The future of 10x engineering
This tweet from Andrej Karpathy is as good as advertised. Here's one of the many good questions he asks:
What happens to the "10X engineer" - the ratio of productivity between the mean and the max engineer? It's quite possible that this grows a lot.
Some notes:
- Karpathy's question assumes that 10x engineers exist. This is an old debate (and, for a while, a battle ground in the culture wars). I do still think that some engineers are at least 10x as productive as median ones, for any reasonable sense of "productive."
- I can't see a good argument for generative AI not making the best engineers better, as measured by the ratio to the median engineer. If that ends up not being true, it's probably because "median engineers," as we now think of them, are no longer in the profession.
- What if we ask about the productivity ratio between the 99th-percentile engineer and the 95th- or 98th-percentile one (instead of the 50th-percentile one)? There's an argument to be made that all excellent engineers will be able to use generative AI to avoid big mistakes, correct design problems, achieve excellent observability, and so on, and that gap (from 95th- to 99th-percentile) won't grow as much, or might even shrink.
- But if you replace "99th" with "99.9th" above, I'm back to thinking the gap will get much larger.
- Relatedly: part of my argument for the existence of 10x engineering is how much time gets spent on avoidable, entrenched problems. It's easy for 20% of a team's effort to be spent, e.g., maintaining something that shouldn't exist or debugging problems that only exist because a single abstraction boundary wasn't created. There's plenty of hype about all aspects of AI, but "we can build so quickly!" gets so much attention that "we can prototype highly maintainable designs!" and "we can reliably avoid certain common mistakes in defining data structures!" are comparatively overlooked.
- Speaking of "the best engineers" or "10x engineers" can lead one to think that top engineers now are the same as future top engineers. I think this is a trap: some crucial skills will be the same, but others will be different. For example, I suspect that sheer stamina and work ethic will become even more important to one's output.