I prefer to use none AI at all.
Are they in order? If not, I would like to see your order.
That was chatGPT's answer, intended te be seen as a joke answer.
I follow --
1. Online Communities and Forums
2. Online Tutorials
3. Experiment and Learn:
4. Arduino IDE and Documentation
5. Be Patient
thank you for actually answering this guys question. why cant people focus on finding answers, as you have done?
try this link.Reddit - Dive into anything
@J-M-L the new poster above evidently did not recognize the joke.
ChatGPT be8ng happy about a chatGPT answer maybe? 
That is certainly among the many joys of AI blather.
Hi,
ChatGPT is a great help to me, it is still very early days but as one of you A.I. related YouTubers I follow often says "Remember that it will only get better, what we are seeing now is the worst it will ever be".
In the 80s when I got myself a computer people scorned and laughed at the idea that every home will have one some day...My computer was not really a lot of actual use, but entertaining to learn about and learn to code etc. but knowing about them when most people had tried to just ignore them served me very well 
ChatGPT is great for asking how you would do something or for suggesting improvements to code I have written, if you have some mundane task such as creating some Json or a simple procedure it will do it all for you instantly and be able to answer questions about it or make further modifications.
e.g. I recently made myself a DRO for my milling machine and I wanted it to be able to read gcode so I asked ChatGPT to create a procedure which will take in some gcode and extract x,y co-ordinates from it which it quickly did and I then just took this and modified it rather than start from scratch which saved me an awful lot of time and effort.
If you find some code you want to use you can paste it in and ChatGPT will instantly understand it and be able to explain to you how it works and answer questions.
As I say, it is not perfect but every day it improves. Have a look at "Devin" for an idea of the way things are heading.
I installed an accelerometer on a childrens swing and output the movement speed on ws2812b lights. chatgpt helped in problems about which forums said:
"oh its not possible with commercial accelerometers" in a few ways.
-
getting true acceleration free of gravity effect. The actual acceleration value free of the tilt of accelerometer.
This was said to be either not possible or barely possible with little guidance on how it can be calculated. Chatgpt provided some guidance which actually made it much better than anything i had tried so far... -
calculating of the speed value based on the movement start point at the end of a swing. This was an easier problem however the guidance in code regarding time and acceleration value was very easy to understand.
Im not here to discuss if ai is good for microcontroller use. I know it is better than anything else. The topic i wanted to explore was more related to prompt engineering and how prompts could be made better, basically communicating with ai with a language stemming from microcontroller use... however it struck a wall with the "well if you get what you want what else do you want to know... "
I can't really help with that as I have only really done the basic stuff - it is a great subject though and would be very helpful to see what other people are using.
Well, if you conider beeing ignorant about physics 101 and then getting "help" by AI (which is missing the I part toattally) in the form of resurrugating physics 101 in an "intelligent" manner ... yes, then AI is definitly helpful. But keep in mind, this is "Ideocracy" in its purest form (search on youtue, you might find the full movie) - it does not help you understand the funamentlas, it just feeds you with the results of a search.
I'd be interested to see the code you/it came up with for:
ChatGPT is very patient with questions that people more familiar with the domain material might think tedious or dumb.
If you are looking for good prompts that lead quickly to a good solution, starting with a good understanding of the material and a specific question would be key.
If you don't have good domain knowledge or a clear question, a lot of the prompting is iteratively leading you to some local optimum that makes you believe you have a good answer.
I disagree I can think for myself just fine and I use Chatgpt. Chatgpt is just a tool and a useful one if you use it right.
I have only a rudimentary knowledge of coding, and not time to learn more.
From this standpoint, ChatGPT have been awesome. I have made Arduino projects far more advanced (for me) in a shorter time than I ever could have achieved without. Forums are great, but they are slow and they don't write the code for you.
Nothing works on the first try, but ChatGPT is a great debugger, but some knowledge of coding helps understand the process.
The downside is that I don't understand what my code does, but the result is my goal for these projects and I live happily with this.
One example is my outdoor Arduino. It samples the temperature and pressure every 10 minutes and displays the previous 144 readings as a graph served on a webpage. With max and min temps/pressure. On another port, the current temperature is displayed with a "word of the day" grabbed from the web.
These types of projects are way above my reach should I do them myself. But I made them in hours with CharGPT.
Im hoping this topic aged well, here are my recent suggestions:
-
step by step solutions are much better than all at once. i usually find myself typing "create a function that..."
-
you can paste serial data, chatgpt understands and even graphs it for solutions
I know very little about coding. I use Chat GPT to create code to enable me to make a clock for example. As we all know, Chat GPT is not perfect, and I am fine with that, because it is a learning experience. When AI creates the code it also creates comments on what that bit of code does. So I can see what each part of the code is doing. Then if something doesn't compile correctly, I can google the error or even post the error to AI and it will come back with a reason why it didn't work and offer a solution, debug code, and even revised code. The latest clock I put together was totally coded by AI, and I wanted to see if the clock was really updating with NTP like I wanted it to. So looking thru the code and seeing other "Serial.print" statements in the code, I was able to add a statement that told me when the clock updated the time. I realize that an LCD clock, or blinking some LEDs randomly is pretty easy for someone who knows what they are doing. but for a total beginner like myself, AI makes it easier to understand what's happening. I have been able to learn from AI. So even tho it is basically a search engine, it is helpful.. Hopefully someday soon, I will be able to write my own sketch to do something. Like everyone did at one time or another we all started at the beginning. My thought is go with what works for you..
AI-powered coding assistants have revolutionized development, but are they creating more problems than they solve? In 2025, AI-generated code is faster than ever, yet debugging it has become a major challenge.
- Lack of Readability – AI often produces complex, unstructured, or overly optimized code that's hard to understand.
- Hidden Logical Errors – Since AI predicts rather than “understands,” it can introduce subtle, hard-to-detect bugs.
- Unexpected Edge Cases – AI struggles with real-world exceptions, leading to unpredictable failures.
- Dependency Issues – Auto-generated code may not align well with existing frameworks, causing integration headaches.
While AI speeds up coding, does it actually save time when debugging takes twice as long? How can developers balance speed with maintainability in AI-generated code?
One possible solution is the use of AI-based validation tools such as CriticGPT or Google’s Jules. These tools can identify potential bugs and inefficiencies in AI-generated code.
So you want to "improve" AI-generated crap by trowing the next layer of AI-crap aka "Jules" on top of that that pile of garbage and call it an improvement then? Great, you should apply at german education ministry for the next great thing in learning how not to learn - just in case they missed some bad ideas.