Tribalism in AI

You'd think we'd just use the best models. But tribalism is baked into us—especially as Americans. We're tribal about sports, politics, brands, identities. It's just how we're wired. So when AI models show up, we're not doing anything new. We're picking teams, defending them, building ourselves around them.

And then there's the switching cost on top of that. The workflows you've learned, the muscle memory, the way each model thinks—rebuilding all of that is real friction. So tribalism in AI isn't just preference. It's the combination of who we are as people, plus the actual cost of jumping ship. They reinforce each other.

But here's what's interesting: we don't have to accept that. If we build products and workflows that are model-agnostic, we sidestep it. The concepts are the same across providers. It just takes first-principles thinking about how LLMs actually work, rather than getting locked into what each provider's tooling makes easy.

Design around the fundamentals, not the tribalism. That's where the opportunity is.