zlacker

Domain discovery is so underrated. LLMs remove so much friction that makes everything so incredibly accessible.

If LLMs couldn't do anything else then that alone would still warrant an invention of a century sticker.

>>wrapti+(OP)
No doubt. The tedious stuff is just a time saver. The domain discovery is often the difference between making progress and giving up.

>>wrapti+(OP)
Couldn't agree more. The first time I used Claude Code was for something very much like this. We had a PDF rendering issue with Unicode characters in one of our libraries. We ultimately needed to implement a sort of bespoke font fallback system.

With the help of the agent, I was able to iterate through several potential approaches and find the gaps and limitations within the space of an afternoon. By the time we got to the end of that process the LLM wrote up a nice doc of notes on the experiments, and *I* knew what I wanted to do next. Knowing that, I was able to give a more detailed and specific prompt to Claude which then scaffolded out a solution. I spent probably another day tweaking, testing, and cleaning up.

Overall I think it's completely fair to say that Claude saved me a week of dev time on this particular task. The amount of reading and learning and iterating I'd have had to do to get the same result would have just taken 3-4 days of work. (not to mention the number of hours I might have wasted when I got stuck and scrolled HN for an hour or whatever).

So it still needed my discernment and guidance - but there's no question that I moved through the process much quicker than I would have unassisted.

That's worth the $8 in API credit ten times over and no amount of parroting the "stochastic parrot" phrase (see what I did there?) would change my mind.

>>wrapti+(OP)
> Domain discovery is so underrated. LLMs remove so much friction that makes everything so incredibly accessible.

And, unfortunately, a lot of friction from not having access to information in the first place. I've read a bunch of docs from people talking to Glean in order to explore a new topic; if it's a topic I'm actually very familiar with then four out of five times it is somewhere from misleading to catastrophically wrong. Any internal terminology that doesn't match the common usage outside of our organization poisons the whole session and it will make up things to join the meanings together, and the prompter is none the wiser.

I trust AI only as a gap filler in domains that I'm already an expert in or where there's little internal context, anything else is intellectual suicide.

>>epicco+g4
Yes, that’s true and very cool but you’re an expert. Where do the next generation you’s come from? The ones that did not do weeks of dead-end research which built resilience, skill and the experience to tell Claude now saves them time? You cannot skip that admittedly tedious part of life for free.

I think pro-AI people sometimes forget/ignore the second order effects on society. I worry about that.

>>whatno+Zc
On the other hand, I remember lots of stupid beginners questions I had, when learning to programm. My peers did not know them either and I had to wait sometimes days for the opportunity to ask someone advanced who knew. Blocking my progress.

(Asking online was a possibility, but instead of helpful answers, insults for being newb was the standard response)

With a LLM I would have had a likely correct answer immediately.

And yes, yes what if it is wrong?

Well, I was also taught plenty of wrong stuff from human teachers as well. I learned to think for myself. I doubt anyone decently smart who now grews up with those tools, think they are flawless.

In the end, you are responsible for the product. If it works, if it passes the tests, you succeeded. That did not change.

>>aloha2+19
I feel like if you're using it in the right way, asking the AI to write code, or give domain context in a specific way then the answers it gives are easily verifiable enough- and it's domain knowledge you wouldn't have gotten very easily through a series of google searches- LLMs as a kind of search can work great.

>>lukan+nn
> My peers did not know them either and I had to wait sometimes days for the opportunity to ask someone advanced who knew. Blocking my progress.

Hypothetically, a solution to a problem that preoccupied you for days would translate into a more stable and long-lasting neuron configration in your brain (i.e. be remembered) than a solution to a problem that preoccupied you only for the time taken to type the prompt in.

>>Gensho+Xx
That is somewhat true, figuring things out on my own makes me really understand something.

But I don't have the time and energy to figure everything out on my own and I stopped learning many things, where some useful hints in time likely would have kept the joy for me to master that topic.

So it is probably about the right balance.

>>lukan+sB
There's definitely a balance. Someone told me years ago that when they'd look for one bug to try and fix it, they'd realize a bunch of other stuff about their code along the way. You learn a lot by struggling with a problem exactly when it feels unproductive. On the other hand, there are cases when maybe it's better to get an answer today than spend a week really learning something. For example if you don't care about how a library itself works, AI helps abstract the details away and maybe there really is no cost to that as long as you can see it works.

>>whatno+Zc
I think there will always be people who want to look under the layers of abstraction, and people who don't care. All the abstractions we've created for computing has lowered the barrier of entry for people who want to create useful applications and otherwise don't care. If anything, LLMs make the process of learning for those in the former group much easier, something that only search really did previously.

I do think it's entirely plausible that a lot of people who otherwise would have wanted to learn more will grow up getting used to instant results and will simply not do anything the LLM can't do or tell them. Kind of similar to how my social media addicted brain gets antsy if it goes more than an hour without a fast dopamine hit (hence me being on HN right now...).

>>whatno+Zc
> Yes, that’s true and very cool but you’re an expert. Where do the next generation you’s come from?

I agree that this is a concern, and I even worry about it for myself. Did I miss the opportunity to add another brick to the foundation of my expertise because Claude helped me out? Would I be marginally better at solving the next problem if I'd worked through the week I saved?

Even if the concern isn't some specific knowledge I'd have gained - did I lose out on a few "reps" to build grit, determination? Am I training myself to only like easy solutions that come out of Claude? Are there problems I won't want to solve because they're too difficult for this new "augmented" workflow?

I don't know the answers - I can only say that I do care, and at the very least I'm aware that there are new dynamics affecting my work and expertise that are worthy of consideration.

>>lukan+nn
When I was a beginner programmer, I was 13 years old. I remember noticing that one kid in our class managed to do and use things that no one else in our class did. I asked him how, and he said "it's built-in, I read about it right here" and pointed to the Java API docs.

Assuming you're literate, there's no age or skill level at which it's necessary to get stuck churning on beginner-level questions. The option to RTFM is always available, right from the start.

To this day, readiness to RTFM (along with RTDS: read the damn source) is the biggest factor I can identify in the technical competency of my peers.

>>pxc+Gq2
So you say, you are a great autodidact. Good for you.

Well, I guess I am, too, but I still see great value in asking specific questions to competent persons.

Or don't you think asking teachers/instructors questions is helpful?

>>lukan+033
I can’t remember ever asking a question and getting a helpful answer to be completely honest..

I feel weird when I read about people needing support. Maybe there is something wrong with me.

>>whatno+od3
To also be completely honest, either you have been really, really unlucky with your teachers, or you should improve on the way you ask questions.

I know I had mostly bad teachers and am largely a autodidact myself. But the few good teachers/instructors I had, were really helpful for my learning progress.

>>lukan+033
Yes, definitely. But I think reaching for an LLM can mean failing to build that reading muscle in the same way that leaning on teachers can. And I also think that many people never learn to read documentation not because they can't but because of a lack of willingness to try to learn to read specialized genres (of which technical documentation is just one).

A teacher can be a unique resource, but asking the teacher is often more of a reflexive shortcut than the thoughtful use of a unique resource.

I think use of LLMs (like StackOverflow before them) are more likely to discourage people from seriously or patiently reading documentation than they are to act as a stepping stone to a habit of more serious inquiry for most people.