AI doesnt do everything 20000 times better than you ever will.

Don’t tell yourself that.

It’s better than you right now, but every mid-level tech worker will tell you it sucks. You have significantly higher capacity than it, you just have to learn to get there and that isn’t easy or quick.

Al does everything 20000 times better than I ever will

This is where you're wrong. Models you can afford are not that good, and you can actually learn to do a lot of stuff better than affordable LLMs ever will

The main issue here is that googling becomes less and less attractive due to the emergence of copycat AI generated articles. Oftentimes, you have to wade through several identical copies of an original tutorial with ads crammed every other line before you find an actual resource.

What I find to work reasonably well:

• IF I use AI, I instruct it to never change files or run console command, but instead explain to me what to do and why
• If not straightforward advice, I will check with official documentation pages and correct if needed

This way, AI acts as glorified Google that can read your local context. For me it works quite well and I learn a lot. Nonetheless I find myself often wondering if I am doing the right thing.

now that AI does everything 20000 times better than I ever will

Now I’m no cognitive behavioral therapist, but, uh… this is what I think is catching you.

I use AI plenty. I write most of my code with it. The garbage it spews out if you don’t hand-edit or get ultra-specific with design instructions is just atrocious, even if it looks sleek in the gestalt. You can give it small tasks and one-shots that it handles perfectly, sure, but it will rot a large codebase to the core if you start letting it make any real structural decisions.

There’s simply no replacing architecture and focused thought, and writing code by hand isn’t going anywhere, even if the majority of the industry works with AI. I mentor junior engineers, I’ve seen the worst spaghetti they can make, and I can nonetheless confidently assure you that you’re better than “frontier” AI across multiple domains.

Don't use it.

I know it seems like everyone is doing it, and it's really good and cool, but that's not reality, and you can certainly just write code the old fashioned way by thinking about what you want to do, and then typing some words on into an IDE or text editor.

Yeah, it's a mental thing. Why draw something when there's proper artists out there? Why learn to play the guitar when Mark Knopfler is out there? Why assemble a Jacuzzi in your backyard, when there's a waterpark in the country? Why do your own barbecue when restaurants exist?

The satisfaction you're looking for comes from the process of learning and doing, not having the final product explained/handed to you. You have to remind yourself that the joy is in the making, not the final product.

Do stuff that you find interesting. I can buy good looking cups, but I still run the pottery wheel to make something by my own hands. It's wonky, not perfect, but it's made by me and I'm proud of that

@super_user_do@fedd.it, don't fall for the (insanely voluminous) hype around LLM capabilities.

LLMs aren't better than you even today. They will extrude mediocre code, far too verbosely, and inevitably lose track of what they are supposed to be doing. When asked to make a change, they will frequently change far too much. They will make a mess that no one understands, least of all the LLM.

You can do better than them, today and in the future, by being curious, reading to understand, intending what you write, and improving from mistakes. LLMs can do none of this, their design does not allow them to.

AI does everything 20000 times better than I ever will

I find it hard to believe that AI throwing pointers and references at random counts as "20000 times better" than anything.

Besides, saying that means that you fell for the worst trap, of believing that everything it does is flawless. It's not. You're in uni, so you don't really have experience in ordering people around, which is what fucking around with AI is - you give an order and wait for a result. If you trust 100% of the end result, the problem is you.

AI is great at stuff you don't know about, because when you do know about it, you'll notice errors and mistakes, making you realize you cannot trust it.

A very long time ago I was taught that the lifecycle of a program is 20% design, 10% coding and 70% maintenance. The LLM can do some of the design & coding for you (which is the stuff you need to learn) but it's going to (if it works) produce functional but impossible to maintain (for which read - change later for any reason) code. You're training your brain to know how to do this stuff, and that's important in and of itself.

It can do classroom bullshit because it's well-defined, small in scope, and generally isolated. If an entire app can be expressed in 1000 lines of code, AI can probably do it faster and better than you. But when you're talking about a full microservice or game, it can't.

Coding skill is about knowing what to build and levels of abstraction. Writing the code itself is more tedium than skill. If you can understand code well enough to see AI output and edit it to repair the bullshit and keep the boilerplate, you're doing great.

The challenge is AI code looks really good while sucking hind tit. Even I go through this after thirty years of experience. AI writes code that looks great because the stupid stuff it does is largely hidden from you — it is stupid decisions it made and then wrote code consistent with that decision. You haven't thought through the hard bits so it's solution looks simple and right. But it did the thing wrong.

I use AI all the time, and I find it helpful, but it's FAR from perfect. So very far.

I always either cede to temptation to use it (thus nullifying learning) or just get bored so quickly I just turn off the computer

This just sounds like a question of discipline, and AI is likely not the issue.

I can relate. Whenever I wanted to learn programming, it felt like a chore to sit down and do it. This happens when you're kinda compelled to learn it — such is the case in university.

There is no magic trick. The answer is discipline and fighting your urges. Just tell yourself you're gonna sit down for 90 minutes or however long, and you will work on X task without using AI (or whatever other guardrails you decide on). When you explicitly tell yourself that, and set up explicit boundaries and guard rails, you will do it. You just keep at it until it works.

As I said earlier, this is how I felt even before AI. After a few months of persistence, it suddenly clicked, seemingly overnight. I became addicted. It became more difficult to stop doing programming. You just need to get in the flow long enough for programming to become natural and enjoyable to your brain.

I have found a few of my friends doing the same, and I have used AI on occasion. What I would suggest as someone who went through post secondary education without AI is to limit it, and see it as a tool. You can use it for more than code generation. I've used it to lay out plans in formats I don't care to learn, but work well with CLI tools, help with debugging, and generate trivial code, like boilerplate or refactors. I admittedly am a bit of coding addict, so your mileage may vary, but i do think that it can help adjust the challenge you deal with, and, by that, give you the sense of achievement that will build a lasting joy from the process.

have a vision and a project that you want to actually control the design of. Have a goal about being able to re-use and re-learn from your project when you look at the source of again, so that things are as well structured and as re-usable as possible. Make your goal to be a better programmer and to fully understand your code.

Don't use LLMs to learn, if you insist on using them use them to automate things you already can do yourself. Treat university like a 40 hour a week job, you need to put that time in to build knowledge & retain it through practice if you want to develop your knowledge and intuitions. If you're going to dedicate that time to it you'll be less inclined to take shortcuts since saving time won't matter because you'll spend the same time working at it either way.

I've been thinking of all the places computers have democratized industries. Music is produced by someone on a laptop now more than a collective group of practicing and talented performers. 3d printing produces objects of great sophistication without really any craft skills at all. Chess is a bigger industry now than it was, and no one can beat a computer.

But even then, sports have been something that anyone can participate in to various degrees, yet we still admire people who reach the pinnacle of performance.

Before starting my own aquarium, a wise person said "Aquariums aren't about keeping fish, they are a hobby for people who like to keep water, and the fish are just enjoying that". Every craft, every sport, every performance has practice and maintenance that itself has to bring the enjoyment, the puzzle, and the promise.

I see that for programming. It will be for someone who desires the craft of computing more than the outcome of the application, and they will achieve things that are simply beautiful. Because they love the craft.

You can understand things in principle, implement them once and still use AI afterwards. That's basically how uni will work from now (at best). I don't mourn programming on its own tbh, because in the field you rarely do exiting stuff. And the exiting stuff you will likely still have to do, because it is the hard stuff

is there a way to revive the interest for programming and building projects

Read some books on programming. Or read some programs others have written. This was good advice for finding inspiration even before AI.

These models can create apps that are great for a single person with no skill but they all miss something. its a great tool for analyzing. but it misses in style and domain.

here is my best analogy. i love mesding around with suno the ai music creator. it offers similar fun as mad libs does for me. i plug dhit in snd see what i get. well i am 50 and miss 90s alternative rick/grunge music. i try to see what suno can come up with. recently i tried to have it mimic the wallflowers. it created a song thay seemed technically the same but didnt have the same feel.

a non dev can plug shit in and have an app but it takes skill to use that tool to make a solid piece of software.

All AI did was move the bar up SLIGHTLY for new devs.

You're alright.

Don't know how to escape that though...

Put it in coaching modę. Write the code yourself and only invoke it when you need a principal explained. AI doesn't understand things, but you do. Use what it's good at — writing text in statistically predictable ways to help you get better at what you're good at.

If you can't drop the use of AI for your own learning, use it for learning. Tell it specifically that you want guidance on how to do something, not to do it for you, perhaps have it create a html guide you can follow without it giving you all the answers.

If it does something in a way you didn't expect, ask why. Ask if the way it has done something is the best way of doing it, and then do a bit of research and challenge it when it's wrong, and it will definitely be at times.