Meta’s philosophy has bit them before, but they at least do it better than anyone else. Other companies hear the Meta philosophy and their CEOs take that as an excuse to underfund development to the point of constant errors and shipping broken products.
They don’t seem to realize that the reason that Meta can operate that way is because they are / were relentlessly focused on figuring out why things broke and then building out new products and systems to let them keep working fast and breaking things without their being a big downstream impact.
They have incredibly robust testing, monitoring, and alerting systems in place for all of their products, including newly developed ones. They found it faster to work in a giant monorepo and share code, but they actually monitored and recognized when it scaled too big and was slowing development down and had teams building out custom version control software and virtual disk utilities to fix this (Microsoft did similar with Git when they moved Windows development to it), and when Meta found that coding in raw JavaScript and HTML was creating scaling difficulties with their app, they built React. Same thing with their customized version of PHP on the backend.
I don’t think Meta’s impact on the world has been positive, and I don’t think they should move fast and break things from a product design and ethics standpoint, but from an engineering standpoint, I do have respect for how they have executed that philosophy, and think that literally everyone else who tries it fails because they view it as a way of cutting short term costs, instead of as a way to identify and build and fix long term infrastructure.
The reason Meta could operate that way was because they were a platform for people sending funny texts to each other with no promises of security or privacy.
By the way, even they don’t operate like that anymore.
I’m basing that on my experience contracting there ~ 1.5 years ago. They’ve added new control systems to address things like the GDPR, but they are all still designed to be fully productized parts of their developer framework so that developers don’t have to think about them and can still move just as fast with product / feature development.
And while their product market had a little bit to do with it, they quite frankly have buggy software in production for less time than most major SAAS vendors or contract built systems.
As you noticed, they have had a quality assurance structure for way longer than 2 years. They’ve had it for close to 20 years now.
When they used to have this philosophy, they did always have something broken on their site, and go out of air once in a while. And they did benefit greatly from the speed they got from it, for a while, until it started being harmful.
Going slow doesn’t mean you don’t break things either. If you don’t want to break things, you need test plans, logging and alerts
Meta’s philosophy has bit them before, but they at least do it better than anyone else. Other companies hear the Meta philosophy and their CEOs take that as an excuse to underfund development to the point of constant errors and shipping broken products.
They don’t seem to realize that the reason that Meta can operate that way is because they are / were relentlessly focused on figuring out why things broke and then building out new products and systems to let them keep working fast and breaking things without their being a big downstream impact.
They have incredibly robust testing, monitoring, and alerting systems in place for all of their products, including newly developed ones. They found it faster to work in a giant monorepo and share code, but they actually monitored and recognized when it scaled too big and was slowing development down and had teams building out custom version control software and virtual disk utilities to fix this (Microsoft did similar with Git when they moved Windows development to it), and when Meta found that coding in raw JavaScript and HTML was creating scaling difficulties with their app, they built React. Same thing with their customized version of PHP on the backend.
I don’t think Meta’s impact on the world has been positive, and I don’t think they should move fast and break things from a product design and ethics standpoint, but from an engineering standpoint, I do have respect for how they have executed that philosophy, and think that literally everyone else who tries it fails because they view it as a way of cutting short term costs, instead of as a way to identify and build and fix long term infrastructure.
The reason Meta could operate that way was because they were a platform for people sending funny texts to each other with no promises of security or privacy.
By the way, even they don’t operate like that anymore.
I’m basing that on my experience contracting there ~ 1.5 years ago. They’ve added new control systems to address things like the GDPR, but they are all still designed to be fully productized parts of their developer framework so that developers don’t have to think about them and can still move just as fast with product / feature development.
And while their product market had a little bit to do with it, they quite frankly have buggy software in production for less time than most major SAAS vendors or contract built systems.
As you noticed, they have had a quality assurance structure for way longer than 2 years. They’ve had it for close to 20 years now.
When they used to have this philosophy, they did always have something broken on their site, and go out of air once in a while. And they did benefit greatly from the speed they got from it, for a while, until it started being harmful.