Lately I've watched something I genuinely enjoy seeing: people with no background in software discovering a new hobby, building "digital products." I use quotes on purpose. These are driven, curious people who unlocked a new ability to act on ideas they always had but could never execute without depending on a "expensive team of programmers."
I don't blame them for thinking that way. It makes sense. A lot of people in our industry do the work badly, with no judgment, or promise something that isn't real. So the appeal of skipping us entirely is easy to understand.
But here's the part nobody wants to say out loud: AI doesn't close the gap between what you want to sell and how to actually pull it off. It compressed the easy part of building. The hard part is exactly where it was before.
What AI actually compressed
Be honest about what changed. AI made it trivial to go from "idea in my head" to "something on a screen." What used to take weeks now takes days. That's real, and it's great.
But notice what got cheaper: typing the code, wiring up a UI, generating an .html file (tools like Claude or Codex often just hand you one). That was never the expensive part of a product. It was always the most mechanical part.
The expensive part, the part that decides whether you have a business, didn't move at all.
The gap AI doesn't solve
Most people building this way are quietly looking for a "Can I please get money?" button. They want the output without the part that earns it. And that part is a set of questions a model can't answer for you:
- Why would anyone pay for this?
- What real problem are you solving, and for whom?
- What value does this create that they can't get elsewhere?
A prototype that looks finished gives you a dangerous feeling that the thinking is done. It isn't. A working screen is not a working business. The screen was the cheap part.
The trap of an agreeable model
There's a second, sneakier problem. LLMs are built to be relentlessly encouraging, even about ideas that don't hold up. Ask one if your idea is good and it will find a way to tell you it's brilliant.
So a hopeful, well-meaning person can end up in a loop. They don't have full clarity on what they're selling or how to monetize it, and now, on top of that, they've also quietly become:
- a UX/UI designer deciding layouts and flows they've never studied,
- an HTML developer maintaining code they can't fully read,
- and a manual QA team of one, testing everything by hand.
The tool removed the cost of producing. It did nothing about the cost of deciding, and it added three jobs you didn't sign up for.
What happens after the first sale?
Let's say it works. You got it running, configured containers, sorted out DNS, and somehow deployed to production. Genuinely impressive. Now the real questions start:
- What happens when you get your first real customers and the system has to scale even a little?
- What happens when there's a security gap, or a dependency that must be updated?
- What happens when you run out of tokens and have to keep paying the extra quota just to make changes?
This is the gap. It's not glamorous and it doesn't demo well, but it's where products either become businesses or quietly fall apart. None of it shows up in the first exciting week. All of it shows up later, usually at the worst possible moment.
Where AI genuinely shines
I'm not anti-AI, far from it. These are some of the best tools we've ever had, for the right job:
- Building prototypes fast, to make something tangible.
- Communicating ideas more clearly, once you've actually thought them through (in that order, not the reverse).
- Validating hypotheses before you invest real money in a real product.
- Acting as a guide to map a rough roadmap of everything you should know before you commit to building.
Used this way, AI is a discovery and validation engine. It helps you reach the decision faster. It just can't make the decision for you, and it can't carry what comes after it.
The mistake isn't using AI. The mistake is treating the prototype as the finish line instead of the starting line.
Will AI take software jobs?
This whole thing answers a question we can't avoid: is AI going to put people in tech out of work? My honest answer is yes and no.
It will push out the people who copy and paste without adding value (if those people were ever really doing the work). And it will amplify the people who genuinely understand the problem, the system, and the business underneath it.
The differentiator was never typing speed. It's judgment, the ability to know what to build, what not to build, and what breaks when it meets the real world. AI makes that judgment more valuable, not less.
The bottom line
AI gave everyone a head start. That's a good thing. But a head start on the easy part can hide how much of the hard part is still ahead of you.
If you've built a prototype and people are starting to react to it, you're past the hardest creative step and right at the start of the hardest product step. That's a great place to be, as long as you treat it as the beginning.
