I’ve been meaning to make new dev projects for a while. But finding the time to sit down to just code is difficult for me these days. However I wanted to make a watch face for my new watch and I just said screw it.

Through a mix of using the Pebble SDK documentation, and referencing a few AI chatbots, mainly GPT-OSS I’m running in Alpaca, I was able to make it, make it run, and since it’s in C, I was able to have no memory leaks (according to the logs).

I’ve uploaded the source code to Codeberg if you want to look at it, I’ve yet do document it, so hopefully it’s not too hard to understand and not a spaghetti mess.

https://codeberg.org/the16bitgamer/poketech-dx

** Notes on my experience using AI to aid in coding **

When I was taught to coding, Google-fu was the goto way if/when the documentation fails you. However with both how old, and incomplete the Pebble SDK docs are, Google-fu wasn’t helping since the info it would pull was about a decade or older and mostly out of date or hard to decipher information on Stack Overflow. I would almost say using an AI along with the actual documentation and tutorials was a necessity for me to even finish this project.

The AI wasn’t perfect, endlessly gas lighting me, telling me to do over the top things for functions I already implemented, and putting me down rabbit holes going no-where. But compared to what I could get from Google, and it’s ability to try and break down what the code is doing, it was a lot more helpful than I thought it would be.

  • vrighter@discuss.tchncs.de
    5·
    2 days ago

    you know what? I have had several projects where finding documentation online was hard, sometimes impossible for some proprietary stuff. Shit still got done.

    The excuse for rationalizing ai reads as “when it got too hard, i gave up and gave the task to someone else”. And it wasn’t even the task itself that was hard, it’s just the information gathering and digesting stage (the part in which you learn stuff).

    Edit: I also wrote my own watcfaces back when I had the first pebble model. Llms didn’t even exist yet. Documentation was there and it was very clear. Writing my first watchface was done within the day. And it was also one of my first times with c

    • the16bitgamer@programming.devOP
      5·
      2 days ago

      I could write the watch face app in a day, their tutorials were clear enough for that. But holes started appearing for things like, how to utilize 1 or 2bit Palettes. How to implement a sprite sheet and select a character, and the formats which are being returned from Clay when you update the settings.

      There are no guides, just example projects, which I could use and learn from. But man the LLM made it easier. Even when it was wrong, it gave me enough that I could go to the docs and say, oh that’s what that meant.