Comment by mplanchard
5 days ago
So? The better these tools get, the easier they will be to get value out of. It seems not unwise to let them stabilize before investing the effort and getting the value out, especially if you’re working in one of the areas/languages where they’re still not as useful.
Learning how to use a tool once is easy, relearning how to use a tool every six months because of the rapid pace of change is a pain.
This isn't responsive to what I wrote. Letting the tools stabilize is one thing, makes perfect sense. "Waiting until the hype cycle dies" is another.
I suspect the hype cycle and the stabilization curves are relatively in-sync. While the tools are constantly changing, there's always a fresh source of hype, and a fresh variant of "oh you're just not using the right/newest/best model/agent/etc." from those on the hype train.
This is the thing. I do not agree with that, at all. We can just disagree, and that's fine, but let's be clear about what we're disagreeing about, because the whole goddam point of this piece is that nobody in this "debate" is saying the same thing. I think the hype is going to scale out practically indefinitely, because this stuff actually works spookily well. The hype will remain irrational longer than you can remain solvent.
3 replies →