If They Ask About OpenAI, They Did Not Read Your Deck

Every era has its boogeyman question. Ours was Theranos. Here is how to know when it actually matters, and when someone simply has not done the work.

Every founder raising in AI right now is being coached to have an answer ready for one question. What happens when OpenAI or Anthropic just builds this?

We think most founders are preparing for the wrong thing.

This week made the point better than we could. Apple, the most valuable company in the world, effectively conceded it could not win the model race on its own. It shipped its new AI built on foundation models it developed with Google, and bet its future on the experience layer instead. If leaning on a frontier lab does not doom Apple, it should tell you something about your own odds.

We have heard a version of this question in every season of our journey. Paul Graham made the same point this week. “What if the model companies do this” is just the new “what if Google does this.” He is right. There is a question like this in every era.

Every Era Has Its Boogeyman Question

The form changes. The function does not. In 2000 it was “isn’t this just Pets.com?” A decade later it was “what if Google or Facebook builds this?” After 2019 it was “is this another WeWork?” After FTX it was “is this going to blow up like crypto?” Today it is the labs.

It is the lazy pattern match an investor reaches for when something makes them skeptical and they do not yet know why. It feels like diligence. It is not.

Mine Was Theranos

When Matt and I were raising for our diagnostics company, mine was Theranos. We were building animal diagnostics in 2021, and every few meetings someone would lean back and ask, “so, are you like Theranos?”

I would roll my eyes. Theranos claimed to run more than thirty diagnostics out of a single mysterious box. We were doing one diagnostic. We were applying computer vision to the narrow problem of identifying cells. The only thing the two companies shared was a loose association with blood.

Anyone who had actually read what we were building would never have asked. That was the tell.

The Question Is Usually a Tell

This is the part founders get backwards. The boogeyman question is what someone asks instead of doing the work. The honest answer to defensibility was never a clever line. It lives in the details. How you go to market. The proprietary data you accumulate as you serve customers. The reputation you compound over time.

Someone asking the lazy version has not read your deck. They have not researched your company. Sometimes they have not really been listening on the call. If they had, the question would have answered itself.

So do not contort your pitch into a soundbite built to defeat “why can’t OpenAI just do this?” You will spend your best energy on the people who were never going to dig in. Know your real moat cold, and bring the conversation back to the specifics. The investors worth having will follow you there.

Know When It Is a Fair Question

Now be honest with yourself, because there is a version of this question that is completely fair.

If you are building a frontier lab, competing directly with OpenAI and Anthropic to train the most capable general model, then “what if the labs do this” is not lazy at all. It is the question, and you had better be able to answer it. But that is a different point, and a different company.

Most startups are not on that path. If your edge is a remarkable customer experience and the data you acquire as you actually service customers, the labs are not on your path, and they are not going to be. They are building the broad magic. You are building the specific, compounding thing in front of a customer who trusts you. Theranos was a box that promised everything. We were a narrow tool that did one real thing. That is the same gap that separates a frontier lab from almost every AI startup being asked about one today.

Final Thought

The boogeyman changes every cycle. Pets.com, Google, Theranos, WeWork, FTX, the labs. There will be a new one the moment this one gets boring. The discipline does not change. Build the narrow, real thing only you are positioned to build. Compound the data and the trust you earn serving customers. Answer the fear underneath the question, not the costume it is wearing this year.

And when someone leans back and reaches for this era’s version of that question, let it tell you how closely they were actually listening.