justified the original software market are I think a lot harder to encode in Markdown than the 1 week they gave it, and also verification becomes a bear as more ground-truth business logic is replaced with nondeterministic AI output. There's a thing happening here for sure, I just don't buy it's as decisive as they say.

3. Public data access: I 100% buy this. If this was a real moat, it's dead.

4. Talent scarcity: same deal. Remember, we're talking about vertical software, where the underlying technical work is fairly repetitive and best-practices driven; it's the exact slice of software development work LLMs excel at.

5. Bundling (you get IB messaging along with your charting and your news service); maybe. This point feels tautological. Work out what LLMs do to each of the bundled experiences and there's your answer for how resilient that moat is.

6. Proprietary data: I think they're just dead on right here, and it does indeed seem to be a good time to be a company like Bloomberg?

7. Regs lock-in: half credit, because AI does make regs compliance a lot easier, and I think we're at the very early stages of seeing how.

8. Network effects seems like a repetition of "bundling" and if I have a qualm about this rubric it's that they made it look like an even 10, so they could have clean wins and losses.

9. Transaction embedding (ie, being a payment processor or a loan originator) also seems tautological; it's a moat, sure, but they're begging the question of whether AI enables people to stand up viable competitors.

10. I think "system of record" and "transaction embedding" are kind of the same moat.

I wish people would not blog on X (I will call it X when it's used as a crappy blog platform); these ASCII charts are awful. But that's neither here nor &c.

Great theory, in practice I just expect things to relocate and shuffle around as it usually goes.

> The S&P 500 Software & Services Index, 140 companies, fell 20% year to date.

The overall index has been pretty well flat. What sectors gained?

And surely there aren't 140 "software and services" companies in the top 500 by market cap?

Pretty sure there are 500 "software and services" companies in the The S&P 500 Software & Services Index.

500 is the target, not a hard requirement. The current SP500 index is comprised of 503 constituents.

I’m also pretty sure S&P doesn’t maintain a ‘S&P 500 Software & Services Index‘.

There is a ‘S&P Software & Services Select Industry Index’ with 140 constituents. That’s probably the index in question.

Right; so if it's not "the software and services subset of S&P 500", then TFA was mistaken in including "500" in the name.

That's what I was trying to establish.

Either that, or there really are 140/503 software and service companies, and I learned something surprising.

Agreed. It’s likely just a typo.

> The overall index has been pretty well flat. What sectors gained?

Data centers and AI.

The current US economy is flat except for AI and data centers.[1]

[1] https://fortune.com/2025/10/07/data-centers-gdp-growth-zero-...

From the guidelines:

> Please don't use HN primarily for promotion. It's ok to post your own stuff part of the time, but the primary use of the site should be for curiosity.

Was this generated by AI?

Anyway, I dispute that there's "zero switching" costs to go from proprietary keyboard shortcuts to english sentences as the interface.

In the bloomberg example, the shortcuts are precise. LLM's responses are not always what you want.

Imagine being a vim or emacs user and have those replaced by something you have to type entire sentences for functionality.

"Public Data Access → Commoditized"

Also, no. Today I tried to have gemini pro give me some data from wikipedia for a list of countries that I supplied. It gave me data and source links, but the numbers were all wrong! I would have zero confidence in this for anything serious.

Even if it wasn't generated by AI, it reads like an advertisement for some company which sure sounds like a wrapper on top of a LLM. I think the moat for that type of company is a lot less defensible than any of the Vertically Integrated category leaders by a country mile.

> Imagine being a vim or emacs user and have those replaced by something you have to type entire sentences for functionality.

As a former vim user who uses cursor, I've found that as the models get better I'm typing less and less. I appreciate the vim key bindings, but eventually I can imagine not missing them.

> Imagine being a vim or emacs user and have those replaced by something you have to type entire sentences for functionality.

This has already happened for a non-negligible amount of people on this site. I still use magit to prepare commits and review diffs. Everything else is English and Claude code or opencode.

Ironically, part of the reason I like emacs and formerly liked vim was because it reduced the amount of time my hands had to leave the keyboard. I simply look where I want to type, press the chord to jump with avy, and then begin typing. New tools are spiritually aligned with this goal. I look at the screen, I think about what I don’t like, and now instead of translating the criticism into code changes, I just stream my thoughts directly into the tool.

So many hallmarks of LLM authorship (the arrows, the framing, the unnecessary fluff, the big important-sounding words). Not to mention the vested interests of the author. Let’s just call it what it is: bullshit. No signal, just noise.

LLMs do a lot of things, but none of them well enough.

Now that this guy is selling AI, surprise, surprise, its going to revolution everything else...But its misleadingly written as "we" are cooked.

Its another ad, no?

Look at his submission history, nothing but self promotion. His entire presence here is just an ad.

You both need to take this stuff to hn@ycombinator.com instead of hashing it out here. I think you're wrong, but either way: stop crudding up the threads when all you've got to say is "this person submitted their own post". People do that all the time.

Again: if you think they've crossed the line, the guidelines specifically ask you not take to the threads with it. Mail Dan and Tom. They'll get back to you quickly.

yes.jpg

Getting really sick and tired of the AI doomerism.

I’m sick and tired of people forgetting that with every field, the first 95% is easy, the last 5% is almost insurmountably hard and what people get paid for.

Yelling “it’s 95% there, we’re so cooked” when anyone can be 95% there for any field since Google was invented doesn’t show much.

That may be true, but it can mean layoffs of 95% of the employees.

You completely missed the point.

Technology, Google, all of it, makes it easy enough to learn how to do 95% of many common roles and fields. However, you will never be hired, or replaced, until you master that last 5%. Every role, even some of the lowest ranking, has its own insurmountable 5%. How well has fully replacing cashiers with self checkout gone, after… two decades… of trying?

If we can’t fully replace a cashier yet, the most automatable role in existence, politely stuff your pie hole when saying AI will replace skilled roles.

Reply on news.ycombinator.com