Complex internet services fail in interesting ways as they grow in size and complexity. Twitter’s recent issues show how failures emerge slowly over time as relationships between components degrade. Meta’s quick launch of Threads demonstrates how platform investments can compound over time, allowing them to quickly build on existing infrastructure and expertise. While layoffs may be needed, companies must be strategic to maintain what matters most - the ability to navigate complex systems and deliver value. Twitter’s inability to ship new features shows they have lost this expertise, while Threads may out-execute them due to Meta’s platform advantages. The case of Twitter and Threads provides a lesson for companies on who they want to be during times of optimization.
@scrubbles My favorite early moment was him firing people based on lines of code written… which of course meant he fired all of his best because the worst programmers write many lines that do less while great programmers write few lines that do more.
The last few months have been me going back and removing code lol. Most of my time is spent reviewing my jrs code now, less even writing my own! But no he very smart for line count = good programming
Wasn’t that an old example of perverse incentives? IBM ranked or paid bonuses based on lines of code. In short order, all their code became bloated and inefficient.
This was an old example in the 90’s and maybe the 80’s, so could have been over of the other OG computer companies (Digital, Sun, HP, etc). Could also be apocryphal. Point is, it’s a classic example of dumb management ideas.
i remember a corporate rule came down that we needed something like 70% of all code unit tested for stability.
Damn were our getters and setters rock solid. No errors there. Business logic however…
Well, then the developers committed fraud, as getters and setters generally have very little logic. I’m surprised the code coverage reports failed to show the low coverage… You did have code coverage reports, rright?
Bruh it’s a story, not a trial.
I find it a bit obnoxious to claim unit testing is a waste of time and then point to worthless testing of logicless code as proof.
All that illustrates is that worthless tests are worthless. Basically, a tautology. If one wants to convince people that tests are worthless, show how actual test coverage added no value.
The reason most coverage requirements are about 80%, is precisely that testing should not be done on code that has no business logic, like getters and setters.
So, testing the one thing for which tests are worthless is fraudulent behavior and ironically just makes their own jobs that much more painful.
Yes. That was the joke of it all. That a useless business rule that came down made developers more focused on hitting a metric rather than building useful tests. Thank you for explaining my own story to me.
Aha, well I like to think I would have picked up on the joke if this was an in-person discussion. I’ve heard that talking point as a serious condemnation of automated unit tests.
Meh, the best programmers are probably somewhere in the middle.
This also depends on what kind of work you’re doing.
Writing some frontend with lots of Boilerplate? That’s lots of lines.
Writing efficient code that for example runs on embedded systems? That’s different. My entire master’s thesis code project on an embedded system consisted of around 600 lines of C code, and it did exactly what it should, efficiently.
A better metric to that effect would be the git activity graph. People that do important changes don’t commit 20 times a day - they push a commit usually once a day tops to once every 2 weeks