[monkaiju]

I always find folks bringing up rubber ducking as a thing LLMs are good at to be misguided. IMO, what defines rubber ducking as a concept is that it is just the developer explaining what their doing to themselves. Not to another person, and not to a thing pretending to be a person. If you have a "two way" or "conversational" debugging/designing experience it isnt rubber ducking, its just normal design/debugging.

The moment I bring in a conversational element, I want a being that actually has problem comprehension and creativity which an LLM by definition does not.

[satvikpendem]

Sometimes I don't want creativity though, I'm just not familiar enough with the solution space and I use the LLM as a sort of gradient descent simulator to the right solution to my problem (the LLM which itself used gradient descent when trained, meta, I know). I am not looking for wholly new solutions, just one that fits the problem the best, just as one could Google that information but LLMs save even that searching time.

[bandrami]

> I'm just not familiar enough with the solution space

Neither is the LLM

[apsurd]

(Trying to find where you might still see this)

I've read the thread and in my mind you're missing that LLMs increase the surface area of visibility of a thing. It's a probe. It adds known unknowns to your train of thought. It doesn't need to be "creative" about it. It doesn't need to be complete or even "right". You can validate the unknown unknown since it is now known. It doesn't need to have a measured opinion (even though it acts as it does), it's really just topography expansion. We're getting in the weeds of creativity and idea synthesis, but if something is net-new to you right now in your topography map, what's so bad about attributing relative synthesis to the AI?

[bandrami]

Because if that's it we've made a ludicrously expensive i-ching.

[imtringued]

If there is something LLMs are good at it's knowing some obscure fact that only 10 other people on this planet know.

[bandrami]

They're also very good at almost knowing an obscure fact that only 10 people know but getting a detail catastrophically wrong about it

[jimbokun]

No, this is the kind of thing LLMs are very good at. Knowing the specifics and details and minutiae about technologies, programming languages, etc.

[bandrami]

Oh Lord, no. Not at all. That's what they're terrible at. They are ok-ish at superficial overviews and catastrophically bad at specific minutiae

[darth_aardvark]

Honest, non-confrontational, non-passive aggressive question: Have you used any of the latest models in the last 6 months to do coding? Or frankly, in the last year?

[bigstrat2003]

I have. And the people who say "use a frontier" model are full of it. The frontier models aren't any better than the free ones.

[satvikpendem]

They note in another comment they don't even use search engines so I don't think they're the right person to ask regarding frontier models.

[satvikpendem]

Oftentimes it is though, good enough for my purposes.

[bandrami]

If you're not familiar with the problem space, by definition you don't know whether or not that's the case. The problem spaces I do know well, I know the LLM isn't good at it, so why would I assume it's better at spaces I don't know?

[satvikpendem]

I said familiar enough, not familiar. For example, let's say I'm building an app I know needs caching, the LLM is very good at telling me what types of caching to use, what libraries to use for each type, and so on, for which I can do more research if I really want to know specifically what the best library out of all the rest are, but oftentimes its top suggestion is, like I said, good enough for my purpose of e.g. caching.

[bandrami]

I still don't get what you're saying. If you possess enough information to accurately judge the LLM's suggestions you possess enough information to decide on your own. There's not really a way around that.

[WCSTombs]

Absolutely, the whole point of the rubber duck is that it's inanimate. The act of talking to the rubber duck makes you first of all describe your problem in words, and secondly hear (or read) it back and reprocess it in a slightly different way. It's a completely free way to use more parts of your brain when you need to.

LLMs are a non-free way for you to make use of less of your brain. It seems to me that these are not the same thing.

[Waterluvian]

Maybe it’s just a semantic distinction, which, sure. I guess I’d just call it research? It’s basically the “I’m reading blogs, repos, issue trackers, api docs etc. to get a feel for the problem space” step of meaningful engineering.

But I definitely reach for a clear and concise way to describe that my brain and fingers are a firewall between the LLM and my code/workspace. I’m using it to help frame my thinking but I’m the one making the decisions. And I’m intentionally keeping context in my brain, not the LLM, by not exposing my workspace to it.

[dexwiz]

Sometimes people just need something else to tell them their ideas are valid. Validation is a core principle of therapeutic care. Procrastination is tightly linked to fear of a negative outcome. LLMs can help with both of these. They can validate ideas in the now which can help overcome some of that anxiety.

Unfortunately they can also validate some really bad ideas.

[jrowen]

I feel I've had the most success with treating it like another developer. One that has specific strengths (reference/checklists/scanning) and weaknesses (big picture/creativity). But definitely bouncing actual questions that I would say to a person off it.

[__mharrison__]

My understanding was that rubber ducking was using a different portion of your brain by speaking the words.

The same discovery often happens when you explain a problem to a coworker and midway through the explanation you say "nvm, I know what I did wrong"