Stephen, why are you so critical of Facebook and not the other big tech companies?
A 🧵 for well-intentioned engineers about how to navigate the complexities of big tech.
There's a simple inescapable truth about the distinction in kind between Facebook and the others.
Google could fix its content moderation and stop its military contracts and its business model would still be Google. (1/)
Apple could fix up its supply chain, raises the unit prices on its products for more sustainable and environmentally sound sourcing, and its business model would still be Apple. (2/)
Amazon could treat its workers better and stop self-dealing in adjacent markets with its own products, and its business model would still be Amazon. (3/)
Netflix could ... actually don't really have here to complain about here really. Maybe less Tiger King. (4/)
That leads us to the key point, the subtle point that makes all the difference. All companies have psychopath executives, all companies cut contracts with unsavoury third parties, all companies have employee issues, all companies have ethical problems in their supply chain. (5/)
These things are not immutable facts about a company, they can and regularly do change. What *cannot* change is its core business model. The raison d'etre for its existence and the mechanism by which it continues its own existence, pays its shareholders and issues bonuses. (6/)
And that leads to Facebook, the singularly worst company in the tech sector ever. Its business model is to strip mine data from your relationships to your friends and family. It does this by showing you addictive, misleading and divisive content to optimise your screen time. (7/)
Their ambition is to create an ML-driven driven system designed to optimize your addiction to its apps to extract even more time and data from you. There are no limits or bounds on what it will show you to maximize your addiction to its platform. So long as it makes money. (8/)
Facebook is destroying the very fabric of human relationships, the democratic functioning of our state, and behaves like a dealer peddling in an addictive drug called anger. (9/)
Hell, Facebook even wants to issue their own sovereign currency so that they can track every cent of every purchase you make globally. To become a shadow bank that integrates purchasing data with behavioural data, to finally complete their user manipulation loop. (10/)
As the Wall Street Journal reported, any effort to reform moderation is second to one thing: user engagement. A public corporation must show quarterly growth, and Facebook's growth is built on ever-increasing user engagement to toxic content. (11/) wsj.com/articles/faceb…
If you work at Facebook, you dump poison into the body politic of society every day you log into Workplace. Every Phabricator task is a step forward in Facebook's executives unbounded psychopathic ambition for nightmare surveillance capitalism based on digital addiction. (12/)
So, no Facebook is not "just another company" like the other tech giants. The ethical distinction is one in kind, not degree. Its business model is corrupt to the core. It is the Big Tobacco of the information age. (13/)
There can be no redemption or internal reform for companies whose business model is based on human suffering. And the only answer is burn it to the ground, salt the earth and ensure our children's generations never builds the same horrors. (14/)
/end rant
• • •
Missing some Tweet in this thread? You can try to
force a refresh
LLVM is really great for languages that to first approximation are sugar on top of C (like C++, Rust, Swift etc). If you don't use System V calling convention, have non-standard register use, closures, or objects with dynamic lifetimes then it's basically just a clever assembler.
It's absurd to propose that somehow LLVM has solved compiler backends (because of Rust handwaving), because it absolutely hasn't. It's an amazing toolchain but it is by no means the end of history and most of the interesting work is still open problems for the non-C family.
You can get quite far with the approach where you write the whole runtime in C, compile out to IR and then let the interprocedural optimizer and LTO have at the final artifacts. But that's very brittle about guaranteeing performance and has a ton of compile-time overhead.
So let's finally talk about Coinbase. Let's not mince words ... they are a venture-funded gambling company. (1/)
But Stephen, Aren't they a bitcoin exchange?
Yes they are. And bitcoin IS a gambling product. It has no fundamentals, no exposure to the economy and trading it is a negative sum activity. It is a greater fools gold. (2/)
When you invest in a non-productive asset that isn't tied to any economic activity, the only thing you can do is try to find someone dumber than you to pay more for it. Nothing of value is created in this transaction. (3/)
So let's talk about "no-code" this morning. Sure, it's a silly buzzword and pretty much every time our industry has tried this for the last 30 years it's always ended badly. But why? (1/)
We can go back to the IT literature from back in the 2000s and it was clear that Forrester et al thought that developers were soon going to be a thing of the past. Frame-based development and GUIs would enable management to simply encode their business intent directly. (2/)
Truth has a way of asserting itself and in hindsight the entire premise was absurd. The most succesful examples from that era were things like FrontPage ... which in hindsight was a generally terrible idea. (3/)