We are certainly closer now to being able to prototype and go to market faster with a product. In one weekend is a little much but I think its hard to deny that building will continue to expedite. What most developers don't think about is that the marketing, sales, customer service are all non-trivial parts of the business/product and all require legwork that is more than just sitting at an IDE. The nail in the coffin is that the data is a large part of company moats, and new products need time in the market to get that. Migration is also a long process and risky...so to get customers, a newcomer needs to provide way more value than what the incumbent gives.
I imagine you're going to have people trying to automate the whole GTM lifecycle, but eventually the developer that thinks they can bootstrap a one man enterprise without actually doing any kind of social interaction will run into a wall.
> We are certainly closer now to being able to prototype and go to market faster with a product
Absolutely. But this begs the question that businesses want to also sign up to maintain whatever product they've built, on top of their core business.
"Service" is the word that people seem to be forgetting in SaaS. If you roll you own, all you have is software.
The "service" part can still come from internal sources. Plenty of companies have internal support for internal tooling. With a good foundation of infrastructure and a lean team that knows how to build, not just vibe code, there is most definitely some ROI there.
But yes, this brings us back to the point that simply building the tool is only a small part of the process...and it has often been one of the most expensive parts of the process.
> We are certainly closer now to being able to prototype and go to market faster with a product.
What are the higher-order effects when anyone can do this, and *aaS becomes a market for Lemons?
I think that just because anyone can do it, doesn't mean they will. Lots of people have really great ideas but very few actually commit to execution. Ultimately ROI will go down, deincentivizing the commercialization of that thing someone wanted to bang out in a weekend.
In the very long term, software will become a commodity, as you mentioned. Process and workflow may move into JIT delivery for the need at hand, in theory the data layer will be comprehensive and clean and the days of clicking around a bunch of stuff to fulfill process needs will move into a lower latency activity like...talking to your agent.
I saw a quote today by Brian Eno(1995) that said: "So the question becomes not whether you can do it or not, because any drudge can do it if they're prepared to sit in front of the computer for a few days; the question then is: of all the things you can now do, which do you choose to do?" and it resonated with me a lot.
> Lots of people have really great ideas but very few actually commit to execution.
This is true when you had to work hard for those ideas. Now you have LLMs. It means more people can sling a lot more crap at walls with fewer barriers to entry.
Good execution doesn't get easier with an LLM.
No, and I agree with the conservative sentiments here. However, putting together a SaaS alternative that frees up money during a crunch, and now with the pet features your boss has always wanted, is potent indeed.
You've hit the nail on the head. Immediate gratification.
AI is like sugar. It tastes nice, gives you energy quickly - what's not to like? The gratification is immediate, and if "today is all that matters" it's brilliant.
The problem with sugar (and AI) is medium term. So sure, that junior dev whipped up the whole framework in ClaudeCode, and it's humming nicely. Junior dev gets credit, and after a couple years moves on somewhere else.
Then something changes. Windows. TLS. Code Signing, whatever. We need to update the program to the change. Just a small tweak. Junior Dev has gone (or is otherwise occupied) so we'll get new-Junior-dev to do it. Is he expected to do the change at the code level? Or at the prompt level? Will ClaudeCode in 2029 be able to maintain ClaudeCode Code from 2026? Or will it want to rewrite everything? Will new-junior-dev have the skillset to prompt as well as first-junior-dev? Was the code good enough that a dev could just "take it over"? Or was it "it works, let's use it" standard?
AI makes everyone look good in the short term. But it worries me for what happens in 5 years, 10 years, and so on. Sugar is great, but you can't live (long term) on sugar. Sometimes you need a proper meal plan.
That is why I said "throw crap".
I think my point is that not everyone is suddenly going to go and start making stuff. There will definitely be an increase, which is a net positive because we can start exploring unknown areas of interest more rapidly, we can fail faster, etc. Fewer barriers to entry will increase competition, lower prices, increase efficiency, and theoretically benefit the consumer.
here's reality before claude:
- nearly every enterprise IT project is a failure anyway
- "can i do this for free?" savvy people write "thing i don't want to pay for github".
- ??? "stupid smelly nerds!" (https://www.reddit.com/r/github/comments/1at9br4)
okay, what was the actual obstacle? it's really simple: in order to use something FREE, you had to touch GITHUB, which meant GIT. and people hate git.
today, with LLMs:
- "can i do this for free?"
- LLM dutifully does the needful, using projects it finds and code it learned from github, and doing the prosaic tasks of launching them for you, whatever that means.
people are getting way up into their heads about what matters, psychosocial and management and whatever bs. chatgpt is FREE. it will fix your problems for FREE. people will put up with ANYTHING for FREE.
the real innovation is laundering all that inaccessible, pre-existing solution space into a format that doesn't require transiting git and giving it away for free.
don't believe me? all of the most profitable SaaS businesses in technology are the packaging, deployment and customization of pre-existing open source free software, whether it is linux, kvm, postgres, etc. they are factories to turn stuff that is inaccessible because it is in GIT, which SUCKS - that is the hard part for people to wrap their minds around, that GIT sucks - into websites you can pay for. now LLMs do that.
Exactly. Once the market tastes like lemonade, everyone will be afraid of trying new apps in the way that they are afraid to accept phone calls from unknown numbers now.
You will trade initial development budget for advertising budget, trying to position your product in proximity with people who are known quantities.
Tbh I think you’re fundamentally misunderstanding the issue (or I am).
It’s not about some single dude disrupting the saas market. It’s about largish companies who already have internal dev teams, slowly weening their company off these ginormous one size fits all saas products and building local, tailored solutions.
It’s death by a thousand cuts from the erosion of their highest paying customers.
I feel the market forces kinda point the other way, though, since the customization of the SaaS is also cheapening, but faster and more targeted than these internal teams. Over time I believe that’ll lead to more, not less, SaaS consolidation.
Let’s put the cost of code production at 0: regulatory compliance with payment processing laws or industry oversight is a recurring job that’s common for the whole industry when it changes. SaaS companies have hundreds of customers to attend, these become first class business functions. New demands won’t be in training data for LLMs, so someone needs to be doing this. SaaS has the funds and customer base to have dedicated experts at these functions, but it’s dead capital and nigh-impossible hiring in a tiny talent pool for the rest of the market… the delta to get Salesforce or SharePoint not to be total ass and fully customized is orders of magnitude smaller than detailing those foundations, and as people who sharecrop on platforms like those know, the devil is always in the details. Those internal teams just aren’t positioned to juggle both sides of that coin, they can’t be experts, mistakes can be existential, and the liability picture is so very ugly… coding is the least of it.
Into this, MBAs are not static. It’s not gonna take more than a few “vibe coding ate our CRM data” high profile snafus, or industry think pieces to map out why customization is faster/better/smarter, to get clear business dogma around this. A witty turn of phrase about focusing on your actual business.
I think ‘no one ever got fired for hiring IBM’ x 5 is on the horizon, and the evil marketers at Salesforce, MS, and the rest are gonna work hard to grow their piece of the pie. They have LLMs too, only with better models and unlimited tokens. And our executives will be checking directly with their LLMs about how to invest (the consultants, journalists, fanboys, and social media bots too…).
My current boss is an ex CTO of IBM, and tbh he's proof that more people should have been fired for buying an IBM.
Unrelated to the convo you make some very valid points. I just absolutely detest that saying xD
The history of software has been that once it becomes cheap enough for teams to flood the market with “existing product” + x feature for y users. The market consolidates around a leader who does all features for all customers.
I’d bet that we skip SaaS entirely and go to Anthropic directly. This means the ai has to understand that there are different users with conflicting requirements and that we all need the same exact copy of the burn rate report.
It's not a market for lemons. We can share info about the lemons and all choose to use the good ones. There's no information asymmetry.
> We can share info about the lemons
That might turn out to be less than reliable over time, as bots are already screwing up systems with fake information and it's probably going to get worse.
I don't disagree with that, but the market for lemons still doesn't really fit.
If I remember my econ class correctly it uses used cars as an example. If you're neighbor bought a used Toyota and tells you about it being a great purchase, you can't go out and buy another used Toyota and expect it to also be in great condition. Every car is a gamble.
But if you use something like Hubspot and tell your neighbor it's really good/bad you can expect to receive the same Hubspot service they did.
If you can gin these things up in a weekend then why would you bother with a monthly subscription model for software? The only valuable part is the specification and possibly the hardware to run it. If I were a CTO trying to save money I might pay for the labor to develop good specs, but I would prioritize getting out from under software companies with a rent seeking models and 80 to 90% margins
> If I were a CTO trying to save money
A CTOs job isn't to save money but to spend money effectively. Saving money by increasing risk is not neccesarily a prudent move.
Even if AI were erased today, most SaaS companies wouldn’t exist in 5-10 years. Doing business with a small tech company that could run out of borrowed money or sell to someone who will destroy the product or just arbitrarily change the terms of service tomorrow is a liability. That’s assuming the hypothetical tech spend isn’t just eroding margin anyway.
True, but that's one of the reason established SaaS can charge an arm and a leg. Not every SaaS is a start up.
On prudent choices: one thing I'm surprised about is that LLMs are showing me libraries and tools that I'd not found via search.
A boring one from today was about select, datalist or some custome element (which LLM can prototype) or some JS libs. Good breakdown; links to playgrounds, rough mocks so team could kick tires. It raises points the team had and had counterpoint to help drive decisions.
it depends a lot on the application, I think, though certainly you can point to cloud services like Cloudflare or whatever Burger King was using to track how many times a clerk said "You Rule" (while capturing all customer audio data, which was then stolen by low-effort attackers) as high-risk; just because you don't feel the safety risks of outsourcing data to a black box on the cloud doesn't mean they don't exist, it just means you get to neglect them. when I headed IT at an SMB, I was given a lot of leeway, and our department had its own budget, so cutting out SaaS was a high priority so we could do more. if I were heading today with LLMs' present competency, I would have replaced much more, up to and including Salesforce which was draining the heck out of our budget despite us not doing anything technically interesting with it.
$40/head/year (including employees no longer with company) for a call metrics suite is low-stakes and relatively easy to replace what we want out of it, and this is an example of something we did replace with a $0 solution with my own abysmal-at-the-time coding skills. ~nobody's about to replace Microsoft suite, though (a couple replacements before me, they earnestly tried; there were still some laptops with OpenOffice on it; I admire them, but I'm not dealing with our sales team trying to figure out what an ODF is).
I love this "petty kingdom" budget model, by the way, as someone whose work personality could be described as "cheap analyst." I'm paying $40/month per head for Software X in your department, and I have an inferior replacement for $0/month/head which meets specs and which you can't quantify productivity loss for (essentially, it just looks ugly and feels bad). I can therefor cut that out of my budget entirely while meeting my obligations, and if *you* really want the decadent solution, *your* department can bear that cost. Either way, I get plenty more money to basically not have to be a dick (like charging careless employees for broken/stolen equipment, or getting an above-expectations solution for ADA employees); and sometimes, maybe some antennas show up on the roof which would be difficult to justify cost for if asked, but I'm way under-budget so nobody would.
in the 90-ies anyone could easily prototype with tools like Access (and all the other "4GL" tools which were similarly all the rage back then). That still didn't preclude companies from buying their major software from software vendors instead of doing it themselves.
In some sense having customer able to prototype what they want is a good thing. I did it myself as i was at the time on that side, and having a quick-whip-it tool was a good thing to quickly get some feature that was missing in the major software before that major software would add it (if at all). (And if one remembers for example Crystal Reports - while for "reports", it and the likes were in many senses such quick-whip-it tools for a lot of such customization that was doable by the customer.)
So, after initial aftershock - "Ahhhh, we don't need software companies anymore!" - we'll get to the state with software companies still doing their thing just with a lot of AI as specialization is one of the main thing in modern economy and AI becomes most powerful tools of the trade. (and various AI components themselves will be part of software delivery, like say a very fine-tuned model (hosted or on-premise) specific to the customer and software - Clippy on steroids)
(Of course some companies wouldn't survive the transition just like some companies didn't survive the transitions to client/server, cloud, etc. while some new companies will emerge like Anthropic has today or Borland had at the time)
Access is not as dead as you might hope. The long tail of internal tools written with Access continues to shamble along. I had to figure out how to dump MDB files on Windows last year for just this reason. As an industry I think we often fail to grasp how much outsider art there is, in the form of internal departmental tools.
LLM coding is going to create a cambrian explosion of these tools. It’s going to be very interesting to see the remnants of this wave 30 years down the line.
One of the key questions here - will LLM coding decrease the proliferation of app-specific Excel files (by for example accelerating and simplifying Excel-to-webapp conversion) or would result in an opposite outcome by making feasible managing even orders of magnitude more of those disparate Excel files :)
I wouldn’t bet against cramming more and more business processes into Excel. The guy who was copying cells from one workbook to another yesterday, tomorrow can have a single mega-workbook with all the macros more or less deconflicted.
It means the same: random lottery of mass, with everuone else failing.
American capitalism hides the depressing fact that rarely does the best succees.
AAI momentum is parallel to just buying lottery tickets and doing so with the belief that you know the real odds, so one can overwhelm with quantity of tickets.
I sure hope I never have to hunt down any GTM options myself soon and I can tell the AI to do what it should be doing. However AI adoption may be getting slowed down by profit motives because what Google should have already been doing is letting me git clone the entirety of GTM with all its configs to a local folder so I can treat it like code because it is code. The difficulty with AI adoption will be to make all products be like this so they can interact on a code level instead of me having to press buttons in different UIs to make thing shappen. E.g cloudflare should be letting me git clone its entire config, everything I did in the dashboard, to my folder too.
> bootstrap a one man enterprise without actually doing any kind of social interaction will run into a wall.
But you are not limited to only using LLM for coding.
I agree that marketing and sales is as important as product and technology, but they are not necessarily safe.
In any usable product making a product is like 20% or less. Enter compliance, security, payments and a million other things.
Even if you can build it in a day B2B SaaS will continue to prosper because they sell peace of mind, reliability and compliance. Not features.
Also due to economy of scale it will always be cheaper to buy something from a vendor that sells it to many clients than to DIY it.
Yep. Wait until they realize one of their vibe coded apps validates everything client side and their entire DB is open to the world.
Like that never happened with non-vibecoded stuff.
Of course it has, but it's more likely to happen when nobody is paying attention.
There's a difference between someone breaking into your house to steal your valuables, and you getting robbed after leaving your valuables on your lawn with a sign saying "Look At My Valuables"
Startups of the last decade were routinely doing this. Building their fresh ideas on the newfangled databse without bothering to investigste how to secure it in any way against malicious actors.
Yep. It's a funny thing.
You build a Twitter. Profiles have posts, posts can have images, etc. It's very easy to model the database.
But then how do you make money with it? Now you need to build a separate system for advertising? Or do you want to sell subscriptions? Which means you need to build a separate system to handle payments. This is usually the big one, because when you handle money, what happens if there is a bug and you charge someone without delivering anything? How do you prevent fraud? How do you handle disputes?
Someone posted something illegal. What do you do in this situation? Do you call the police? The FBI? What kind of data do you give the authorities? How much data SHOULD you have been logging in the first place in case something like this happens?
One user doesn't like you so he bought a botnet to DDoS your website. How do you handle this? Are they mass posting? Mass creating accounts? Is it possible for them to exhaust all the usernames possible and then nobody can create an account anymore?
Your website is online but if the server blows up you'll lose all the data in the database. You need backups. You need a system to ensure the backups are actually working. But then some guy from the UK said he wants his posts all deleted. What are you going to do now, because his posts are also in the backups, and you don't want to touch those.
Trolls are posting things against the ToS. Who handles these things? Shadowban? So there needs to be a shadowban system? Moderators? So there needs to be a moderator-only section of the website? Should this be integrated with the main website or not?
Then you look at this horrendous mess of 6 paragraphs and you think back about the first paragraph that already did everything you wanted from Twitter. All these other systems, most of the work, and all you actually wanted was the first paragraph.
All those things are true. It still doesn’t sound like 1000+ engineers at 350k/yr.
What actually happens in a startup is you encounter these problems one at a time as they arise.
Twitter wasn't built by 1000 engineers at 350k/yr.
It had to hire them later on. Because when there are users - you need support, take out fires etc.
And this exact thing will happen with any homebrewed SaaS.
You either run a business or play tech company making your own saas instead of focusing on your business.
Sure you can do both in very rare cases - if you are SpaceX or similar, otherwise you are shooting yourself in the foot.
No. They hired 1000 people to help them justify funding rounds.
Even then, startups prioritize growth over efficiency. So maybe 100 people would have been fine but 1000 gets them a 5% growth improvement in growth.
Startups have no users and no data to start with, and if they fuck up security, well, they just fail sooner than expected.
Once you get past a certain size, you have very different sorts of problems. Any idiot can vibe code a facebook lookalike, but the real one has to handle hundreds of millions of users and posts while being a target for state actors.
TLDR; yes you do need that many
> yes you do need that many
You absolutely do not. what do you think about the website we are using right now! It has half of the problems listed above.
> facebook
Your work project doesn’t have a billion users.
We were talking about what it would take to fix the technical problems resulting from taking a working program to something people use.
> Any idiot can vibe code
I didn’t say that either.
How is it the HN opinion that it’s impossible to make a web application a lot of people use?
These are two different problems.
AI has solved the 'coding' part. The business is still very human because they are the ones buying, for now.
It's still possible. Have the product and hire other people for these things.
Use stripe, cloudflare, whatever the legal equivalent of these stuff, S3.
Yes they might take most potential profit, but you'll also not have a huge payroll.
> We are certainly closer now to being able to prototype and go to market faster with a product.
Prototype maybe. Go to market maybe not so. It's giving false hope. You're just taking more shortcuts with prototyping.