Is SaaS dead? Wrong question
What's dying is generic software sold to people who can now build it themselves. Software tied to a specific, painful problem is selling fine.
2 min read · Updated 1 Nov 2026
The short answer
SaaS isn't dead, but one kind of it is in trouble: thin, generic tools that customers or AI can now copy in an afternoon. Software that solves a specific, costly problem for a customer you can reach, with something hard to copy around it, still sells. A first-time founder should choose a customer first and a business model second.
“SaaS is dead” has been one of the loudest arguments among founders this year. It’s worth separating what’s actually happening from the headline, because the answer changes what you should build.
What’s actually dying
Three things are true at once.
Generic tools are easy to copy. In July 2026 a post titled “Micro-SaaS Is Dead” reached the front page of Hacker News. Its author had built a tool in hours that he said outperformed paid products, and concluded: “The code was never the moat” (Zeli).
Thin AI products are under pressure. In February 2026, Google’s Darren Mowry, who leads its startup organisation, told TechCrunch that for thin wrappers around AI models “the industry doesn’t have a lot of patience for that anymore”, and advised founders to “stay out of the aggregator business” (TechCrunch).
Distribution has become the hard part. In June, Pieter Levels described a room of indie hackers with impressive AI pipelines where “almost none of them have money or traffic” (levels.io).
What still works
Read those three together and the pattern is clear. Building got cheap, so building stopped being the advantage. A founder whose studio had built more than 15 products put it plainly on Indie Hackers: “Micro SaaS isn’t dead because demand disappeared” (Indie Hackers).
What still sells is software attached to something hard to copy:
- A specific customer with a costly problem. Generic tools compete with everyone. A tool for one kind of business with one expensive problem competes with that business’s current workaround.
- A way to reach them. If you can name the people and get in front of them, you have something a better-funded copy doesn’t.
- Knowledge the software carries. The method, the data, the judgement built in. That’s the “intellectual property” Mowry said users want.
- A service around it. Delivering the result yourself, using tools you built, is often the faster first business, and the HN post argued it’s where durable value now sits.
When everyone can build, taste, distribution, and retention become the game.
What this means if you’re starting out
Don’t choose a business model first. Choose a customer and a problem, then pick whatever model gets the result to them fastest. Very often that’s a service delivered by hand for the first ten customers, with software growing out of what you learn.
The useful question for this week is whether there’s a specific person whose problem is costly enough to pay you, and whether you can reach ten of them.
The Sunday email
Get one of these every Sunday.
One founder problem and one thing to do before Monday.
Questions
Should I still build a SaaS product in 2026?
If you have a specific customer with a costly problem and a way to reach them, yes. If you're picking SaaS because it sounds like a good model, start from a customer instead.
What is an LLM wrapper?
A product that adds a thin interface on top of someone else's AI model. Google's Darren Mowry said in February 2026 that the industry has little patience left for thin wrappers.
What's "service with a software"?
A business where you deliver a result for clients, using tools you built for yourself. It's often the faster first business, and it teaches you what a product would need to do.
Next step
Is your idea worth 30 days? Find out free.
Five questions about your idea and a straight answer. It's week zero of Humble's 30 days, and it costs nothing.