Production is getting close to free. Anyone can prompt a copy of your product into existence, so software on its own drifts toward a price of zero. Attention did not get cheaper. There are still only so many hours in a day, and everything is fighting for them. That leaves two games worth winning: distribution, getting in front of people at all, and trust, being the one they choose. This chapter is about building a real moat in that world. The public part covers where the defensible value moves to. Your email opens the tactical part: how to use AI on your pitch without it quietly working against you.
When production gets cheap, distribution and brand decide
Here is the shift underneath everything else. When it costs almost nothing to make a thing, far more of that thing gets made. More games, more apps, more content, more of everything, all at once. The number of hours people have to spend on any of it did not move. Same day, same attention, much more competing for it.
So the hard problem quietly changes address. It used to live in production: can we even build this. Now it lives in distribution: will anyone ever find it. Most titles never clear meaningful revenue, and the reason is rarely quality. It is that nobody saw them.
This is not a passing weather report about one funding season. It is a direction, and it holds. When production gets cheap, distribution and brand decide. Build that into your problem slide as a tailwind, not a threat: the flood of new supply is exactly why the thing you do to get found, and the reason people pick you over the next lookalike, is worth more than any feature list.
One more thing an investor is doing in their head while you talk: discounting any long timeline. Nobody knows how fast these tools speed up, and the weakest version of them you will ever work with is the one in your hands. A plan that pays off far out competes against a version someone builds much faster once the tools improve. Short, provable, soon beats grand and distant.
Software is not a moat anymore
So what actually defends you, if the software itself does not. Strip out the parts anyone can copy and look at what is left. The code is copyable. The clever feature is copyable, and faster every month. What stays hard to copy is everything that took real time and real relationships to build.
Four layers hold their value when production goes cheap: the data only your product produces, a community that already gathers around you, a brand people trust, and IP that is legally yours. A competitor can clone your screens. They cannot clone the data your users generated inside them, or fork the community that lives there.
What trends to zero, and what holds
| Copyable, drifts toward zero | Hard to copy, holds its value |
|---|---|
| The code itself | The data only your product produces |
| This month's clever feature | A community that already gathers around you |
| The generated look of the app | A brand people reach for without checking |
| Discovery: finding the right thing for someone | IP: the rights, characters and worlds that are legally yours |
Notice what sits on the losing side: discovery. Helping people find the right thing feels like a moat, and it is a shrinking one, because finding things is exactly what AI does better than you will. Community is the moat that gets stronger as AI gets better, not weaker. Discovery is the one that gets eaten.
Distribution is the hardest thing, and the flywheel test proves it
Distribution is the hardest problem in a world of cheap production, and most decks wave at it in a single line. Getting found. Actually reaching the people who would love the thing, at a cost that works. Building the product is no longer the bottleneck. Getting it in front of a stranger is.
Investors know this, so they have gotten sharp about one slide in particular: the flywheel. A flywheel is a loop where each turn makes the next turn easier, and it only earns the name if it spins out and fetches you the next user. Draw your loop and follow the arrows all the way around. If the last arrow does not land on "and that gets us more users", you do not have a flywheel. You have a diagram of a hope.
When a general model threatens to eat your category
If you build anything on top of AI, one objection is coming, and it is the scariest one on the list: what happens when the big general model just does this itself. It is a fair question, and freezing on it loses the room. The answer is not to claim the big models are bad. They are not, and everyone knows it. The answer is to be precise about where raw generation stops and your product starts.