The Karpathy-Rubin Spectrum
I like to joke that work now sits on a spectrum from Andrej Karpathy to Rick Rubin. Technical ability at one end and taste at the other.
Rubin’s famous 60 Minutes interview: ~40 sec YouTube Clip
Sitting at the Karpathy end means you understand how something works well enough to build it yourself, or close enough. When it breaks you can find the cause and hopefully fix it. It means you can get under the hood and hands on the tools.
At the Rubin end, you know what good looks like. You can spot what’s wrong even when everyone else thinks it’s finished. This kind of taste-making is deciding what to keep, what to change, and (most importantly) what to leave out.
Taste develops through making decisions and living with the results. The feature that was removed or rejected I think can say as much about your ability as the work that was delivered.
People can have both (it’s an AI generated diagram with 2 heads, so let’s not ask too much of it), but I think that’s uncommon. The useful part about this tongue in cheek post is placing yourself (with honestly) based on the work you’re doing so you can find great collaborators. And for different kinds of work you’ll be at different points on the spectrum. Due to the nature of how AI works it is broadl optimizing towards averages or regressing to a mean. It is trying to find the most likely answer, the least risky one, the "most not incorrect" one. I don’t think you want to compete head to head with AI right now. What I mean by that is, since its fundamentally the average of all the world’s knowledge, it generates things that are the most average correct answer. It is winning and will continue to win on breadth and correctness, on average. That’s where we can shine as humans, by going beyond.
So much is changing so fast. Keep your knees bent.
END.