# Code is fungible. Trust isn't.

Chris Dixon talks about three exponential forces that drive company growth: Moore's Law, composability, and network effects. He says they're growth drivers, and to some degree he's right, but I think they're something else. They're the only three ways to accumulate the one thing AI can't replicate: trust.

## Moore's Law is the storm

Moore's Law, generously transposed to include doubling AI capability too, is pushing the cost of all software toward zero. It's getting easier to spin up a competitor and easier to migrate off an incumbent, and both of those trends are accelerating. If the only thing protecting your product is that it's annoying to reimplement, that protection is fading fast.

So the question shifts. It's no longer can you build it — everyone will be able to soon. It's do people trust you. And trust, unlike code, gets more valuable as everything else gets noisier. The products people already know, already use, already depend on have an advantage that widens as the noise increases.

These three forces are really three different ways to accumulate that trust, and three different buffers against it fading away.

## Trust is the shelter

### Network effects

When your product gets more valuable as more people use it — the social graph, the data, the relationships inside it — you've built something that doesn't care how cheap the code gets. The fact that 10,000 people trusted you enough to connect their social graph or build on your platform can't be copied, bought, or generated.

The network is the product, and networks don't fork.

That trust barely drains at all.

This is the gold standard, but true network effects are vanishingly rare and ferociously hard to create.

### Composability

Most products are tools, not networks. But if your tool is woven into a customer's workflow — especially where the stuff of business piles up: relationships, notes, processes, data, integrations — you've built a buffer.

Any single customer could leave without much effort, but they probably won't. If it's not broke, don't fix it. You're not irreplaceable, you're just low ROI to replace. That's trust measured differently. It's not love, but the quiet confidence that this works well enough and there are better uses of my team's time than reinventing it.

This kind of trust drains slowly. Slow enough to buy you time to respond if your base starts to erode.

### Discoverability

AI is the panopticon. Every tutorial, blog post, Reddit thread, and youtube video that mentions your product is now training data. It makes LLMs more likely to recommend you the next time someone asks "what should I use for X?"

Usage generates references, references generate recommendations, recommendations bring in more users who write more references. It's the internet's collective memory of people choosing you, and it compounds.

This is also the most accessible of the three to start building. You don't need 10,000 users. You need to be present and referenced in the right places within your niche.

Stop asking "can someone replicate what I've built?" Of course they can.

Instead, ask: how much trust have I built, and how fast will it drain?

Code is fungible. Trust isn't. That's the whole game now.