We back smart people who use their intelligence to create intellectual property that supercharges our businesses. Proprietary information or knowledge, even if only for a few short years, allows these companies to build and scale.

But the ground rules are changing because of AI.

In Phase 1, AI builds efficient systems: more, cheaper, faster, better. The value of content starts trending toward zero. The value of labor also trends toward zero. Productivity goes through the roof.

In Phase 2, AI builds AI: efficiency applied to AI engineering itself. In a year or two, most next-generation foundational models (Grok, OpenAI, DeepSeek, Manus, etc.) will be significantly developed by their current versions. Pretty soon, full recursion begins in AI construction. The value of knowledge itself tends toward zero.

What? Yes, think about it: it will soon be hard for governments to issue patents when we can generate significant novelty with five minutes of compute. The economic value of knowledge starts to move toward zero.

Phases 1 and 2 will combine to commoditize intelligence. All investments in AI tech—and often in layers built on top of foundational models—will become cost-plus, utility-like businesses. Even intelligence itself will be commoditized.

For example, in Phase 1, imagine you invest $50 million to build the world’s best AI lawyer. When Phase 2 arrives, a new competitor could simply instruct an AI to clone your AI lawyer, creating one that’s nearly identical—or perhaps even better—than your carefully crafted version.

Another example: Midjourney generated $50 million in its first year and $200 million in 2023. But pundits are questioning its future as its functions are absorbed into foundational models. What it does is being commoditized.

Now what? It took me some time mulling over this question: Can we still build sustainably differentiated companies? And if so, what would be the nature of that sustainability?

‍

‍

I have one answer so far. I suspect it might be the only answer (though friends disagree): Proprietary Information on People (PIoP).

Why? PIoP enables you to deliver superior products and services tailored to those people.

You can obtain PIoP in one of two ways:

1. Spy on people (accumulate observational data).
2. Earn people’s trust, and they willingly give you the information.

Option 1 yields inferior-quality information and isn’t a sustainable model if people have a choice.

So, I think option 2—building trust relationships with your audience—is the only way to create sustainably defensible businesses in the long haul.

This leads to Phase 3: AI builds trust relationships. If we conduct a proper game-theoretic analysis of the conditions, we might realize this has to happen.

There are many details to flesh out over time. I suspect your imaginative mind will conjure countless visions of the future from here.

I started this journey with a simple question: How do I effectively invest in a world with AI? I didn’t expect to come full circle and conclude that we must return to building businesses that foster trust relationships. This isn’t just because it’s what I want or hope for (though I do). It will happen because the conditions of commoditized intelligence will drive our business models in a free market toward that direction. Unsurprising, perhaps. It’s pushing our businesses to prioritize what matters most to us all: trust relationships.

‍