The editorial highlights the top HN comment's framing: platforms didn't abandon the social graph by accident — they actively deprioritized it because friend content is inconsistent and posts irregularly, while algorithmically-surfaced fad content scales and retains attention reliably. This reframes the shift as a deliberate economic choice, not an emergent failure.
The BBC piece argues that with Meta disclosing ~50% of Facebook feed time and >30% of Instagram time coming from AI-recommended non-followed content — and TikTok/X/YouTube already defaulting to algorithmic surfaces — the friend graph has effectively died as the organizing principle of feeds. What remains is a ranker with a comments section attached.
By submitting the BBC piece and driving it to 598 points, the submitter amplifies the thesis that feeds are no longer social in any meaningful sense. The framing 'Anti-social: fads, not friends' positions algorithmic dominance as a categorical change in what these products are.
The article coins the second-order effect: everyone is exposed to the same trends, but privately, with no awareness that peers saw the same thing. Discovery rises while connection flatlines, because the feed delivers culture one-to-one rather than through visible peer endorsement.
The editorial argues the practical consequence for builders: 'build a social graph, monetize the network effect' no longer works because the unit of distribution shifted from your follower list to how well a ranker thinks your content embedding will perform. This explains why new accounts can outdraw 2M-follower creators and why creators now optimize for 'the algo' over their audience.
The BBC's Worklife piece — currently at 598 on Hacker News — puts a number on something developers have felt for two years: the share of content in the average feed coming from accounts the user actually follows has collapsed. Meta's own disclosures peg AI-recommended (non-followed) content at roughly 50% of Facebook feed time and over 30% on Instagram, and Mark Zuckerberg has publicly forecast both will keep climbing. TikTok started at ~100% recommended and never pretended otherwise. X's algorithmic 'For You' tab is the default. YouTube's homepage hasn't been a subscription feed in a decade.
The 'social' in social media is now legacy branding for what is, mechanically, a recommendation system with a comments section bolted on. The BBC frames it as the death of the friend graph; the HN thread (598 upvotes, ~700 comments at time of writing) is sharper, with the top comment arguing the graph didn't die — it was demoted, because friend content has worse retention numbers than algorithmically-surfaced fad content. Friends post inconsistently. Fads scale.
The second-order effect is what the article calls 'parallel monoculture': everyone watches the same five trends, but with no shared social context. You and your coworker both saw the same dance, but neither of you knows the other saw it, because neither of you posted it. The feed delivered it to both of you privately. Discovery is up. Connection is flat.
For anyone building consumer software, the practical implication is that the entire 2010s playbook — 'build a social graph, monetize the network effect' — is now a museum piece. The unit of distribution stopped being 'your followers' and became 'whatever the ranker thinks performs against your content embedding'. This is why a brand-new TikTok account can outdraw a creator with 2M followers on the same platform in the same hour, and why creators on every platform now obsessively talk about 'the algo' instead of their audience.
The technical shift underneath is unglamorous but worth naming. Recommendation at this scale is a retrieval problem, not a ranking problem — two-tower neural retrieval over content embeddings, ANN indices on billions of items, sub-100ms p99 budgets per impression. Meta's open-sourced work on DLRM, ByteDance's Monolith, YouTube's two-tower paper from 2019 — the public literature on this is now mature enough that a small team with a GPU and a good embedding model can build a defensible For You feed. The moat moved from 'who has the bigger graph' to 'who has the better embeddings and the cheaper retrieval'.
The HN discussion surfaces the uncomfortable corollary: if the graph doesn't matter, neither does network-effect lock-in. Users don't stay on Instagram because their friends are there; they stay because the ranker is good. That's a much weaker moat. Threads bled users back to Bluesky and Mastodon not because the graph followed but because the ranker on Threads is, by most accounts, worse. Reddit's recent algorithm changes — surfacing more 'recommended communities' over subscribed ones — generated the loudest organic backlash since the API debacle, because power users correctly read it as the platform admitting their subscriptions are no longer the primary signal.
There's a secondary thread in the comments worth pulling on: the death of asynchronous social context. When everyone saw the same posts from the same friends, you could reference 'did you see what Mark posted' as social currency. With personalized feeds, that's gone. Group chats absorbed the function — which is why Discord, iMessage, WhatsApp, and Signal have all quietly become the actual 'social' layer of the internet, and the public feeds are now broadcast television with a like button.
If you're shipping any product with a social component, three things change immediately. First, stop optimizing for follower count as a north-star metric; it predicts almost nothing about distribution on any modern platform. Optimize for content-level engagement (watch time, completion rate, save rate) because that's what every ranker actually scores. Second, if you have a creator-facing product, the analytics page needs to expose per-post reach distribution, not per-account follower growth — creators are flying blind on the metric that now determines their income.
Third, and this is the one most teams get wrong: if you're building a new social product, building a follower graph first is the wrong order of operations. TikTok proved you can launch a successful social app with no graph at all, because the ranker substitutes for one. The graph is now a retention mechanic, not a distribution mechanic — it's how you stop users from churning after the algorithm cools on them, not how you get them content. Bluesky's bet on a portable, federated graph is interesting precisely because it's trying to make the graph valuable again as infrastructure, but the jury is out on whether users care enough to pay the switching cost.
For backend engineers: the infra implications are real. A feed driven by friends-of-friends is a graph traversal problem; you can serve it from Postgres with the right indices and a Redis hot cache. A feed driven by personalized recommendation is a vector retrieval problem; you need an ANN index (FAISS, ScaNN, pgvector with HNSW), an embedding pipeline, a feature store, and a real-time inference path. The teams that retrofitted this onto graph-shaped infra (early Twitter, pre-2020 LinkedIn) burned years of engineering time getting it wrong.
The interesting question isn't whether the trend reverses — it won't, because the engagement numbers are unambiguous — but where the actual social interaction migrates. Group chats and Discord servers are absorbing it, but they don't scale to discovery. Some bet is on AI-mediated social: agents that summarize what your friends posted across fragmented platforms, or LLMs that surface 'people you should talk to' instead of 'content you should watch.' If that's where it goes, the next decade's social products won't have feeds at all — they'll have inboxes. And the developers who keep building 2014-era follower-graph products will find themselves shipping the social equivalent of an RSS reader: technically correct, structurally obsolete.
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