I have the same problem. My biggest issue is looking for a resource for "something that does the thing I want but I don't know what it's called."
I can read code pretty well otherwise. It's depressing because over the decades I've always wanted to program, and the motivation is high, but the results just don't show.
That’s true. These days I don’t feel embarrassed but I hadn’t noticed that that had even happened.
Not sure if you've tried reading any textbooks? (I only see mention of Documentation or youtube)
But depending on the language you'd like to learn there's a wide range of textbooks with different writing styles! Some that are more dry, dense, hard to parse; and others that are written in a friendlier manner which can be helpful for newer folks.
A big plus they typically also have excercises at the end of chapters along the lines of "write a program that accomplishes this goal" or "write a guessing game that works like this".
You don't have to hit them exactly but they can also be helpful for sparking ideas of a program that you would actually like to write!
There's also things like Rustlings or Ziglings that are designed to help you learn the language through solving code exercises, personally I really enjoyed Ziglings
It's not wrong the vast majority of the time. In fact, it's petty damn good at explaining principles and giving feedback on code, but it doesn't read minds, and so it can be wrong sometimes. It sometimes tells you you're wrong while explaining the exact same thing you just told it back to it.
People can be wrong, too. I just saw a teacher explain something wrong in my daughter's online school video. She had the right answer but explained it wrong and confused my daughter. I think it's fine to prefer human assistance, but AI is really good at having infinite patience and stepping back and covering fundamentals when you aren't ready to understand a thing.
My point is: however you find the answers you are seeking when confused, keep seeking. Keep asking questions. Keep experimenting.
The difference between a senior and a junior is how long it takes before you feel embarrassed of your code.
Once someone is no longer embarrassed by their code, but can explain it in terms of compromises necessary to get a project to completion on time and within budget, and what they wanted to do at the time but circumstances didn't allow — I'd say that person is ready for a team/tech lead position.
I didn't have this issue, but I will share my opinion anyway.
What do you mean by "actual programming"? I would expect that to include not knowing stuff beforehand, to explore and prototype solutions and missing information. That's the opposite of what you seem to expect, judging from the rest of your sentence? (missing know how, unclear docs, no youtube course to follow along to)
I learned through hacking website and related stuff in php, going into a popular C++ FOSS project, and other projects. I oriented myself through what already existed and documentation.
The best way to learn is through self interest, through actual projects or functionality you try to implement. If it's scoped small enough, with a doable challenge, that's how you grow.
Well, depends on the person. And I think the era--I'm getting the impression what it takes to be a dev in 2020 isn't quite the same as what it took in the 90s. But in any case, I said toward the senior dev level in that timeframe.
Appreciate the insight. It's becoming pretty disheartening to work directly with a knowledgeable person only to hear "have you tried ChatGPT?". The whole point is to be mentored and not have questions rerouted to something that's generally wrong a vast majority of the time, and that's not even factoring in all the unethical uses of AI.
fixing bugs also help you learn
Ok. I’m not convinced of most of what you say, certainly not to the level of certainty that you have. I’d still love to see actual evidence of what you’re saying. Extrapolating 3 years of progress, especially when that progress is rapid, out to infinity is a classic mistake. Do you really think that, for example, the progress that models have made in 2026 is anywhere near the level that they made from, say, Jan 2024 to Nov 2024?
I do agree that things have changed and that maybe it is progress.
Radicle is pretty good. If it gets easier to work with it will be really good.
So basically, this guy is wrong.
I've had this (controversial) theory for a while that I've yet to see proven wrong. Before AI, it's basically been known that most programmers "age out" of the craft, because it's really difficult to keep up with it forever, and most people in middle of the bell curve simply can't do it. Big tech used to get sued for ageism all the time, because they would do shit like only promote or hire young people (which is illegal).
The aging millennials of today were the young rockstar ninja programmers of the past 2~3 decades, but now most of them are getting too old to keep up. That's why you see AI psychosis hitting the older coders harder than anyone else, since it provides a convenient way to not face the realities of aging. It's easier to think you can still be a rockstar by "pivoting" to AI, than it is to admit that you don't have the energy or motivation to keep your skills sharp. Tim Sweeney (and to a certain extent John Carmack) come to mind as high profile examples of this.
A pre-AI version of this is pivoting to management. That's what pretty much everyone does, and it's common to find people like that who haven't written a line of code themselves in years. But I guess becoming an AI cheerleader is easier to do, and the number of open management roles certainly isn't big enough for everyone.
Personally, LLMs are too often wrong and overly confident to be useful for me. Plus, unless you're running a local model, yet another subscription is not fun, especially by the likes of OpenAI and Anthropic. And you have all the ethical issues with LLMs that I'm sure you've heard about already.
It would be much better if you used the vast swath of online resources we are very lucky to have nowadays, don't take it for granted as the internet becomes increasingly saturated with low-effort LLM generated garbage.
Again, the only scenario where LLMs are potentially useful is translation in my opinion, and even then, you have to mindful it is often pretty wrong.
I have yet to see much evidence that indicates that I won’t be needed indefinitely.
I've seen models going from pissing shitty 10-liners that didn't compile to reworking entire projects halfway convincingly in only a few short years. I see no reason why they won't get good enough to do that well enough to render your senior software architect ass obsolete in a few more.
"Today’s AI is the worst you’ll ever use" is correct. You might get some respite for a few more years when the AI bubble bursts and we go back to doing shit ourselves for a while. But the technology isn't going away. And bad luck for us, it's almost perfectly suited to replace software engineers, of all jobs.
If you haven't seen this, you've been living under a massive rock. It's clear as day.
lot of questionable fear mongering going around
Well, I will say this: all I'm saying is that OUR jobs as WE know them are gone. They either don't exist anymore, or they're on the verge of disappearing. What I'm emphatically NOT saying is that it's wrong: it's progress I guess. We're just the latest round of professionals whose profession has been rendered obsolete by technology.
I do worry that so many professions are in the crosshairs that it'll basically break society though, because the changes are so radical and so quick.
a very clear possibility of a talent shortage if we keep going on this trajectory.
It's not a possibility, it's a certainty. And it won't matter: again, I'm 100% convinced the machines will do a better job than you or me very soon.
Don't use AI. Don't do it. People telling you "it's great for learning" are full of shit and going off of gut feeling, not experience. These LLMs have only been around for a few years, and have only gotten semi competent at generating code for even less time. There is no data or evidence that it helps people learn programming, and plenty of evidence that they do the opposite.
So avoid it. Pick a simple project, and see how far you can get. You will probably fail, but that's the point! You'll most likely find something interesting or tough to focus on in the middle of it, and you'll learn a whole lot about whatever that is, and it'll stick with you long after you've abandoned that project.
My advice is to start very small, and avoid anything GUI related. Write like a bare bones terminal only blackjack game, then add a high score table, persist those high scores to disk, etc. If you fail even at step one, that's fine! Just try something else, like tictactoe, Simon says, or a non-game like an app for converting between Fahrenheit and Celsius.
Ask for help on a programming forum somewhere with real people.
Maybe what you really like is theoretical computer science (i.e. glorified discrete math), or computer engineering, or IT or something. Computers, but not software development.
How long do you think you'll be needed to do that?
I have yet to see much evidence that indicates that I won’t be needed indefinitely.
The models are coming for your job.
Gonna need some citations on this, chief. I don’t necessarily disagree, but proper architecture and high level analysis and design, especially of large existing systems, is something that LLMs have always been quite bad at, and I don’t think that’s going to change anytime soon. But I know the state of the art is hard to follow, so if you have any new developments to share I’d love to look at them.
They'll get good enough to replace you, and sooner than you think. If you think you're safe, you're delusional.
I have never thought I was “safe” in this industry, going back 30 years. But I think there’s 1) a lot of questionable fear mongering going around, and 2) a very clear possibility of a talent shortage if we keep going on this trajectory.
toward that senior dev level. But it’s not easy to replicate 3-5 years of on-the-job experience
5 years on the job doesn't remotely qualify you as a senior dev.
Slopmaxxing ftw.
1p per 1000 lines
Yeah.
Get the code generated by some random no-login session and give it to them.
If they end up using it to train their highest tier model, it's their problem.
I still draw on my experience extensively, every day, when designing systems and reviewing code.
How long do you think you'll be needed to do that?
The models are coming for your job. They'll get good enough to replace you, and sooner than you think. If you think you're safe, you're delusional.
And in the long run, nobody will need your finely-honed skills or mine. Young professionals who start a career in software engineering today are basically signing up for long-term unemployement or burger-flipping.
Everyone says it, and I'll say it, too. Make up a project and work on it. You'll hit so many walls, and climbing over them is the best way to learn.
"The expert has failed more times than the beginner has even tried."
As a learner, use AI like this:
Problem solving steps:
This will create a shortage of experienced people who can actually solve real architectural problems and review code accurately.
This is also my current "most likely" scenario. Bob Martin (of Clean Code fame) doesn't read or write code any longer. But he has the luxury of drawing on decades of coding and architecture experience to direct LLMs to do the right thing.
There has been an increase in the number of open positions for senior devs, and a decrease in the number of open positions for junior devs. Median senior salary is $240k. These both indicate value to coding and architectural knowledge.
As seniors retire and quit (a lot of them don't really care for the AI job), we're going to end up with a skills gap.
I teach a couple SE courses at our university and I'm revamping them as we speak to aim higher toward that senior dev level. But it's not easy to replicate 3-5 years of on-the-job experience in 2 quarters.
In China, the gap is being made up for by employers (kinda like it's historically been). I know one thing for sure about the US, though: we'll make corrections when our backs are against the wall and not a moment before. :)
Ok so you want to compress and store my proprietary code in your model which will then allow to to produce my proprietary code.