It's 2025; you are young, ambitious, and preparing for a career in programming.
However, seems like everybody in the industry is discussing how AI...
Some comments have been hidden by the post's author - find out more
For further actions, you may consider blocking this person and/or reporting abuse
I agree that javascript isn't a good language to start with. While it has improved it still has no native types.
I think Python is a good beginner language. I would not start with a compiled language because you are going to spend a lot of time waiting for the compilation during learning.
Once you know the basics, you can learn a compiled language because then you will make less mistakes.
As far as I know any language on a microcontroller is compiled to machine language, whatever you pick. So the hardware doesn't need to more powerful. You can program Arduino's with Python.
I don't think any language will be LLM safe forever. Over time people will feed LLM's all the different languages.
I am not a AI optimist but I don't think niche software is going to be AI free. Just use it to your advantage by learning and keep on learning whatever you set your mind on.
Arduino is not recognized by the industry as a professional platform. Also, the level of optimization and hardware control required in commercial projects is not achievable with Python-to-native compilation.
The natural barriers I mentioned will remain in place for a long time. There is no sense in using AI-generated code to control the pump, for example. That is not so much code, but it must be 100% precise.
P.S. The compilation of C is pretty fast, you won't notice in your learning project (it may even outrun Python script interpretation), and it's not difficult to start with, in opposition to popular opinion.
I agree an Arduino is not a professional microcontroller. I mentioned it because it is not needed to go to C from the start. You can ease into microcontroller programming with a higher level programming language.
Why doesn't it make sense to have AI generated code to control a pump? If the code is correct because of the prompt and context you gave AI, what is the difference who or what generated the code?
It might not make sense when it is the only code that is needed, but most of the time your programming a flow of actions that are started by a trigger or some sensor reading.
I don't know how fast the C compiler is. My experience with compiled languages when learning is that is the compile step is like a threshold that is build just a little too high, so that you hit it often with your foot. After a while it will become that annoying you want to lower it.
And that is what I meant with spending time waiting, a lot was an exaggeration. When the change does not have the expected outcome, there is always that chance that the code wasn't compiled when it was tested if you run the compilation in the background. So you run it again to be sure.
With a script language you know you are the one that did something wrong.
I do think the sectors you are promoting are a valid option. But they are not going to be everyone's cup of tea.
Because trials are more expensive than writing, you need the entire assembly (sources, sensors, tank), multiple people to assist, and a switch of scenario requires a lot of manual labor; it could easily take a week for a single build to be tested. And still, some conditions will be encountered only in the production environment, where the cost of a bug is too high. You don't want LLM code to drive the s**t pump near your house 😀. So let's the professional human write precise, purely hand-crafted code.
My experience is opposite. Compiled language LSP offers more support during writing. With interpreted, you need to run it to know what is wrong in general.
P.S. Arduino is not a recommended approach for learning embedded development. As a result, you are learning the Arduino ecosystem's abstractions, which hide a lot of important stuff from you. For anyone who wants to start — STM Discovery boards are the way to go (it's easier than you think). Select one based on the interfaces/peripherals (do you need Bluetooth, Ethernet, etc), f4discovery is the default choice.
So you can't test everything, which means there always is a chance that there is a bug how good the code is your write?
Did you notice the word correct in the part you quoted? I agree with the precise part in the quote above.
Does it really makes a difference if you type the code or that some LLM wrote it. Most IDE's try to predict what you want to want to write on the current line. An LLM is that but you can write English. People still have the final say, I don't want a vibe coded pump.
You are right. The point I was trying to make was more about removing uncertainty from the learning process.
I mentioned running the compiler step with a background process, because manually triggering a build is annoying. Most things that need compiling nowadays have some hot reload process. But because it is a separate process it can fail or be slower than triggering the test.
With script languages, compiling and running the code is one process.
I mentioned Arduino because it is the thing I know, not being immersed in this field. Thank you for the suggestion for a better board.
I did some research, a search engine and AI question, and I found MicroPython
I want to stress that Python on a microcontroller is a way to ease into that kind of programming. At some point the switch to C/C++ is needed. Not everyone is confident enough to dive into the deep end straight away.
PS: I was scrolling through my RSS feed after I made this comment and I read this. It might be interesting.
Of course, bugs in production happen, and devices or equipment are recalled. However, this is significantly rarer than with a web app, server, or mobile app, which can be almost instantly updated.
That depends on skill, of course. LLM can write better code than some people. However, even a beginner-level coder can solve more complex tasks.
LSP completions and snippets are not the same as logic delegation.
Good luck on your path! It's easier than you may think; you will succeed.
Also, the guy in the article you shared did not produce and bring up his helloword-level PCB. It's like writing the code without running. If interested, you can compare the actual ESP32 S3 board and ESP32 design reference to what he did. Or search for USB impedance and diff pair design.
True, but that is beside the point.
True. I don't trust LLM's to write logic for me, that is not their task. So for me both things are on a similar level.
Not my path, I would go for C or C++. I'm thinking of other people.
Are you implying that person is not skilled enough to see the flaws in the PCB design?
The interesting part is that AI helped making the PCB. From that I get that AI already is in the AI safe sectors. Maybe just only with their toes, but how long will it take for the rest to follow? (I don't humanize AI, but I found it a funny analogy)
Maybe he does not want to, otherwise article makes no sense. Even I can see flaws in his design (I am not a PCB engineer). To be honest, that article is pretty low quality.
No, it does not. "AI" just gave answer as it always does. If LLM tells me cure for cancer, it doesn't mean it's right.
Ok that was my mistake trusting that person.
The complication of C is pretty fast, you won't notice in your learning project (it may even outrun Python script interpretation), and it's easy to start with, in opposition to popular opinion.
Arduino is for DIY and learning projects. That's not a professional tool. The level of optimization and hardware control needed in commercial projects is not approachable with Python/JS/whatever to native compilation.
The natural barriers I mentioned will remain in place for a long time. There is no sense in using AI-generated code to control the pump, for example. That is not so much code, but it must be 100% precise.
My article is addressed to the newcomers. Also, it's not only about AI. Common niches are oversupplied with workforce. Embedded development is an area where demand for specialists will continue to outpace supply.
Nice article on "Career Hacks" thanks for sharing
Wow I haven’t seen articles about programming and AI from this POV yet. Thanks for sharing your experience 🙌🏼