Post

At the risk of starting the flame war to end all flame wars... Modern LLMs (GPT-5.1, Claude 4.5, Gemini 3) produce excellent code and can be a significant productivity boost to software engineers who take the time to learn how to effectively apply them - especially if used with coding agent tools

it's also true I wouldn't accept code from someone who couldn't write that code without an AI assistant and who doesn't know how it works, especially if it's a hit-and-run PR that they won't have to maintain

Same here - my hard rule is that if you can't explain how the code works that code shouldn't end up in a commit to anything other than a toy project

As someone in a leadership role in our software engineering organization, this is actually the sticking point for me: there are often perceived or actual (in our case mostly just perceived) incentives to SHIP FAST in ways that very directly undermine the commitment to understanding generated code.

In my experience, how well the tools work is also (a) very context-dependent, not only in the sense of “how much of this is in the distribution?” but also “what kind of work am I doing?” and (b) depends on how comfortable you are with the totally unsurprising tendency to regress toward the mean.

The latter will leave a very bad taste in any thoughtful software developer’s mouth, and in particular when the push does *not* recognize where these tools have cliffs and sharp edges or even just take (often *much*) more time than just doing it yourself manually would. I find these all the time!

The latter definitely SHOULD leave a bad taste in the mouth! I would hope that the people pushing that aren't themselves experienced software developers

Unfortunately… «points at Shopify and other such places with nasty mandates» Part of the trick is that if you’re in-distribution you get one experience, and if you’re out-of-distribution, you get a wildly different one. It’s really hard to get some folks (not you!) on the booster side to get that.

I find those cliffs and failure modes and “lol, I can do this *way* faster myself” all the time—and, I should add, as someone who does get value out of them in quite a few places. Some of that is the kind of work I do. A lot of it is quite far “out of distribution”… which is what makes the job fun!

The “regression to the mean” part is also something I see your own prompting constantly trying to work around. If you don’t want these tools to generate the most mediocre React code imaginable, you have to *work* at it. And maybe doing that work ends up net positive for you! That’s quite plausible.