4 ms·

Oddly enough I don’t see this being discussed here. There hasn’t been growth in user counts that might bring large technical problems. Sure there’s going to be

Oddly enough I don’t see this being discussed here. There hasn’t been growth in user counts that might bring large technical problems. Sure there’s going to be work needed, optimizations, new features or whatever. I think twitters problems are how to grow users and grow revenue, neither of which require technical solutions to get traction going.

charcircuit 4y ago
Operating with 80% less humans is a technical challenge.
XorNot 4y ago
Twitter is a service which handles bursty type demand. One of these events is the World Cup, which starts on Sunday. Which Twitter engineering has previously noted is one of their most challenging demand times.
- avereveard 4y ago
  I think this is super weird. massive events happen every so often. the first one is a challenge, the second one might bring a new set of challenges, but by the third one you haven't a well oiled process, then there's some serious engineering issues that go deeper than the challenge itself. this is something that was talked well before Elon, they have thousands upon thousands engineers, for a plafromt that had a fixed set of features since forever. I don't claim that twitter could be run by the smart cousing from it's basement, but the effort/result numbers seem way off from here. I think everyone is conveniently setting this topic aside in the ferwor to pile up on Elon (whom actions I don't defent in any way, shape or form) but something in that area was in need of a correction for a while, and apparently the whole stack is politicized enough than steering the ship would have been challenging irregardless by whom would have taken such task.
  - mrguyorama 4y ago
    You should go work for a big retail giant during a christmas season. It's not the pure load that does you in, spinning up more clusters on AWS and bumping up the size of the instance is super easy. What kills you every time is the edge cases and "1 in a million" chances that the load brings. Every single bit of code you've written since the last huge season is suddenly very aggressively being tested in ways you couldn't really simulate. Sure you've fixed the problems that caused fires last year but there is ALWAYS new stuff to start fires. Even if you froze the code base from the previous year, you would still see new patterns and behaviors, and all the problems they will uncover. Twitter was definitely poorly managed, had no vision, and the engineers were probably closing lots of tickets with well written and documented and smart code. The problem is that if the tickets being created are stupid, no amount of "hardcore engineering" can fix that. You can't fix "draw a red line using this blue pen" with better coders. If you actually care about twitter (you shouldn't), what it needed was a plan to make people hate it less, since even the people who use it constantly seem to hate it, and a strategy to make things more reliable and cheaper, essentially pay off all the tech debt they accrued by doing everything before it was done. But that would have required a massive undertaking, and public companies generally hate that kind of thing, and would rather slowly die over decades than make things better by spending a couple bucks up front. None of that makes walking in, shitting on everyone's keyboard, pissing off everyone that could be talented, firing lots of people seemingly randomly or off the cuff, and saddling the company with a bunch of new debt they no longer have revenue streams for because your stupid feature idea was predictably stupid, a smart or correct or laudable action.

charcircuit 4y ago

Operating with 80% less humans is a technical challenge.

XorNot 4y ago

Twitter is a service which handles bursty type demand. One of these events is the World Cup, which starts on Sunday. Which Twitter engineering has previously noted is one of their most challenging demand times.

avereveard 4y ago
I think this is super weird. massive events happen every so often. the first one is a challenge, the second one might bring a new set of challenges, but by the third one you haven't a well oiled process, then there's some serious engineering issues that go deeper than the challenge itself. this is something that was talked well before Elon, they have thousands upon thousands engineers, for a plafromt that had a fixed set of features since forever. I don't claim that twitter could be run by the smart cousing from it's basement, but the effort/result numbers seem way off from here. I think everyone is conveniently setting this topic aside in the ferwor to pile up on Elon (whom actions I don't defent in any way, shape or form) but something in that area was in need of a correction for a while, and apparently the whole stack is politicized enough than steering the ship would have been challenging irregardless by whom would have taken such task.
- mrguyorama 4y ago
  You should go work for a big retail giant during a christmas season. It's not the pure load that does you in, spinning up more clusters on AWS and bumping up the size of the instance is super easy. What kills you every time is the edge cases and "1 in a million" chances that the load brings. Every single bit of code you've written since the last huge season is suddenly very aggressively being tested in ways you couldn't really simulate. Sure you've fixed the problems that caused fires last year but there is ALWAYS new stuff to start fires. Even if you froze the code base from the previous year, you would still see new patterns and behaviors, and all the problems they will uncover. Twitter was definitely poorly managed, had no vision, and the engineers were probably closing lots of tickets with well written and documented and smart code. The problem is that if the tickets being created are stupid, no amount of "hardcore engineering" can fix that. You can't fix "draw a red line using this blue pen" with better coders. If you actually care about twitter (you shouldn't), what it needed was a plan to make people hate it less, since even the people who use it constantly seem to hate it, and a strategy to make things more reliable and cheaper, essentially pay off all the tech debt they accrued by doing everything before it was done. But that would have required a massive undertaking, and public companies generally hate that kind of thing, and would rather slowly die over decades than make things better by spending a couple bucks up front. None of that makes walking in, shitting on everyone's keyboard, pissing off everyone that could be talented, firing lots of people seemingly randomly or off the cuff, and saddling the company with a bunch of new debt they no longer have revenue streams for because your stupid feature idea was predictably stupid, a smart or correct or laudable action.