Some feature requests were a matter of days, like user profiles: https://news.ycombinator.com/item?id=481

A brief essay on some HN design decisions: https://pastebin.com/bSW5dfRQ (from https://news.ycombinator.com/item?id=8424502)

I think the arc codebase is worth studying and understanding, primarily so that you can extend its simplicity into your own projects. The reason HN was such a success is because it handles so many cases in the same way: Stories, comments, and polls are all the same thing: items. If you want to add a new thing, you just create a new item and add whatever fields you want.

These rapid prototyping techniques have downsides, but the cure is to keep in mind what you can't do. (For example, you can't rename item keys without breaking existing items, so be sure to choose good names.)

Like emacs, HN's design is borne out of simplicity and generality. It's what you get when you write the next most important feature as quickly as possible, then cut as much code as possible, every day. Both halves are equally important.

It's fine to say that our modern applications are so much more complicated that the old lessons don't apply. And in extreme cases, that may be true. I don't think SpaceX has the luxury of rapid prototyping their software.

But the typical app is CRUD. For those, data processing flexibility is perhaps the most important factor in whether you can write new features quickly. And since code is data, a lisp master can write systems with a shocking number of features in shockingly few lines of code. (See Jak'n'dexter: https://all-things-andy-gavin.com/2011/03/12/making-crash-ba...)

That they keep changing (many times against express complaints of their users) is not reason for excusing them a bad experience.