I did suspect that the triumphalist pronouncements, and even the seemingly honest and forthright deep dive, were a little premature.

This is a key problem with LLM exuberance: it's very tempting to trade on decades of experience in software using LLMs, because one is tired of typing and manual figuring-out, it seems to work if you're competent, and the payoff is essentially immediate. The real bill comes in the mail much later.

▲

krupan 14 hours ago | parent | next [-]

This is so real now that we have LLMs. The stuff I'm working on doesn't seem to exist in LLM training sets, but I still get tempted to try when I get tired of doing the tedious parts.

▲

yuye a day ago | parent | prev [-]

>because one is tired of typing

I consider typing to be secondary to software engineering, but I do concede that for those cases that typing really is the bottleneck, LLMs can certainly be of value.

Those cases are rare, though.

	▲	abalashov a day ago \| parent [-]
		Oh, I 100% agree. I'm imagining the case where someone asks an LLM to do trivial things just for the novelty, or because they are fatigued of thinking + typing, but conceptualise the thing they are putting off as primarily a problem of typign and not thinking (I often do that).