Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Writing toy software is a joy

blog.jsbarretto.com

Post reply on HN
Print view View original

251–260 of 318 posts

Re: Writing toy software is a joy

#251

Post by SeanDav »

snackbroken wrote: post #132
dakiol wrote: post #39

> Perhaps you’re a user of LLMs. I get it, they’re neat tools. They’re useful for certain kinds of learning. But I might suggest resisting the temptation to use them for projects like this. Knowledge is not supposed to be fed to you on a plate Am I the only one using LLMs as if they were a search engine? So before LLMs I was searching on Google things like "pros cons mysql mongodb". I would read the official document…

The few times I've used LLMs as question answering engines for anything moderately technical, they've given subtly-but-in-important-ways incorrect information such that taking them at face value would've likely lost me hours or days of pursuing something unworkable, even when I ask for references. Whether or not the "references" actually contain the information I'm asking for or merely something tangentially related…

I agree. Use with caution. One of my personal pet peeves with LLM answers is their propensity to give authoritative or definite answers, when in fact they are best guesses, and sometimes pure fantasy.

Re: Writing toy software is a joy

#252

Post by yusina »

dakiol wrote: post #39

> Perhaps you’re a user of LLMs. I get it, they’re neat tools. They’re useful for certain kinds of learning. But I might suggest resisting the temptation to use them for projects like this. Knowledge is not supposed to be fed to you on a plate Am I the only one using LLMs as if they were a search engine? So before LLMs I was searching on Google things like "pros cons mysql mongodb". I would read the official document…

> I would read the official documentation of each db, forums, blog posts, stackoverflow entries, etc. It was time consuming on the searching side. The time it took to read all the sources was fine for me (it's learning time, so that's always welcomed).

This learning time that you welcomed is what you will now miss out on. The LLM gives you an answer, you don't know how good it is, you use it, and soon enough, if running into a similar issue, you will need to ask the LLM again since you were missing out on all that learning the first time which would have enabled you to internalize all the concepts.

It's like the regex engine example from the article. An LLM can create such a thing for you. You can read through it, it might even work, but the learning from this is orders of magnitudes less than what you get if you build this yourself.

Re: Writing toy software is a joy

#253

Post by reyqn »

npodbielski wrote: post #239
pjc50 wrote: post #15

Earlier quoted context omitted.

I used to self-host my email. I don't any more, for exactly this reason. And of course this applies to every other "just do it yourself!" enthusiast appeal. I actually want to delegate it to someone who does it full time.

What kind of issues you have with it? I am running my own email server for.about 10 years and very rarely have to look at it.

I heard a lot of people saying hosting your own email server was the best way to be considered as spam by every big email provider. I think it's a big issue, but I've never actually tried it.

Re: Writing toy software is a joy

#254

Post by nunodonato »

yusina wrote: post #252
dakiol wrote: post #39

> Perhaps you’re a user of LLMs. I get it, they’re neat tools. They’re useful for certain kinds of learning. But I might suggest resisting the temptation to use them for projects like this. Knowledge is not supposed to be fed to you on a plate Am I the only one using LLMs as if they were a search engine? So before LLMs I was searching on Google things like "pros cons mysql mongodb". I would read the official document…

> I would read the official documentation of each db, forums, blog posts, stackoverflow entries, etc. It was time consuming on the searching side. The time it took to read all the sources was fine for me (it's learning time, so that's always welcomed). This learning time that you welcomed is what you will now miss out on. The LLM gives you an answer, you don't know how good it is, you use it, and soon enough, if runn…

Only if you use LLMs wrong. Today's models have deep research which will generate a comprehensive analysis with proper citations

Re: Writing toy software is a joy

#255

Post by msgodel »

nunodonato wrote: post #254
yusina wrote: post #252

Earlier quoted context omitted.

> I would read the official documentation of each db, forums, blog posts, stackoverflow entries, etc. It was time consuming on the searching side. The time it took to read all the sources was fine for me (it's learning time, so that's always welcomed). This learning time that you welcomed is what you will now miss out on. The LLM gives you an answer, you don't know how good it is, you use it, and soon enough, if runn…

Only if you use LLMs wrong. Today's models have deep research which will generate a comprehensive analysis with proper citations

I feel like I should point out that's the dialog engine not the model itself.

Re: Writing toy software is a joy

#256

Post by amunozo »

coolThingsFirst wrote: post #69
amunozo wrote: post #56

Earlier quoted context omitted.

How are you supposed to learn to invent new things if you don't know how existing things work?

There's ZERO room for innovation in the projects suggested. Wanna do some creative thinking? Read some books, learn how normal people think. Explore latest datasets. That's where the meat is. 98% of what you read online has been written by super uncreative(boring) people. They have no imagination whatsoever, IQ is well statistically speaking not too flattering. Internet is ran by the bottom. It's slop by the bottom,…

You learn how to do these things, then innove in other projects. I don't see these two concepts contradictory.

Re: Writing toy software is a joy

#257

Post by amunozo »

bitwize wrote: post #78
amunozo wrote: post #49

Earlier quoted context omitted.

I try to do exactly that. I phrase it as using LLM as teachers instead of interns.

I don't trust Dissociated Press on steroids as an intern; why TF do you think I'd trust it as a teacher?!

It's an integrated search engine, why would you trust random people at StackOverflow?

Re: Writing toy software is a joy

#258

Post by nunodonato »

msgodel wrote: post #255
nunodonato wrote: post #254

Earlier quoted context omitted.

Only if you use LLMs wrong. Today's models have deep research which will generate a comprehensive analysis with proper citations

I feel like I should point out that's the dialog engine not the model itself.

Yes, I think that is understood by everyone

Re: Writing toy software is a joy

#259

Post by pjc50 »

npodbielski wrote: post #239
pjc50 wrote: post #15

Earlier quoted context omitted.

I used to self-host my email. I don't any more, for exactly this reason. And of course this applies to every other "just do it yourself!" enthusiast appeal. I actually want to delegate it to someone who does it full time.

What kind of issues you have with it? I am running my own email server for.about 10 years and very rarely have to look at it.

It was Exim on Debian; generally it would chug along fine for months and then incur some sort of regular sysadmin issue that nonetheless needs dealing with. Upgrades. Occasionally, breaking upgrades. Changes to anti-spam rules on other systems causing deliverability issues. ISP mandating IP migration. Earlier in the process it was an actual, physical rackmount server not a VPS, which incurred extra effort a couple of times to replace it.

And the big one: inbound anti-spam.

It's not that issues are frequent, it's just that there's no guarantee they won't be badly timed. There may come a point in your life where (a) you don't have a lot of free time and (b) you badly need a specific email.

I still have the mail domain, I just pay an ISP to handle it for me.

Re: Writing toy software is a joy

#260

Post by msgodel »

nunodonato wrote: post #258
msgodel wrote: post #255

Earlier quoted context omitted.

I feel like I should point out that's the dialog engine not the model itself.

Yes, I think that is understood by everyone

You'd be surprised. A number of fairly technical people who are just not that familiar with ML I know got confused by this and believed the models were actually being tuned daily. I don't think that's universally understood at all.

That has actual practical implications and isn't just pedantry. People might like some model and avoid better dialog engines like perplexity believing they'd have to switch.

Post reply on HN