Blain having a slide at a conference saying 'Scaling Rails. Its Easy. Really', before they had all those problems, was, well, sorta asking for it.
And anyway it was 2 years ago, could he have learnt from it? This is, after all, a new medium etc. etc. so a bit of trial and error is expected. As this post points out, the hit rate to date this year is almost perfect.
(edit: I just re-read that post and boy it was a bit harsh, the comments are terrible (from both sides - a lot of grudges are being held over this, I feel).
Doesn't convince me to write-off any trust in TC though, especially not two years later. If I ignored every source that wrote a bad story I would be left with nothing atm).
It'd be nice if there was any form of data to support the position of the article. I know that among my non-technical friends there seems to be no worry about algo feed vs time-based feed - and there doesn't seem to be any slow down in their scrolls through their Facebook/Instagram/Twitter feeds. I get wanting to call a change, but lacking any data it's hard to take seriously.
It was during the time of posting, which kind of proves my point. It is gamed. A statistically random sample of users would have yielded the same results.
I was very ready to suggest this must be an old article, but no it really is only six hours old. I suppose doing good statistical work takes time, but this feels like the obvious truth that every single product manager has known since KPIs were invented for social media engagement - anger and fury are the best for engagement.
This isn't news either. Ryan Holliday's "Trust me, I'm lying" talks about this in depth and was published in 2012. I am quite sure folks before him discussed this as well.
We could fix social media tomorrow if we disabled resharing/retweeting/whatever and removed likes, make timelines/feeds strictly linear time based. You want to be an epic troll? Fine, copy and paste your content in. That's enough of a barrier that social media would go back to being reasonably decent.
Seems like the author just completely made up the numbers here, as well as all underlying technical assumptions. In reality, based on previous public information, Twitter runs most of their infra from their own datacenters, and the number of tweets per day hit 500 million back in 2013.
Not to mention, the author also completely ignores read traffic (which is the majority of traffic for a social network), replication/redundancy/HA, so many other things. It's just a garbage article. It appears the company hosting this blog pays randos to submit posts on any topic? https://www.cohesive.so/write-for-cohesive
The primary source is the liked Twitter thread. I wonder how credible this source is. (I'm not familiar with the norm of ML community - They seem to be Twitter-heavy than other part of tech.)
I can't find a reference off the top of my head but earlier this year I saw a journalist post their engagement data on Twitter and it was shockingly bad. If I recall correctly they had hundreds of thousands of views and hundreds of click-throughs to their website.
People like to dog on their early marketing material, but it's not exactly a far cry from Animojis or Samsung's equivalent Memoji garbage. In this case, it looks like Meta took the criticism to heart and used it to develop something much more advanced.
You can see where some of the HN commenters get their outrageous takes from. You can "reverse engineer" Twitter in a month if you had .01 [1] the traffic and none of the advertising elements.
Here's an even less popular take: as someone just casually browsing twitter, it seems to work pretty much the same or slightly better than it used to, with the same (or slightly better) content. The fact that they achieve this with one quarter (!) or so [*] of their previous headcount is truly amazing, and more interesting than some of the other takes media tries to spin on it.
[* by some accounts one tenth -- this is even more amazing if true]
> Twitter has several Candidate Sources that we use to retrieve recent and relevant Tweets for a user. For each request, we attempt to extract the best 1500 Tweets from a pool of hundreds of millions through these sources. We find candidates from people you follow (In-Network) and from people you don’t follow (Out-of-Network).
> Today, the For You timeline consists of 50% In-Network Tweets and 50% Out-of-Network Tweets on average, though this may vary from user to user.
It would’ve been interesting to see what changes were made since Musk’s takeover. As someone who followed 5,000+ users, I know I never saw a tweet that wasn’t either from nor retweeted by someone I followed — e.g. I never saw those “[user you follow] liked [someone you don’t follow] tweet”
50%/50% in FYP seems to reflect my experience today — which is much worse, to the point that I’ll regularly switch to viewing by List b/c I miss seeing people who I want to read.
I wonder how much testing and analysis went into deciding on the 50/50 ratio — e.g. how does it impact user engagement and behavior. Because it sounds like an easy round value that you’d land on when thinking “users should be pushed out of their bubbles”
Correct, I do not think it is realistic to expect Twitter to behave differently. I do not believe it's Twitter's behavior that the original article is trying to highlight.
A) A good hunk of the article is dedicated to arguing that the people Twitter don’t understand why their algorithm chooses one tweet over another.
B) Even if they do release the entirety of the algorithm, that doesn’t mean anything is going to get better. Likely it will get worse. The scourge of SEO optimization is already having detrimental impacts on finding good information. Imagine the kind of garbage we’ll have to sift through if the people gaming the system are able to see exactly what the system is.
> I'm sure Twitter is well aware of how their algorithm works, and I'm sure it yields the best results for the majority of their users, but I apparently don't belong in that group.
The algorithms are optimised to crank up the engagement metrics. Though that’s certainly good for Twitter, whether that’s actually the best thing for their users is up for debate.
That's the problem with metrics that only measure part of the situation. Maybe having cheap, insincere boilerplate does increase the number of people following through to one's Twitter page, something that is easily measured. But it can't measure the number of people who lose trust in you due to it.
I don't think these graphs are (in all fairness) accurate. Twitter by all accounts has had about a year+ headstart on Friendfeed and if you compare the two assuming the same starting point, Friendfeed are doing OK.
I agree with Paul, it takes years of effort to become an overnight success.
> The extrapolations there are "limit on the post size" and "this funny algorithm".
I disagree. "Limit on the post size" was not arbitrary, it was a technical limitation - Twitter was created as "a social network, but for SMS". SMS was a well-known technology back then. As for "this funny algorithm", in context of the dialogue 'endisneigh posted, the algorithm is just an input. "Hey, AI, use that thing as a ranking function". The "extrapolative" work would be the invention of the algorithm itself.
Blain having a slide at a conference saying 'Scaling Rails. Its Easy. Really', before they had all those problems, was, well, sorta asking for it.
And anyway it was 2 years ago, could he have learnt from it? This is, after all, a new medium etc. etc. so a bit of trial and error is expected. As this post points out, the hit rate to date this year is almost perfect.
(edit: I just re-read that post and boy it was a bit harsh, the comments are terrible (from both sides - a lot of grudges are being held over this, I feel).
Doesn't convince me to write-off any trust in TC though, especially not two years later. If I ignored every source that wrote a bad story I would be left with nothing atm).
reply