Argues the shift from social to recommendation happened in production code between 2020-2023, one A/B test at a time. Cites Meta's 2023 antitrust filings showing only ~7% of Instagram time is spent on followed accounts, framing the follow graph as still present in the database but no longer the dominant ranking signal.
Contends that once TikTok proved strangers' content could be ranked better than friends' posts, the follow graph became a liability that caps content supply at the size of a user's social circle. A recommender drawing from the entire content pool will always have more retention headroom, so every platform that didn't switch lost.
Frames the change as a cultural shift where feeds are now dominated by strangers, brands, and trending content while friends barely appear. Names the phenomenon explicitly: fads beat friends, parasocial beats social, capturing what heavy users have felt for years.
Surfaced the BBC piece on HN where it gained 179 points and 143 comments, indicating strong resonance with the developer community's sense that the 'social' label on these platforms has become misleading marketing for what is actually a content recommendation product.
The BBC's *Worklife* piece this week put a name to something every heavy user has felt for two years: social media stopped being social. The feeds we open in the morning are dominated by strangers, brands, AI slop, and whatever the ranker decided is trending. Friends — the people we explicitly chose to follow — barely appear. The article frames it as a cultural shift: fads beat friends, parasocial beats social.
That framing is correct but undersells the mechanics. This isn't a vibes shift. It's an architecture shift that happened in production code, one A/B test at a time, between roughly 2020 and 2023. Instagram's `Reels` tab cannibalized the friend feed. TikTok's `For You` made the follow graph optional. Twitter's `For You` tab became default. Facebook's main feed quietly stopped being a chronological list of people you know. The follow graph is still in the database — it's just no longer the dominant signal in the ranker.
Meta said the quiet part out loud in its 2023 antitrust filings: roughly 7% of time on Instagram was spent on content from accounts users follow. TikTok has reported similar internal numbers. The product surface still says 'follow' and 'friends.' The backend is a recommendation system that uses your follow graph as one feature among hundreds, and not the most important one.
For engineers, the interesting question is *why* every major platform converged on the same architecture in the same window. The honest answer is that interest-graph ranking wins on retention against follow-graph ranking, and every platform that didn't switch lost. Once TikTok demonstrated that you could rank strangers' content better than friends could rank their own posts, the follow graph became a liability — it caps your supply of high-engagement content at the size of someone's social circle. A recommender that draws from the entire content pool will always have more headroom than one that draws from 300 followed accounts.
The other reason is supply-side. Friends post less than they used to. The behavior that fueled 2010-era Facebook — public status updates from people you know — collapsed into private DMs and group chats years ago. If you're a product manager staring at a feed where half your active users follow people who haven't posted in six months, you don't have a content problem. You have an inventory problem. The interest graph is what you reach for when the social graph runs dry.
What this produces, downstream, is a media environment that *looks* like social media — handles, profiles, follow buttons, reply threads — but functions like cable TV with a comments section. The content is professionally optimized, the audience is unknown to the creator, and the algorithm is the only thing standing between a post and oblivion. The BBC piece's complaint about 'fads' is really a complaint about this: when the ranker rewards format-matching ('do the trend, get the views'), creators converge on whatever the current template is. The platform doesn't reward originality. It rewards legible variants of currently-winning content.
The community reaction in the HN thread is telling. The top comments aren't nostalgic for old Facebook — they're describing how they've already adapted. Group chats, Discord servers, Signal groups, small Mastodon instances. Senior engineers in that thread weren't mourning social media; they were describing a re-decentralization that's already underway in private channels, just not on platforms that VCs can index.
If you're shipping anything that depends on social distribution — a side project launch, a blog post, an open-source repo, a newsletter — you should reason about the platforms you're posting to as classifiers, not as audiences. Your followers are a weak signal to the ranker. The strong signals are: does this post hook in the first second, does it match a format the model recognizes, does early engagement velocity clear the threshold to be shown to non-followers. Optimizing for your followers is optimizing for a feature that the ranker has already partially deprecated.
This has concrete implications. If you're building a product that markets on Twitter, the playbook of 'build an audience, then launch to them' works less well than it did in 2018, because the audience you build doesn't reliably see what you post. The replacement playbook is to optimize the post itself for the cold-start ranker — formats, hooks, mid-tweet engagement bait — and treat your follower count as a vanity number that mostly helps with social proof. Same logic on LinkedIn, TikTok, Instagram, YouTube Shorts.
For developers building social products, the architectural lesson is sharper: don't ship a follow graph unless you're prepared to also ship a recommender that can substitute for it within 18 months. Mastodon and Bluesky are running the experiment in public — chronological follow-graph feeds, no algorithmic ranking by default. The retention numbers compared to TikTok-class incumbents are not close. That's not a bug in their implementation; it's the math of the supply problem. A recommender with no inventory ceiling will always beat a follow feed when the active-poster ratio drops below some threshold.
The other thing worth internalizing: the 'dead internet' theory is wrong about bots but right about the underlying texture. The dead-internet feeling comes from interacting with a system where the entity on the other end of the feed isn't a person you chose, it's a model that chose you. The bots are downstream of that, not the cause.
The next move is already visible in beta. Meta is wiring AI personas into feeds. TikTok has tested AI-generated avatars. Both moves make sense from inside the architecture: if the ranker is the product and the social graph is vestigial, why constrain the supply to humans? The endgame of interest-graph social media is a feed of synthetic content tuned in real time to a model of you, with the occasional human post leaking through because it tested well. Whether anyone wants to use that product is a separate question from whether the platforms will ship it — and the answer to the second question is obviously yes. Build accordingly, and keep your real conversations in places the ranker can't reach.
Yes, the game is over, the corps have won. Where the Internet used to be a forum for creativity, it's now a weapon of influence. Where we used to have an anonymous (or at least pseudonymous) playground, we are now monitored more than anywhere else. Where we used to be able to genuinely connect,
If you're on Android, you can use revanced to patch social network apps, to, among other things, remove content from non-friends (and ads).It's scary how empty the feed is once you do this. It can be full days with the same post at the top. And the worst part is that I hadn't noticed
HN is social media. Social media is a spectrum.You can imagine HN like a documentary channel compared to Facebook’s reality TV, but even “documentaries” can be dopamine sinks that aren’t actually informative (or accurate).(But personally, I see lots of short and pure opinion posts here, documentarie
I've stopped using YouTube and Reddit since early April and it's been a mixed bag.On one side my interest level has adjusted so that normal activities make sense again - like sitting in the garden or playing a game with my kid. I've also completed dozens of projects like replacing old
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
This article has struck a nerve in the comment section. It's describing how traditional social media sites like Facebook and Instagram are not used for social features anymore, but for content discovery. The descriptions of how people are using Facebook to find new content anonymously are not t