I've been a professional software developer in the mobile space for around 15 years now. I started in the Android 2.1 / Eclipse / Ant days and iOS Objective-C days pre-ARC on Xcode 4 (if memory services, I've drifted a bit from iOS since then) doing a mobile app for my university's biology class around reporting field sighting of local fauna and flora.
Probably like most software developers, solving real world problems with code tickles something in my brain that makes it almost addictive. There is this inherent drive to type weird words into a computer that makes fancy sand do weird things on these devices that have become so important to us.
That addiction / dopamine loop is in part what I think makes the current era of software so dangerous. Being able to 10, 20, or 100x that dopamine loop to solve a problem (this is to say nothing about the path that was used to solve it) has a tendency to give a false sense of confidence and a cheap thrill.
I've been struggling a lot with this lately as I've recently been laid off (this is the 4th time in my career). To the public, there were no 'AI related' reasons for the layoff, but more related to the post-COVID era WFH contraction that has been slowly occurring even before the recent AI boom.
Being laid off is not the hard part of this all. I am extremely grateful that every time I've been laid off, there has been a good enough cushion of severance that I don't have to be in panic mode right away taking the first job that comes my way, and I'm able to be a little more selective.
The hardest part of this current layoff revolves more around the interview loops. I have alluded to some of my misgivings against AI, and regardless of those misgivings, it feels like my employment in this industry is intrinsically linked to this tool that has been laid out in front of us. Every interview has such a large focus on AI and how to use it effectively, that there is very little room in the conversation to have real thoughts exchanged on other topics.
Putting my own conflicting feelings on the topic aside, I did my level best effort to dive into AI head first. I followed up on all of the Twitter threads on hot new topics like loop engineering, spec-driven development, etc. I created three projects of varying complexity with varying levels of input from myself to understand these workflows. One of those projects more or less acts as my main AI harness that I did most of my development through on projects where the blast radius is not 'production' and I don't have to worry about passing along any maintenance burden.
Low - https://github.com/j-roskopf/Lexikon
Medium - https://github.com/j-roskopf/Phoebe
High - https://github.com/j-roskopf/Andy
I saw an internet thread awhile ago that hinted at the idea of personalized software from AI, and I did find some truth to that. Lexikon is a Wordle forever clone that is ad free and non-intrusive after my wife and I struggled to find a Wordle forever site that didn't shove a million ads down your throat.
Phoebe is a Plex / Jellyfin / Navidrome / Emby music player for web, desktop, and mobile after ditching Spotify and not being happy with the existing choices.
Andy is a desktop Android companion app that has a lot of features I found useful while working on Phoebe when I didn't want to open Android Studio due to its heavy resource consumption.
There was a nugget of truth to that earlier sentiment of using AI to create this very personalized layer of software that is useful to only me, that I'm not trying to sell or force anyone else to use. But it makes me examine the cost it took to get this software on my devices.
Not just the financial cost, but the data centers it took to run these AI models, and the data that these AI models were trained from, almost certainly illegally.
And then there is the internal and intrinsic cost to me. I have no way to quantify this (others do, though), but my critical thinking skills felt more deteriorated while being laid off and working on these projects with AI versus if I had just played video games all day. I also lost a ton of human connection in the process. The friends, peers, and Slack groups I'd usually go to for assistance when I'm trying to get GPU acceleration working while mirroring emulators in Andy for example were gone. Why would I post a question to the internet void when the AI can do it?
I struggle with the ease that 'solving' a problem with AI brings versus talking it out with other humans. Ultimately, I miss the human connection before AI. There is a void where people used to ask questions that is filled with sycophancy from AI.
I have no real conclusion. I find AI to be inevitable in the short term, at least until the cyclical nature of technology rears its head and there is a return, in some form, to the way things were. I am not naïve though, I know these tools are not going away, but I am hopeful we are able to find the medium where coding with Claude feels analogous to coding with an IDE and the humanity isn't driven out of the equation.
