Hacker News Reader

Comment by hn_throwaway_99 | Hacker News Reader

I would just add that gen AI can actually be a fantastic aid when you're building some toy projects just for yourself.

I am primarily a backend engineer, though I am capable on the front end, just not particularly fast. I am especially slow at the voodoo that is CSS. In the past, I would have been turned off from doing some end-to-end personal projects because I know I would have been bogged down just battling CSS for hours and hours (it's particularly problematic that I'm slow at CSS but also a stickler for how I want things to look). With AI tools I can literally just say "make it pretty". It gets me about 85% of the way there, but importantly it gets me to the point where it puts in all the styling framework and then I just have to fix bugs/edit/tweak it, which goes very quickly.

So now I've found myself much more likely to build personal projects because I know I can get past things that in the past felt a bit like quicksand to me.

View on HN

It's a balance I find.

On the one hand, the machines can spin up these bridges over unfamiliar code chasms.

But on the other hand, you (at least I) have no investment in the code. No familiarity with it. No "skin in the game" when I can tell the AI to just redo it.

One way around that is by using the "Chat bots as search engine", and then, rather than cut and pasting stuff into your project, type it in by hand. Even if you're just straight up copying it, by typing it in you have a bit more connection to it. When I do that I can better see the unfamiliar patterns rise out of the code. It's much better for me than just reading it.

Also gives me opportunity to tweak it as I like, which can break things, which can teach me things.

Mind, I've been advocating this "type it, don't paste it" philosophy for years and years, for any code snippet found in a book or tutorial or wherever. It's just now with the bots, the code is more tuned to our exact use case, and the incentive for moving it over is less.

But I find grabbing a blob of code, pasting it into my project, and getting errors so that I can tell the bot whats wrong is not particularly satisfying.

Watching Claude figure out its own errors even less so. (It's very cool to watch, just not very joyful to me outside of "wow, that's cool".)

On the other hand, I have no problem pasting a snippet back with a "what exactly is this doing here" to get better explanations of what I'm seeing.

> rather than cut and pasting stuff into your project, type it in by hand

This type of mechanical copying is probably very important for learning and staying familiar with what it is doing.

Just last night I had ChatGPT help me with some algebra and geometry for a hobby project I'm working on. I'm rusty and got stumped so I asked the bot to unblock me. It helpfully walked through the entire simplification / substitution / solving process step-by-step, which I manually copied down using pencil and paper. Following along like that, mechanically, really helped it click for me. I plan to do more of that.

Hard agree on this.

I've come to the point of using current AI tools as "search on steroids". For coding tasks, I now ask questions using more generic language, looking for references, and when the AI comes back with specific examples, I try to refine the question in terms of the example that the AI has given me. "Kinda like that, but inside of Book, I actually want to do X with the Author"

This makes it so that I have to manually translate code that the AI is showing me to my specific case, typing it out myself in my code's context. This has helped immensely with developing my familiarity with a couple of new (to me) frameworks, and with slowly adding to my overall understanding of what I'm /actually/ building.

The joy of personal projects is they can look like shit! You'd also be surprised how much you can learn by just screwing around with something for a day. A big part of the whole exercise for me was getting out of my comfort zone and learning that I can learn. I'm primarily backend as well, and came up through the ops side (as opposed to formal CS), which always led to a lot of imposter syndrome, so a lot of the benefit came from just putting myself in an uncomfortable position and just staying there long enough to work my way through it.

That said, my goal for this was to wear the hair shirt (and also this was before GenAI was really a Thing) - if you just want to make something neat, absolutely vibe your way there.

But I don't want them to look like shit.

When I have gen AI create a starting point for the CSS for me, I still go over every line, and I always have to do a bunch of fixes and edits to get this to look how I want. Frankly, dicking around with the esotera of CSS is something I absolutely don't want to do and don't care about - it's really just reminding me of syntax that I forget because I don't do it often enough. This is also why I find gen AI invaluable for shell scripting. Hunting through reams of man pages doesn't make me learn any faster than just looking at the specific arg flags gen AI outputs (and potentially asking AI more about the details of that command). I just don't do enough shell scripting to really remember the command/arg specifics long term for things I don't commonly do.

> So now I've found myself much more likely to build personal projects because I know I can get past things that in the past felt a bit like quicksand to me.

I'm the opposite of you (my strength is in voodoo) but my experience is the same. Whereas before I'd spend 2-3 weekends getting bogged down on elements that aren't my strengths and never complete, now I can easily whip up a mostly functioning web app (just for me) in just a few hours.

This is how 100% of non-technical executives view software engineers, really. And probably explains a lot of the hysteria (unfounded as I may believe it to be) around LLMs eating the industry.

A good reality check is: if a stranger asks you about a specific part of your toy project, would you be able to explain it?

If you can't, it means there's something there you don't understand, and you lost an opportunity to learn. Of course, this is also true for reusing libraries and stuff.

Within a job, what matters is the _team learning_, and it's roughly the same idea. At least one person in the team should be able to explain anything the team does. Trusting it to an AI puts the whole team in a precarious situation, even if it is something as simple as a CSS layout.

>A good reality check is: if a stranger asks you about a specific part of your toy project, would you be able to explain it?

I like that as a benchmark. I think it also works with LLMs too. I have had best results with AI generation of code when I pass that check. When using a LLM, the chatbot is the stranger.

In fact much of the frustration I have with using AIs seems to be for models tuned to the situation where the user does not have the ability to explain it specifically and it has to make broad assumptions. This may be the best result for a lowest common denominator, but it's frustrating when it provides a generic solution to a specific problem.

I can see specialised models diverging to focus on different levels of assumed knowledge and expectations. Perhaps to have modes of thinking, maybe as a more tone based MOE. Is anyone doing mixture of finetunes?

I would argue that, at least for me, gen AI helps me learn things much faster.

E.g., like I said, previously I would have been turned off from doing some projects because I wouldn't have spent a ton of time dicking around with things that are just inefficient. Now I have gen AI create a starting point, and it's very clear to me "Ah, this is how I would use grid layout to get what I want". I can tweak things and it's very clear to me how things are working.

my current opinion is that uf i can get an ai to write details into a file that another team member's ai can read and then know whats up, thats close to the same as the team member knowing it, as long as theyre either reading the file that the ai read, or working through the ai and providing that file(s) as context.

unless you own/write the whole software stack(not the web stack), i doubt anyone can fully explain their project.

For example, if in your project you have to write a file, and someone ask you how does the kernel cache file writting, that would not be an easy question.

You're missing the point and taking it to an extreme.

Let's say I write a toy project to log all sites I visit to a text file. The goal is to learn browser extensions, not learn kernel cache. There's a clear boundary, and anyone asking me about kernel cache would sound very weird. It's obviously a lot of layers of knowledge deep into the stuff I'm learning.

If my toy project is a beautiful web app, then CSS is _not_ several layers deep. It is in the front of it. It's a dependency I _need to learn_ in order to own my project.

If my toy project is a web app, and doesn't matter if it looks beautiful or not, then I want to spend my time away from CSS (either writing it by hand or interacting with an AI). I am fine with having it ugly.

Understanding these dependencies and layers of knowledge is part of the learning as well.

Again, this works for teams as well. There are certain things, at a certain level, that if you don't have someone in the team that knows it, leads to a precarious situation.

The fact that people don't recognize this, and mistake it for not made here syndrome and other nonsense, is actually scary. It means those recipes (don't reinvent the wheel, good is better than perfect, etc) became mantras that people repeat without actually understand why they exist, and those boundaries have become arbitrary and meaningless.

> If you can't, it means there's something there you don't understand, and you lost an opportunity to learn. Of course, this is also true for reusing libraries and stuff.

Yeah, in libraries that's the aim, software composability and reusability, you don't need to keep everything in your head. Same here.

A "lost opportunity to learn" can be preferable to not making the thing or making it much slower.

Reply on news.ycombinator.com