Everything Is a Skill Issue

Everything is a skill issue.

Just listened to @saranormous interview @karpathy on No Priors and one line hit like a truck.

The models already have the raw capability.

The jagged intelligence, the PhD moments mixed with 10-year-old reasoning, the “almost there” outputs, it’s not a model limit anymore.

It’s almost always that you didn’t give it good enough instructions, context, memory, or orchestration.

Karpathy put it exactly like this: “Even if they don’t work, I think to a large extent you feel like it’s a skill issue. It’s not that the capability is not there. It’s that you just haven’t given it good enough instructions… I don’t have a nice enough memory tool… So it all kind of feels like skill issue when it doesn’t work.”

And later: “I’m the binding constraint… Yeah, it’s a skill issue. Which is very empowering because you could be getting better.”

You’re not bottlenecked by intelligence.

You’re bottlenecked by your ability to direct it.

The new meta-skill isn’t writing code.

It’s becoming a world-class AI director: persistent agents, parallel “claws”, AutoResearch loops, knowing exactly which few high-signal tokens actually matter.

Everything else, from debugging to research to building products, collapses into one question: “Did I level up my directing skill enough?”

If something feels stuck, broken, or “not quite right” with AI today…

99% chance it’s still a skill issue.

The era of “the model can’t do that” is ending.

The era of “I’m not good enough at telling it what to do” has begun.