Hacker News Reader

Comment by poisonarena | Hacker News Reader

[flagged]

This is such a disingenuous take. Twitter isn’t just a web page that displays the last thing saved to a database. It curates a different version of that for the millions of users reading it. That’s before you think about Spaces, the whole advertising platform that supports it, the abuse systems, and the infrastructure required to do all of that at a scale far beyond some bootcamp clone.

HN discourse should be above this.

Just to give one example, I (and many others) have tens of thousands of blocked accounts. How do you implement that efficiently, at scale? Are you gonna do an extra database check for every single tweet that comes my way? Hopefully not.

And if you think about how Twitter works for five seconds, just consider everything that has to happen when an account with 100M followers posts a tweet. It's an absolute nightmare.

A bit off topic, but any book/article recommendations for building things at Twitter/IG scale? I know for example IG switched to a "fan-out" approach early on to handle accounts with a lot of users. Would love to learn more about this.

In some previous HN thread, I dialed in on a recommendation for "Designing Data-Intensive Applications" by Martin Kleppmann, which I honestly still haven't gotten around to reading so....YMMV. Maybe if you search HN for that book title, it might lead you to other useful discussions?

Out of curiosity, what do you mean by tens of thousands of blocked accounts? Are those your bot accounts that get blocked or it is you blocking people from your feed?

I think they’re referring to accounts that they blocked, so that content from those accounts wont show up in the feed.

> HN discourse should be above this.

vomit inducing "Be better" post..

"I disagree with you so your comment shouldn't be allowed here"

back when I was doing personal projects or working at small early-stage startups I used to think the same, but then I've worked at later stage scaleups and even large publicly traded companies and realised that once a project becomes large enough, that prototyping phase that you just described is like 1% or 2% of the total workload. There is so much more to the SDLC that is not just writing code (some examples: liaising with stakeholders, requirements gathering, clients acceptance testing, QA, integration testing, certifications, regulatory frameworks etc. etc. etc. -> all things that take much of your time and involve nearly 0 lines of code written).

> liaising with stakeholders, requirements gathering, clients acceptance testing, QA, integration testing, certifications, regulatory frameworks

Some of that sounds like bloat to me. The fact that Twitter is still here and building features faster with 20% of the workforce may indicate that a lot of what was thought needed really wasn't. You might point to the current outage or other bugs but you have to ask yourself, how serious is it really and is it worth 400% more employees to mitigate them?

None of that is bloat. It's all necessary once you get past the tinkertoy stage.

Also, they aren't building features. The only features that rolled out are the ones that were already done. They haven't released anything.

Long form video, Twitter blue at the very least, which seems like more than Twitter managed to do per year before then with 4x the manpower.

Many of those things you listed are what people who have been in the institutionalized employees find important, essentially process for its own sake.

Before the layoffs Twitter had about 7500 employees, while SpaceX at 12000 builds frickin' rockets, have rocket launches every week, designing and building the next level launch platform that will revolutionize space travel, all the while revolutionizing satellite internet. Musk knows a thing or two about what smart, focused people are capable of.

So you're saying that at SpaceX or Tesla people don't work on any of those things (requirements gathering, clients acceptance testing, QA, integration testing, certifications, regulatory frameworks) and they just happily code away?

I very much doubt that's the case.

> thousands of people in bootcamps are making twitter clones

Heh Heh Heh

Those twitter clones will tend to be of the "looks pretty on the front end, but a fucking nightmare on the backend" type. ;)

Likely with part of the backend nightmare being security issues as well. From what I've seen of bootcamp output, they seem to rarely explain much about security to the people doing the learning. :(

That's not even the problem. The Stack Overflow guy wrote a very informative response to people claiming they could clone SO in a weekend (I can't find it now)

The core of the problem is not even the quality of the code but rather the thousands of things the site does besides letting you post a tweet and displaying it to other people.

Spam protection and protection from people falsely claiming spam on posts they don't like is already enough for me to laugh off the "clone in a weekend" camp.

What's stopping you son?

If it's capital, just get an SBA and start working.

Well that's the easy part, and probably like half a percent of the work that goes on at a scale like that

Apparently you do, given that we're posting in a thread where Twitter is self-imploding due to amateurish technical decisions foisted down by a God King.

Reply on news.ycombinator.com