Everyone has an opinion about vibe coded software
There's a lot of debate around my corner of the internet, among friends, and between work peers lately about whether or not vibe-coded/AI-assisted software is a good thing, a bad thing, or the end of our professional careers/society. I've found myself in a bit of a strange position where I'm somewhat forced to learn how to become what might be described as a vibe-coder at work. The pace that things can be built now is pretty mind boggling and the feeling of not being able to keep up, just like how I felt when I first started working in software, has come right back with this weird kind of lazy vengeance.
The alternative I face might look like getting "left behind", at least in a corporate/employability sense. Everyone at the office has basically been told we need to upskill and work out how we can work with LLMs to do our jobs, not necessarily because we believe it'll automatically 10x everyone but because everyone else is also doing it. Our competitors are allegedly slopping out lots of new features, faster, that is selling and changing our day to day priorities to keep up with them. The people who are driving our priorities are also putting pressure on us to do something with AI in the product, personally I don't agree this is the way, building AI for AI's sake... but that's where we're at now it seems.
So I've spent the last, maybe 18 or so months, figuring out how to vibe-code. It's probably not fair to say "I haven't written any code by hand in ages" which I'm hearing a lot from my peers... but at least for the last 6 months I can guarantee the code I've committed has been almost entirely AI generated. And for the last 2 months done nearly entirely autonomously. Primarily with Claude Code, because I don't have to pay for the extremely generous limits we get access to, but also with a small Codex Plus subscription for myself and my side projects.
Where I'm at
It's got this oddly addictive sort of feel to it when you're in the flow state, if you could call it that (deep chat?). Chatting with this model about architecture, tooling, performance, logic, then watching it write what I might've written but significantly faster than me, getting it right the first time (usually), then seeing things just work like I'd planned is pretty incredible. While I watch the craft that I've loved to do disappear almost completely behind this single text input, I'm left feeling both deeply sad but also quite optimistic.
I worry for the junior devs only now starting out that won't get the same kind of understanding or even joy from coding that my 14 or so years of software experience might've brought them. Not to mention the how high the bar must be now for interviewing and landing that first job out of uni/school, it must be daunting.
At the same time I have a lot of optimism for how much access these models will bring to those that otherwise wouldn't have gotten into coding at all. There's a lot of people with great ideas that just got their own team of software engineers to build those ideas. For me personally (and I assume many others) it's unlocked a new learning pathway with a much lower time requirement, and the ability to build projects in literally any language or framework I'd like. I'm now building things in languages I've not touched in years or frameworks I've been too overwhelmed by to get past the "Getting started" section in their docs, both at full speed.
So what have I built?
Well a few things so far, mostly "personal software" projects that I wouldn't have otherwise built because I didn't want to commit the weeks or months it would take to build something. These are 2 that are public so far, the others unfortunately must be kept private as they relate to software tooling that's internal to my workplace.
Fold
Fold is a CalDAV client I've always kind of wanted, but never had the time or motivation to build myself. You can think of it like a Todoist inspired frontend for your own CalDAV server. It only manages todos, no Calendar, notes, or anything special, and it does it the way I'd like it done. The motivation was primarily: I want my todos on my own infra, I don't want to pay what I'd consider a "premium" price for software if I'm only using maybe 20% of their features, and I want to change the software when my habits/needs/ideas change.
I don't expect anyone to use this, it does todos my way and has some quirky features that are important to me, but if you do try it and you have ideas on how it could be better, then by all means give it a go and make an issue or PR.
Links: GitHub Repo, Docs
Xaero Sync
Xaero Sync is a Minecraft client mod and PaperMC server plugin that I've wanted for a very long time that synchronises Xaero Minimap waypoints to the server. Again this is something I want done my way, I'm aware there are other plugins and mods that achieve this in some way that are all slightly... wrong? Or at least not how I'd want it. The difference with mine is that it isn't designed to add anything to the world, or change the vanilla experience (no waystones, portals, sign teleports, etc.). All it does it sync your waypoints to the server, so when you break your MC instance, or forget to copy your waypoints to your new Fedora 44 installation, all you have to do is connect to the server and they'll all download and be there again. It has a sharing feature where new players could copy your waypoint set to their client, but beyond that it's just a server-specific Dropbox/Google Drive for your waypoints.
Link: GitHub Repo
What next?
This has been a bit of brain dump of thoughts and there's a lot more issues attached to the rise of AI (it still feels silly to call it that honestly) that's not here which I'm struggling with too. Like the environmental impact of AI data centres, the memory and storage crisis (and the impact it has on PC gaming, home servers, local AI feasibility, etc.), AI generated content generally (propaganda/advertising, image generation and abuse, slop/low-quality software), or literally every other social issue we need to deal with collectively for years to come.
I don't have well formed ideas on how to tackle anything outside my own bubble of software development, I don't know if I want to try to form them yet either. I often need to remind myself that my greatest impact will always be local-first; my partner, friends, family, workplace, and local community. Maybe posting this will cause one person think about things differently for the better and make some kind of positive change in their community, maybe not. I'm hopeful that we can find comfort in some cautious optimism for the positive things AI could bring to a future that doesn't seem so bright right now.