Name another approach (you can’t, because there isn’t one).
You can use the best engineering judgment you have, and you’ll never get an answer better than “I’m like 99% sure that isn’t necessary” (if anyone says they’re 100% sure, they’re ignorant or lying).
You have to pick a confidence threshold. Do you pick 90%, 99%, 99.99%, 100%? Too high, and you’ll have false negatives. Too low, false positives.
Musk’s point is that if you never have false positives, then the threshold is set too high. The false negatives have a cost too, it’s just less visible. So set the threshold so that you get some small number of false positives. The less the cost of a false positive, the lower the threshold should be.
I don’t know if any of the software I’ve worked on is considered “large” (I actually fucking hope not), and I have thought “Yeah, let it burn down, fall into the swamp and we’ll rebuild it.” but I wouldn’t decide that the place I need to try my idiotic ideas out is on the entire user base. It’s like the man has no idea of what a limited roll out, beta environment or staged decommission could be.
That’s not the methodology. You are still using your best judgment and evidence to identify the unneeded pieces. The question is about your level of confidence. Do you set it high enough that you’ll never make a mistake? Or do you set just a bit lower so that, say, one out of ten deletions have to be reversed? The difference in the false negatives rate will be enormous.
And the actual positives could be small, but have a large impact on security. You SHOULD have the ability to asses what the microservices do, and how they will impact your customers before you decide to turn them off.
Great. So what if you don’t? What if there’s a bunch of cruft put in by employees that haven’t been there for years, for which source control archaeology is giving limited results, etc.? I run into this stuff all the time at work. In fact was just the recipient of a service that got shut off because the providers thought it wasn’t being used. I was annoyed, but could hardly blame them, and it got reenabled in a state better than it was.
That’s great in a perfect world where every service has perfect documentation right from the start and you never have to depend on institutional knowledge. In the real world, documentation sucks, when it even exists, and infrastructure grows organically, often far out of bounds of what the authors ever intended it for. You’re left with spaghetti infrastructure with no real owners, and at best some maintainers that got coerced into the job because the thing they were really trying to do depended on it.
It’s interesting watching the popular consensus narrative shift to whatever confirms most people’s existing priors, independent of factual accuracy (eg: the emerald mine stuff that now gains an outsized prominence in his story).
What’s happened to Musk is not unique and not not especially mysterious. Musk made his fortune making bold, asymmetric and correct bets. He was both smart and incredibly perceptive of the world in a way that allowed him to make inferences several steps ahead of other people and then had the personality and corporate resources to bet big on where the world was headed. It’s a difficult thing to do because you have to have courage in your convictions and any move you make, you will have very smart, very thoughtful people give a convincing laying out of conventional wisdom as to why your asymmetric bet will probably not pay out. Your entire life is the entire world telling you everything you’re doing is incorrect so you learn to trust nobody but your own gut instincts and the infrastructure of independent minded people you trust to refine those gut instincts.
However, the problem with this kind of investing strategy is that you need all 3 of the characteristics to apply to continue making money. And what almost inevitably happens as you get richer is that you stay the same level of smart but you start losing touch with the perceptiveness that made you the money in the first place. The money insulates you from the outside world and gives you a false perception on the lives of everyday people and slowly, your inner circle of people you trust to bounce ideas off of gets replaced from the original independent thinkers into a coterie of yes people and functionaries who have no incentive to ever push back against you.
Look at Mark Zuckerberg, everyone was convinced that his acquisition of Instagram for $1B and WhatsApp for $22B were crazy overpaying and a total waste of money. No normal CEO could have pushed that through because any “rational” analysis of the situation could have never justified those moves. But he had the courage of his convictions over where the future of social was headed and he had the corporate structure to allow him to make those moves. Fast forward a decade later and he’s doing the same thing with the metaverse, his courage in his convictions are making him see a path there that nobody else is seeing except now he’s wrong and because he’s become high off his own supply and his reasoning process has been hijacked by a small group of insiders who are off in crazy land.
You saw the same thing with Steve Jobs, twice. By the end of his first tenure at Apple, he was off into wacky land making products that appealed to him but wasn’t what Apple needed at the time, he went off into the wilderness for a bit and went even more crazy with NeXT computers, obsessing about stuff like the factory the computers being made in being spotless and it was only the failure at NeXT that made him reflect and come back to Apple a second time with renewed correct judgement. But towards the end of his tenure at Apple, you could see the same reality distortion field working on himself again and Apple living with the legacy of a few of his bad decisions (like the ultra-thin, USB-C only macbooks or the trash can mac pro) for a decade before it managed to correct itself.
You saw the same things from the likes of Howard Hughes and a bunch of other historical business figures that I can’t name off the top of my head, it’s a occupational hazard for any business person who makes their fortune on bold, asymmetric and correct bets.
That’s part of it, but the other part is just dumb luck. Which isn’t to discredit the process; it’s just a simple fact that not all bets will pay off. As you noted, Steve Jobs made some dumb bets and then came back and made some great ones. Did his judgment actually improve, or was it just that any one of his ideas had about 50% odds of being decent, and therefore some hit and some didn’t?
Although it’s undoubtedly true that someone that starts by bucking the trend is likely to fall off over time as they start to drink their own Kool-Aid, I’d suggest the more dominant factor is just that bold bets are also risky ones, and not all are winners. Some dumb-looking ones pay off handsomely and some smart-looking ones don’t, and there’s no magic oracle to tell which is which.
Yep, have dealt with that before. I was tasked with debugging a bunch of nonsense code that had comments like “This is beyond stupid, and we should make the daemon not expect this to ever work. I don’t have the time to fix it now” and another program that I swear was written by meth addicts that started with “aa” as its first variable and then looped around to using “AA”, etc. Around halfway getting my mind around it, the person managing the whole situation said “OK, we’re burning it down and letting it fall into the swamp. Go work on making this usable instead.” But even then, the new hotness was rolled out on people that we could convince to go to it for a couple of years. Eventually the old system wouldn’t run on the new hardware available.
So yeah, if you really think you have to destroy the village in order to save it, do it in the most graduated method you can. Maybe Musk is really cash strapped and must take drastic actions to keep the whole affair afloat, I don’t know enough to comment on whether his pace is necessary or not. His changes so far seem unnecessarily painful and disruptive to his user base to this observer.
Well, they’re obviously disruptive. But… so what? It’s Twitter; planes aren’t going to fall out of the sky if there’s a 10-minute outage. This stuff happens all the time anyway (often because some rickety old infrastructure went down on its own).
There’s an image problem I guess, that it looks chaotic from the outside. But ultimately I expect that in a month or so, things will have settled down, and people will have largely forgotten who the owner even is.