[rybosworld]

This is spot on.

It's essentially like saying "What I do in particular, is much too difficult for an AI to ever replicate." It is always in part, humble bragging.

I think some developers like to pretend that they are exclusively solving problems that have never been solved before. Which sure, the LLM architecture in particular might never be better than a person for the novel class of problem.

But the reality is, an extremely high percentage of all problems (and by reduction, the lines of code that build that solution) are not novel. I would guesstimate that less than 1 out of 10,000 developers are solving truly novel problems with any regularity. And those folks tend to work at places like Google Brain.

That's relevant because LLM's can likely scale forever in terms of solving the already solved.

[whtsthmttrmn]

> I would guesstimate that less than 1 out of 10,000 developers are solving truly novel problems with any regularity. And those folks tend to work at places like Google Brain.

Looks like the virtue signalling is done on both sides of the AI fence.

[rybosworld]

Not sure how you equate that statement with virtue signaling.

This is just a natural consequence of the ever growing repository of solved problems.

For example, consider that sorting of lists is agreed upon as a solved problem. Sure you could re-discover quick sort on your own, but that's not novel.

[raincole]

How is it humble bragging when there isn't "humble" part?

[rybosworld]

Fair point