Exactly: that’s tight feedback loops. Agents are also capable of reading docs and source code prior to generating new function calls, so they benefit from both of the solutions that I said people benefit from.

As an even more obvious example: students who put wrong answers on tests are “hallucinating” by the definition we apply to LLMs.

making the same mistakes

This is key, and I feel like a lot of people arguing about “hallucinations” don’t recognize it. Human memory is extremely fallible; we “hallucinate” wrong information all the time. If you’ve ever forgotten the name of a method, or whether that method even exists in the API you’re using, and started typing it out to see if your autocompleter recognizes it, you’ve just “hallucinated” in the same way an LLM would. The solution isn’t to require programmers to have perfect memory, but to have easily-searchable reference information (e.g. the ability to actually read or search through a class’s method signatures) and tight feedback loops (e.g. the autocompleter and other LSP/IDE features).

This seems like it doesn’t really answer OP’s question, which is specifically about the practical uses or misuses of LLMs, not about whether the “I” in “AI” is really “intelligent” or not.

Fair, but it’s one that the typical tools for finding bugs, tests and static analysis, cannot actually help with.

For what it’s worth, I agree with you about branches, and there are various ongoing discussions about how to make working with branches more convenient. I use an experimental feature called “advance branches” that makes it mostly fit my workflows, and the other benefits of jj are sufficient that I haven’t switched back to git.

I create log files of runs, temporary helper scripts, build output, etc. in my working copy all the time.

The solution to this is to just have a more aggressive .gitignore. But also, note that the “working copy commit” isn’t generally something you want to push or keep; think of it more like a combination of the git staging index and an automatic stash.

Apparently the JS name was selected and announced in partnership with Sun from the very beginning, and Sun had the copyright over both Java and JapaScript up until the acquisition by Oracle. I had no idea, but that makes perfect sense.

Oracle? Oracle owns Java, not JavaScript.

Edit: mea culpa! Sun owned both!

Do you mean Dan Luu, or one of the studies reviewed in the post?

Yeah, I understand that Option and Maybe aren’t new, but they’ve only recently become popular. IIRC several of the studies use Java, which is certainly safer than C++ and is technically statically typed, but in my opinion doesn’t do much to help ensure correctness compared to Rust, Swift, Kotlin, etc.

I don’t know; I haven’t caught up on the research over the past decade. But it’s worth noting that this body of evidence is from before the surge in popularity of strongly typed languages such as Swift, Rust, and TypeScript. In particular, mainstream “statically typed” languages still had null values rather than Option or Maybe.

Note that this post is from 2014.

Partly because it’s from 2014, so the modern static typing renaissance was barely starting (TypeScript was only two years old; Rust hadn’t hit 1.0; Swift was mere months old). And partly because true evidence-based software research is very difficult (how can you possibly measure the impact of a programming language on a large-scale project without having different teams write the same project in different languages?) and it’s rarely even attempted.

Notably, this article is from 2014.

Most of those comments are actually just random people arguing about the merits of the experiment, not continued discussion with the bot.

Also, the bot is supposed to be able to run builds to verify its work, but is currently prevented from doing so by a firewall rule they’re trying to fix, so its feedback is limited to what the comments provide. Humans wouldn’t do great in that scenario either. (Not to say the AI is doing “great” here, just that we’re not actually seeing the best-case scenario yet.)

You don’t have to imagine it; you can browse the Linux Kernel mailing list!

That’s called a mailing list

/s

TypeScript is a language, and traditionally languages are considered separate from their implementations. When I first saw the headline I hoped maybe it meant a non-JS runtime for compiled TS, and I’m well aware of the difference. Yes, that would be a much larger undertaking than porting the compiler to a new language, but the headline doesn’t indicate how large a project this is, and Microsoft certainly has the resources to write a new backend (even a native-code one) for the TS compiler.

This headline is based on responses to the question “what are your biggest worries about the future of Rust”, not “are you worried about the future of Rust.” So of course most of the respondents answered with a concern about the language.

In addition to not actually being correct, I don’t think the information you’ve provided is particularly helpful in answering OP’s question.