[tptacek]

A piece of software like Pinboard can stay running on autopilot for a very long time, because it's simple (that's a compliment, not a dig). A piece of software like Twitter, comprised of a bunch of random interlocking services, is much less likely to stay up unattended. All the random stupid shit that might lock up Pinboard once a year (a full log drive, a run of block errors on a storage device, a network hiccup long enough to blow through a timer, &c) happens to all the services; most of these systems aren't aerodynamically stable: once something burps, the whole system oscillates and wedges until someone goes in and unkinks it.

(Maybe this has been my experience over the past 2 years because our team is just bad at this stuff, but when I relate it to other people working on software like this, I mostly get commiseration).

I'm not taking a position on whether Twitter needs 3000 more engineers to stand a chance of running. I think reports of the death of Twitter are also exaggerated, and I remember reading Dan Luu about bloat at Twitter long before Musk expressed an interest. I'm just saying: it's shocking that the service is running as well as it is right now. Some group of people on that team did a lot of stuff right.