Shop Talk Show · Hands-On Engineering Podcasts · November 2026
Dave was explaining why he thinks AI coding tools will survive even if the AI bubble deflates. His argument is that software already has safety nets that professions like accounting or law lack.
…to having it either basic autocomplete or like scaffold out a whole application in a big loop, you know? And I think, will those exist in big data centers? I don't know. Would it exist on a private local LLM? Ideally, you know, would you buy a special GPU for that? Maybe, you know, like, um, like
a possible outcome or a likely one in your book.
Yeah. Well, I just think, I think the reason it works for coding and, you know, like the reason it works for coding is because we built up all these structures, testing frameworks, linters, all this stuff, with all these guardrails around preventing bad code from going out to the customer. You know, like we have built those like ancient the ancients before us and, you know, like have built all this tooling to stop bad things from going into production. Are we perfect? No. Are there a lot of slop and security errors now? Yes. You know, but like, and like when Anthropic is like, hey, we found 70 bugs in Linux and it's actually like 10 and they're actually super minor. Like, is that is also true? You know, like, I think like we have to, you know, I think like there is a utility for these things in coding, even if the industry pops. And I think that was sort of like a, again, like trying to be intellectually honest. I think like that was sort of like the moment I was like, gosh, it's going to be here for a while. And is it going to be large models? I don't know. You know, I think those have still ethical concerns and potentially legal exposures. But then somebody's…
utility at the moment. And yeah,
it works. And that's like where like an accountant or a lawyer probably isn't going to just because it needs or like will not be the early adopter, you know?
Yeah.