Comment by KaiserPro | Hacker News Reader

Comment by KaiserPro | Hacker News Reader

Disclaimer: I work at a FAANG with exceptionally good integration of LLM into my IDE.

For me its been a everso slight net positive.

In terms of in-IDE productivity it has improved a little bit. Stuff that is mostly repetitive can be autocompleted by the LLM. It can, in some cases provide function names from other files that traditional intelliCode can't do because of codebase size.

However it also hallucinates plausible shit, which significantly undermines the productivity gains above.

I suspect that if I ask it directly to create a function to do X, it might work better. rather than expecting it to work like autocomplete (even though I comment my code much more than my peers)

over all rating: for our code base, its not as good as c# intelliCode/VS code.

Where it is good is asking how I do some basic thing in $language that I have forgotten. Anything harder and it start going into bullshit land.

I think if you have more comprehensive tests it works better.

I have not had much success with agentic workflow, mainly because I've not been using the larger models. (Our internal agentic workflow is limited access)

That's sort of my experience too. It's really good at some auto-complete (especially copying patterns already in your code). For example if you write some cross product function it will easily auto-complete the equations after seeing one. It's obviously not as good as intellisense where that works.

And it's really good for basic stuff in things you don't want to have to look up. E.g. "write me JavaScript to delete all DOM nodes with class 'foo'".

I reckon you're underestimating how much time that saves though. The auto-complete saves me a few seconds many times a day. Maybe 2-3 minutes a day. A whole day in one year.

The "write me some shitty script" stuff saves much more time. Sometimes maybe an hour. That's rarer but even one a month that's still a whole day in a year.

Maybe 2 days in a year doesn't seem significant but given the low cost of these AI tools I think any company is self-harming if they don't pay for them for everyone.

(Also I think the hallucination isn't that bad once you get used to it.)

> A whole day in a year.

Now let's count the hours of "useless meetings" and how much time could be saved there. And that would cost us nothing.

I'm not sure saving two days of work _a year_ is worth the hassle of talking to an LLM like it's a 5 year old - IMHO.

Sure if you have a way to guarantee reducing "useless meetings" we should definitely be doing that.

You don't though.

That's been my experience: it's good at basic tasks. It can be prompted to write idiomatic code in very small amounts. Anything beyond that and it's just as likely to either write non-optimal code or silently delete code while trying to satisfy your ask.

>I suspect that if I ask it directly to create a function to do X

This is how LLMs are most effective, and also the reason why I don't believe non programmers will be crushing code. You do actually need to know how to program.

Reply on news.ycombinator.com