When the Barrier Is Zero
*Published: 2026-03-06 | ID: 60 | URL: https://pgc.ai/when-the-barrier-is-zero/*
Saw someone say: “Whenever I see an article that’s super long, has clear logic, comprehensive citations, not a single typo, and even lists all references at the end, I just close it. Because if you give me the title, I can write the same thing in GPT.”
I thought about that for a while. It’s kind of true for software now too.
With one screen capture, you can describe your app to an AI agent and it’ll rebuild it. Tailored to your needs. Maybe even better. The barrier to copy most software is approaching zero. Which means for most software, the value is approaching zero too.
It’s kind of like most software became pseudo open source without anyone deciding to open source it.
What survives are the standards that already exist. The existing frameworks. That’s probably why every AI startup’s CSS ends up in that same purple-blue gradient. Markdown is the same story. I hated markdown when it first came out. Seemed overcomplicated for no reason. Now it’s the common tongue for everything because it got there first and everything built on top of it. Same reason rail gauges are the width they are. Legacy of carriage wheels. Legacy of what was already there when everyone needed to interoperate.
I was talking with my business partner about this and we noticed something. A lot of the recent AI apps that actually went viral seem to be made by developers with years behind them. My guess is there’s a gap forming. The kind of experience you used to build over years of software jobs, filling in gaps, shipping things, doing R&D under someone senior, that path is closing. There are fewer entry-level jobs to fall into. Fewer chances to make real mistakes on real products and slowly build up the judgment to know what’s actually worth building. AI can write code now but it can’t give you the years of knowing what fails and why. That’s harder and harder to accumulate now.
And the new blood that does break through seems to be less median-developer and more… exceptional outlier. Super smart, figured it out themselves somehow. There might not be a middle rung anymore. Not just “becoming a developer is harder” but the whole ladder is getting pulled up.
The statistics keep saying demand for programmers is increasing since AI emerged. I think that number holds for now but I don’t think it holds for long. Kurzweil’s been predicting AGI by 2029 since 2005, and he reaffirmed it last year. That puts us right in the middle of the run-up. Software won’t disappear, there will probably be more software in the world than ever, but most of it will be cheap and custom-made for one person or one use case. A lot of software will essentially have an audience of one: its own developer.
So the real question becomes what information inside software is actually hard to duplicate.
I think there are two things that hold.
One is that it knows you. Not you generically. You specifically. Your context, your history, your preferences, the things you don’t have to explain because the software already learned them. That’s not copyable from a screenshot.
The other is that it expands what you can do. Not just automates tasks but actually increases the amplitude of what you’re capable of. Makes you more of yourself, not just faster.
Ask yourself honestly: if an AI agent rebuilt your app in an hour, pixel perfect, with all the same features, would your users switch? If yes, there’s nothing there. If no, what is it that makes them stay?
We asked ourselves the same question about what we’re building. I think we know the answer. But you don’t really find out until someone actually tries.