A Series of Vignettes From My Childhood and Early Career

Jason Scheirer opens with a warning he got around 1996: object-oriented programming would soon let one genius write every library once, and everyone else would just snap components together like LEGOs, no programmers needed. Nearly 30 years later he’s still writing software for a living. He stacks five more vignettes on top of that one, each a moment where a new tool or a new hype cycle promised the end of his profession, closing on the LLM wave as simply the latest round of the same prediction.

The piece tracks a pattern he’s watched repeat since middle school: multimedia was going to make every tool obsolete overnight, IntelliJ’s refactoring tools were going to divide developers into haves and have-nots, the dot-com crash was going to kill the internet as a business. Each time, the technology stuck around and got boring, and the collapse never arrived on schedule.

Fichamento

#

• The 1996 warning specifically framed OOP as the end of programming: once a problem was solved by “someone much smarter than any of us,” it would never need solving again.
• The “Multimedia Age” of 1993 demanded every tool be multimedia-ready or get left behind. It fizzled into a <video> tag and normalcy; no industry collapsed, and audio/video specialists stayed rare.
• In 2000, a coworker showed him IntelliJ’s refactoring tools as proof programmers were doomed. Scheirer’s pushback in the moment was practical: the tool moved code around, it didn’t write new logic.
• As a teenager, he wrote a Python script that automated 85% of a contractor’s MUMPS-to-relational migration work. The contractor’s reaction was fear, not gratitude: “You didn’t show this to anyone else, did you?”
• He separately automated his own job of copying Excel sheets into HTML pages, using Windows Automation and scripted FTP uploads disguised with random pauses and typos to look human. The agency hired a full-time developer nine months later anyway; his managers just moved him to other work.
• He frames responsible data collection for NLP training, respecting robots.txt, identifying crawlers, securing explicit permission, as a discipline he practiced years before “AI ethics” became a live debate.
• The dot-com crash is his model for how hype cycles actually resolve: not a clean death, but a slow shift from “we should put this business on the internet because it’s coming” to “the internet is here, let’s build something that makes sense on it.”