Sattva

Own your intelligence

Wooden trays of beans and seeds in many colors, from dark red and black to white and speckled.
Photo by wei on Unsplash

In a recent Request for Startups, Jon Xu of Y Combinator wrote (emphasis mine):

Today, nearly every company building with AI rents its intelligence. You send tokens to a frontier lab, get answers back, and pay the bill. That was an easy trade when the best models were years ahead of everything else, and nothing you did with your own data could close the gap.

That has changed. Open-weight models are now competitive with the top frontier models on meaningful benchmarks. And as AI moves to the center of how businesses run, companies have become more aware of their dependency on model rental.

We think more companies will want to own the models they rely on: improve them on their own data, run them where they choose, and decide for themselves when they change. This matters most in regulated industries, where strict data governance and locked-down infrastructure make renting even more expensive for the performance they get.

But there’s still a big gap between downloading an open model and owning intelligence that improves a business. Someone has to turn workflows into evals, expertise into training data, production failures into improvements, and carry all of it forward with each new base model.

Companies will keep renting frontier intelligence for their hardest problems, but we think many will also own specialized models that encode how their business works and that become more valuable as the business learns.

Most nonprofits don’t need to own their models yet. But they will certainly want to choose which models to deploy in certain cases: for instance, an offline model for sensitive data, or a typesafe model like Jev to make work like data classification orders of magnitude cheaper.

I would encourage nonprofits to take Xu’s title literally. The hard-earned institutional knowledge that a nonprofit accrues over years or decades is its DNA: a coaching rubric revised over fifteen years of classroom visits, or the intake questions a legal aid clinic learned to ask through longitudinal studies. Look again at what Xu says is hard to own: workflows, expertise, lessons from failure. None of that is the model itself.

I think we’ll see nonprofits be increasingly thoughtful about how they nurture, protect, and ultimately distribute their institutional knowledge. The last of these is where they differ from the companies YC is looking to back. A venture-backed startup owns its intelligence so it can keep it. A nonprofit exists to give it away, and the work ahead is doing that in ways that increase access without risking commoditization or slop.