Intelligence is becoming abundant.

Intelligence is getting cheaper. Things that once required years of expertise or entire teams are becoming accessible to almost anyone. It is worth thinking about what happens when intelligence itself is no longer scarce.

Access to the best model isn't enough, what matters is what to do with it, what data it has access to, and who it can be trusted to work for.

It’s not enough for AI systems to answer questions. They need to be able to make decisions, use tools, and get work done independently. But they also need to know when not to. And it needs to be obvious who’s responsible when they make a mistake.

Intelligence should be accessible to as many people as possible. There should be multiple models to choose from, and open models that anyone can use. There should be no lock-in, and claims around safety and performance should be backed up by rigorous testing.

AI systems should be small, fast, and cheap by default, and easily scale up when needed. They should not be more complex than necessary. And the people who make them possible should benefit from them.

A more powerful model is useless if it’s not deployed effectively to create real world impact.

Welcome to nullvelocity