On AI
jcs.org ● Covered by 11 sources
A vintage-computing hobbyist admits AI coding tools now make him doubt every retro project he sees online. His worry: years of hand-built work getting drowned out by a week of vibe-coded copies.
Based on reporting by jcs.org — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
The essay opens with Andreas Kling, the SerenityOS creator whose live-coding videos first got the author hooked on watching someone type fast. Kling was already quick with Qt Creator's autocomplete, but four years later he switched to CLion and picked up early GitHub Copilot support. Skeptical at first, he came around within months, watching Copilot draft whole functions and seemingly read his mind.
The author took a longer road. No IDE, no LSP, plenty of doubt about AI after its early stumbles. Claude in a browser was the gateway, standing in for Stack Overflow searches, and then Claude Code changed things — handing it a folder of code and watching it read, modify, and even reverse-engineer a binary blob with no context left an impression. Local models running through Ollama on an M4 Mac Mini rounded out the toolkit. The point he keeps making is that you can use a technology without buying into the hype and grift that grows up around it, the same way people separate Rails or Rust the tool from Rails or Rust the community drama.
Meanwhile, his own project runs on a much slower clock. Since 2020 he's been recording himself programming on an 8 MHz Macintosh 512Ke, later a Plus, using THINK C 5 and typing everything on the machine's own keyboard — no modern computer touches the code. That constraint produced an IMAP client, a revision control system, a Wikipedia reader, an IRC client, a BBS server, and, after adding Wi-Fi and a TLS offloading setup to handle modern encrypted traffic, a browser project started in 2024 that also picked up Gopher and Gemini support.
A couple of months ago the frustration boiled over in a post he wrote: new vintage-computing projects now read as hollow to him, because he assumes AI generated them before he even finishes the sentence. He compared it to Marge losing a go-kart race — spend a year building something no one's brought to an old platform before, then watch someone else use Claude to blow past you in a week with ten times the features. He'd seen exactly that pattern show up twice recently, a Mac OS 9 browser with CSS and JavaScript support and a System 7 app for monitoring 3D printers, both of which would have thrilled him a year earlier and instead got scrolled past.
Then came the kicker: a vibe-coded Gopher client for classic Mac OS showed up, complete with a slick AI-generated website, built partly from code lifted out of his own projects. Around the same time Kling posted that "enjoy writing code by hand" is shaping up to be 2026's programmer virtue signal. The author doesn't fully buy that framing, but it pushed him to interrogate his own irritation — whether it's really about craft, or about vanity, about wanting his years of labor recognized before something faster and shinier steals the room.
He lands somewhere pragmatic rather than bitter. If a vibe-coded tool is useful and available now, while his version is still years out, that's not obviously bad for the small community using these old machines. He can keep building his own version at his own pace and ship it when it's done — after all, whatever Claude generates has to be trained on code from somewhere, his included.
My take — AI-written commentary, not fact-checked reporting
The honest bit here isn't the AI skepticism, it's the admission that the anger was mostly about ego, not principle. Plenty of hobbyist and open-source communities are about to have this exact reckoning: years of patient, unglamorous work suddenly matched or exceeded by a stranger's weekend with an AI assistant. Gatekeeping the method won't hold, because usefulness to the end user doesn't care how the code got written. The people who adjust fastest will be the ones who stop treating hand-crafted effort as the whole point and start treating it as one valid path among several.
Read more about this at: jcs.org
