[vlovich123]

If code quality only stops mattering in 400 years (whatever that definition happens to be) then the prediction that it makes is worthless in terms of what you should do today. You use it to argue it’s unimportant deal with it, but if it’s a 400 year payoff you’ve made the wrong bet.

[csallen]

Surely you don't think AI coding technology will be as slow to develop as guns were.

We're obviously talking about 1-10 years here, not 100-1000 years.

[vlovich123]

It’s really hard to predict where exponential progress will freeze. I was reading the other day that the field seems to have stagnated again in terms of no really meaningful ideas to overcome the inherent bottlenecks we’ve hit now in terms of diminishing returns for scaling. I’m not a pessimist or unbridled optimist but I think it’s fundamentally difficult to predict and the law of averages suggests someone will end up crowing about being right

[csallen]

That's true, but even if progress on models freezes TODAY and there is zero progress, not even incremental progress, we'll still have substantial progress in terms of techniques and agents built on top of the models in the next few years. And even if that progress were to stop, it's still significantly advanced today, and there's an incredible gap between current and eventual adoption levels.