I find I’m not super convinced that ‘aws exists’ or whatever solves the problems Twitter had. Though I guess a bunch of the problems with eg distributing a notification to millions of followers or super-deep reply chains or football games or new years or whatever can be punted on until you have significant numbers of users which fits ‘low barrier to entry’.
ETA: I want to be careful to say that it's not "lol easy peasy" to build Twitter, but that the question "How likely would an investor be to pass on your Twitter clone startup because they thought the technology couldn't be built by your team with their investment?" is laughable, possibly even if the founding team is non-technical. With $5m, anyone competent enough to even file a YC application could hire a good enough technical consultancy to get a Twitter clone that scales to millions of users.
Second, the toolchain has matured in absolutely huge ways. AWS exists. Twitter had to build their own cloud to meet scaling needs. Languages have matured or been purpose-written to enable building/scaling these types of systems. A huge portion of the N+1s and hidden footguns have been cleaned out of the thousands and thousands of open source libraries that you can glue together to get the system up and running.
Third, Silicon Valley engineers as a whole have spent the last decade-plus building all sorts of Twitter-adjacent Web 2.0 projects, so there's an incredibly deep pool of people that have extremely relevant experience, even if they never stepped foot inside the Twitter building.
Fourth, Twitter (and companies very similar) have been publishing literal engineering designs and post-mortems for public consumption on their engineering blogs. Even if you have no idea what HTML is, your path to self-education and building your own Twitter clone has never been shorter or better paved. They've even published huge chunks of their own work as open source projects.
And on and on and on and on. And here's the thing about growing to Twitter-scale: it doesn't happen all at once. You can build the low-volume version and just follow approximately the same technical scaling path that Twitter itself followed, except you'll be able to skip a whole bunch of mistakes.