What's this about?
An argument that software should be designed for extension by its users now that anyone can generate an extension by asking. The thread produced two sharply different futures.
What exactly changes about extensibility?
Historically, extensibility was for power users: plugin APIs, scripting hooks, configuration languages — all requiring you to program. That limited the audience enough that most software didn't bother, and the ones that did (Emacs, Excel, browsers) built devoted followings. If generating an extension now costs a sentence, the audience for extensibility is potentially everyone — which changes what a well-designed application looks like. The counterargument in the thread is that this confuses capability with desire: most people don't want to extend their software, they want it to work.
What's the community saying?
- The optimistic future: GitHub becomes a platform for prompt-driven manual testing — you write a prompt, the AI proposes a change, you experiment with the UI and attach notes for the next iteration, with AI aggregating user requests into tickets and then into pull requests.
- The pessimistic future, and it's the more interesting one: developers get hired by clients arriving with an LLM-generated program that has reached the point where the LLM can no longer modify it without breaking things. The client's "requirements document" is the broken program.
- "Only power users want to extend software." The strongest objection — normal users want reliable software that does what they need, and building for extension may serve a market that doesn't exist.
- Multiple people had already done it. Several described building their own small tools in half an hour — an app launcher, an MP3 player — and preferring them to the mature alternatives, which is direct evidence for the post's thesis.
- The Cloudflare complaint: several felt the piece read as an advertisement for one vendor's platform, and doubted any single company becomes the default substrate for personal software.