Quote  ·  Added 18 September 2026

Why would I look at code? It’s like assembly, like a compiled artifact.

Andrej KarpathyLet's build GPT: from scratch, in code, spelled out, YouTube, 2023

Filed undercode·assembly·abstraction

Reading

AI generated

When experts stop looking under the hood

There's something fascinating buried in this comment about why we resist looking under the hood at things we use daily. Karpathy, who spent years training AI models, is basically saying that once something gets compiled down to machine-readable form, it stops being knowable in any human way. And he's right—but also, that's kind of the point where most of us check out entirely.

The real tension here is that we've built systems so complex that even their creators can reasonably say "why bother?" Your phone's operating system, your bank's algorithms, the recommendation engine deciding what you watch—they're all "compiled artifacts" now. We're trained to trust them blindly or distrust them blindly, but actually reading them feels pointless. So we don't. We just use them and hope.

But there's a hidden cost to this surrender. When we stop looking, we stop asking whether these systems are doing what we actually want them to do. We can't spot where a bias snuck in, or where something is broken but still running. The comment reveals something uncomfortable: at a certain scale, even experts give up on understanding. That might be practical for engineers, but for the rest of us trying to make sense of the world we're living in, it's worth asking occasionally—what are we blindly trusting today?

The author

Andrej Karpathy

b. 1986

Andrej Karpathy is a computer scientist and entrepreneur known for his work in artificial intelligence and deep learning. He served as the Director of AI at Tesla, where he led the development of computer vision and autonomous driving technologies. Prior to Tesla, he was a research scientist at OpenAI and gained prominence for his contributions to neural networks and for creating the popular course "CS231n: Convolutional Neural Networks for Visual Recognition" at Stanford University.

Comments

Sign in