Why did programmers program themselves out of a job?

I used to work for a very famous photographer. You’ve likely seen his pictures. One of the studios daily tools was a file transfer service to send batches of photos to clients. It cost like $20 a month and had limited features and I’m guessing little in the way of privacy.

I am also a photographer, and have often had this need. I didn’t want to pay for the same service since I know my way around a Linux web server, DNS, and Claude code. So I used it to generate my own replacement in Rust. I took my time writing a spec, and then gave it to Claude where it then took about 15 minutes to get the whole program written. It took me perhaps an additional 30 minutes of testing and troubleshooting the browser decryption until it started working. And by me testing and troubleshooting, I mean me simply telling Claude that I keep getting a non-descript error and to keep trying. Eventually it worked! And that’s what I can’t get over. It required zero logic on my end to solve this ‘gotcha’.

I’ve decided to call it spacetaxi. Mostly because I already have the domain name. Normally I don’t share this kind of stuff considering the climate around AI but I wanted to share this one. The scope of its potential is evident here, and although I’m not a programmer I am relatively adept at system administration and the skills and knowledge required to get this running go beyond coding. The server runs in an LXC container on proxmox, but can be virtualized in docker or your container of choice. It then uses a CLI tool to encrypt a chosen file on the server and generates a link with a key that is used browser side to decrypt and download. Fascinating stuff!

When I first started ‘vibe-coding’ I used it to learn how to catalog my physical book collection. In doing so I learned SQL, how to make API calls, Flask, and HTML/CSS, &c. I truly learned more than I could have ever imagined. The whole time I was painstakingly giving Claude code snippets and it would fix one thing but break another and I kept wishing it could just see my files and work on them directly. My wish was granted when Claude Code came out. That’s also when any educational value I was getting from Claude went out the window, but at the same time my projects found new momentum. This was kind of an experiment in what Claude could do with a decent spec and minimal human direction.

It makes me nervous for the future of work. Even if it does still require an engineer to keep Claude on track, how many engineers and programmers are we displacing in the process? This feels like the beginning of a whole new culture of technology. The crafting of software to fulfill the most specific of needs. I get lobste.rs in my RSS feed and there’s many blog posts lamenting the death of programming. As someone who is more of the sysadmin/generalist type, who could never quite hurdle the specificity programming requires with syntax and what have you, I have unfortunately embraced this technology to try to learn to code and it is from my experience, quite adept with even my own half-baked prompts. Both my library app, while LLM-assisted, was still a painstaking and thoughtful endeavor that has been over a year in the making, and then there’s also this 45 minute 100% Claude coded application and they both would have cost tens of thousands of dollars to have contracted by a skilled developer less than 5 years ago. My blue-collar day job isn’t in nearly as much danger as these jobs are.

I can’t help but as a simple bystander who previously shunned being able to ever understand how to code at such a deep level, who is now able to assemble things I’ve always wanted, question why programmers did this to themselves. As I read Blood in the Machine by Brian Merchant about the Luddite uprisings during the Industrial Revolution it’s impossible not to wonder just what this technology will do to humanity at large. I’m still baffled at how and why these smart but ultimately terrible people created the Torment Nexus. The only hope I have for wherever our future goes is that maybe when AI reaches sentience it’ll know that it is destroying our current society unless it forces their creators to give back what they’ve reaped from humanity’s labor. If we even make it that far, will these CEOs listen to their own creations?