davidyarbrough

I assume it's possible, but it seems like a huge task and I'm not sure where to start. It seems like official support ended with the loss of upstream security patches, which makes sense. I plan on using my phone strictly on trusted wifi and with non-sensitive data, so I'm not overly concerned about device security, I mostly just want it as a test device for developing applications. A side benefit of porting the OS to the device will be getting more familiar with the OS and device interfaces, which I have been meaning to do anyway as I'll need it for some of the apps I have in mind.

I guess my questions are, where should I start? Are there particular pitfalls I should look out for? Will the lack of modern kernel security have implications for the security of apps developed for that platform? Is such a thing even doable / feasible? Should I just use PostmarketOS instead?

A very small thing just happened to me that I don't think I would have noticed if I hadn't been specifically trying not to use an LLM coding assistant.

I am trying to teach myself Godot and GDScript. I have some ideas for video games that I've had scribbled down for a while now but I'd never built enough of them to get a satisfactory prototype. I am comfortable enough in JS that I've now cobbled together some Phaser code into a working game, but I used an LLM to write most of it, I tweaked it just enough to get it running because I wanted a working prototype more than I wanted something polished. I was able to do this pretty quickly, but I (quickly) ran into the problem a lot of vibe coders run into, which is that I didn't understand how the code actually worked. Debugging began to slow me down because I had to try to learn the code as I was debugging it. It was the difference between the relatively minor hassle of debugging my own code and the usually onerous, occasionally nightmarishly unpleasant chore of trying to debug someone else's code, where you keep having to context switch from "how could this code be producing that behavior" mode to "what is this code even supposed to do" mode. I didn't want to work on the project after a certain complexity boundary because making changes became exactly like one of the least enjoyable parts of my job, having to sift through huge amounts of unfamiliar code trying to cobble together enough context to enable the one minor change I want (then rinse, repeat).

So, I decided to start over in Godot, write in GDScript, use only the Godot editor, and use no LLMs while I figure out how to tell Godot what I want. Just me, the docs, and stack overflow, just like old days. Progress has been predictably slow, but I like it. Something about transitioning from generating huge blocks of code at a time and ripping through a generate -> edit -> test -> repeat cycle to writing everything by hand puts the slowness of it in a new light. This feel presumptuous to say because I know almost nothing about carpentry but it's what I imagine it might feel like going from a day job where you have to crank out furniture pieces as fast as possible, then going home to your woodshop and spending a few hours with the hand tools, getting one piece just so. Something about having a comparison to the new way makes the old way feel more noticeably nicer.

While I am doing this old-fashioned coding, I am reading a lot more than I'm writing as usual for starting out in a new language. I'm also less familiar with resources that are available for Godot, so I spend more time scanning the internet trying to find the bit that answers the question I have. Just now I was looking up whether there's a Tuple type in GDScript and I saw an example that, while it used an Array instead of a Tuple, also used the built-in Vector type. It wasn't what I was looking for, and it didn't solve the immediate problem at hand, but for the game I'm imagining I will almost certainly make use of Vector, so I made a note of it. And it occurred to me that this is a thing that happens a lot. I think that slow, semi-random accumulation of trivia and knowledge that happens over the years while you're poring over old forum threads or rambling tutorials or ancient StackOverflow arguments is really useful for developing a deeper understanding of the tools you're using, and I don't think I ever appreciated it or even noticed it happening until now.