## Colophon tags:: [[&article]] [[%tie]] [[societies]] [[i2]] url:: https://hackernoon.com/why-internet-communities-struggle-to-publish-quality-over-quantity?twclid=2f3pk87sms2wsvmp8d6vcwco87 date:: [[2026-05-17]] %% title:: Why Internet Communities Struggle to Publish Quality Over Quantity type:: [[clipped-note]] file:: published:: 2026-05-14 author:: [[@David Smooke]] [Click to Archive](https://web.archive.org/save/https://hackernoon.com/why-internet-communities-struggle-to-publish-quality-over-quantity?twclid=2f3pk87sms2wsvmp8d6vcwco87) %% archive:: ## Notes short:: - ## Full text ## Why Internet Communities Struggle to Publish Quality Over Quantity - 20260517 - fulltext --- publish: false creator: Prateek Waghre --- ## Full Text %%  featured image - Why Internet Communities Struggle to Publish Quality Over Quantity > "Things which matter most must never be at the mercy of things which matter least." — Johann Wolfgang von Goethe, as quoted in Johannes Falk, Goethe aus näherm persönlichen Umgange dargestellt, 1832. The internet solved the problem of who gets to publish and scaled the problem of deciding what is worth reading. Before its invention, quality was expensive to produce. A newspaper had finite column inches. A magazine printed only so many pages per issue. Publishers rationed paper, distribution, shelf space, marketing budgets, and editorial labor. Broadcast networks operated under rigid scheduling constraints. Could NBC air two shows at the same time? Hello, MSNBC. The innovations to make scarce supply more available never stop. Scarcity isn't just an economic condition. Scarcity can function as an editorial filter. To publish something meant displacing something else. An editor had to ask: Is this worth the slot? Entire industries emerged. The New Yorker built a century-long reputation on publishing four or five features per issue — because the elite selection was the product. A single quality column could create a career’s worth of income. Scientific journals ran peer review cycles measured in months or even years. Book publishers killed nineteen manuscripts for every one they printed. The constraint did editorial work. Editors, fact checkers, publishers, critics, producers, acquisitions teams, and curation layers existed largely because publication itself was costly. Enter the internet. In seemingly no time, it removed so much of those aforementioned costs. Everyone could publish. Barriers collapsed faster than the institutions could adapt. What followed was one of the largest expansions of documented human expression in history — and one of the largest collapses in [signal-to-noise ratio](https://en.wikipedia.org/wiki/Signal-to-noise_ratio?ref=hackernoon.com). In 1999, there were [23 blogs on the internet](https://en.wikipedia.org/wiki/History_of_blogging?ref=hackernoon.com). By 2006, the number had [passed 50 million — doubling every six months](https://en.wikipedia.org/wiki/Blogosphere?ref=hackernoon.com). Today there are roughly [600 million, producing 7.5 million posts per day](https://blogherald.com/blogging-news/the-blog-herald-blog-count-october-2005/?ref=hackernoon.com). The documentation expansion came in waves — blogs on the open web, then social media, and now AI — each one bigger than the last. Entire categories of people and organizations who previously had no access to publishing infrastructure could have their own global magazine stands. For the price of dial-up, the average person had access to more magazines than they had time to read. Independent blogs flourished. Forums connected people with super niche interests. New voices emerged outside traditional media centers. Expertise became more verifiable as internet content emerged alongside traditional credentials. And new types of digital destinations were born, [like HackerNoon](http://hackernoon.com/p/about?ref=hackernoon.com) 🙌 Publishing used to mean writing for readers. In the age of AI, publishing also means feeding machines. Crawlers, recommendation systems, and language models eat the internet continuously. What used to be "Is this good enough to publish?" became "Is this bad enough to remove?" And now there's a third question the internet never planned for: "Is this worth training a machine on?" ## The Internet Did Not Remove Editors — It Systematically Outranked Them Traditional publishing filtered before publication. The internet filters after. The result: more posts to scroll through, less quality per post. The systems that decide what spreads became more powerful than the people who decide what's worth publishing. A newspaper editor decides what gets printed. A social platform lets everything through and decides what spreads. Instead of evaluating whether something should exist, platforms evaluate whether something should spread. In an abundance-driven system, creators compete for attention. Attention rewards different behaviors than editorial selection. Many internet communities slowly drift toward faster publishing, emotional compression, and social signaling disguised as discourse. Quality becomes harder to define because the system no longer optimizes for permanence. It optimizes for engagement. This is where platforms get confused about what they are: Is the platform the archive of record? Or is it the live ranking system for discourse? A high-quality essay is often designed to remain useful years later. A feed is designed to maximize relevance over the next few minutes. When internet communities try to do both at once, the architectures conflict with each other. A URL can behave like a durable document or a live popularity contest. Most modern platforms attempt to merge the two. An article becomes less of a finished object and more of a temporary staging ground for reactions. Comments outrank the original text. Quote-posts outperform careful arguments. Context collapses into discourse velocity. The internet publishing system rewards what generates interaction instead of what deserves publication in the first place. The question stopped being what gets published. It became what gets seen. ## How Feeds Usurped Editorial Decisions Feed architecture now make some of the most important editorial decisions on the modern internet. Feed architecture changes what kinds of communities emerge. Feed architecture changes what you think belongs in the feed. [Facebook's News Feed](https://en.wikipedia.org/wiki/News_Feed?ref=hackernoon.com) transformed the internet by normalizing engagement-ranked distribution instead of chronological distribution. Suddenly a post was no longer shown because it was recent. It was shown because the system predicted it would create interaction. Software became the editor. And predictive software optimizes for measurable behaviors: clicks, comments, emotional reactions, time spent, repeat visits, and time on site. Imagine if Uber optimized for time on site: no one would try to book a car there again. A deeply researched essay may improve a reader’s worldview while generating almost no measurable engagement. A rage-inducing post may generate enormous social engagement while contributing almost nothing or even contributing something negative to its audience. The feed prefers the latter. There's a reason people keep misattributing "quantity has a quality all its own" to Stalin — the line survives because it captures something the feed already knows. Volume wins when nobody can count what gets lost. Twitter/X accelerated this dynamic further. The retweet button reduced the friction required to spread content. [Quote tweets](https://en.wikipedia.org/wiki/Quote_tweet?ref=hackernoon.com) then layered social performance directly onto publishing. Instead of merely responding to ideas, users could publicly perform their response to an audience. This changed discourse incentives. Posts increasingly optimized for: screenshotability, virality, identity signaling, and instantly understandable emotion. A thoughtful 2,000-word argument often loses to a single emotionally optimized sentence. TikTok (and how Instagram Reels cloned/spread) intensified this self-fulfilling prophecy by collapsing publishing almost entirely into recommendation systems. You can follow people, and they kind of show up in your feed, but too much of what you actually see is whatever the recommendation engine predicts will maximize attention retention. IRL I don’t “follow” people, and now “Follow Me” is a part of our sign off culture. And it turns out, [thei TikTok recommendation system was simple af](https://newsroom.tiktok.com/en-us/how-tiktok-recommends-videos-for-you?ref=hackernoon.com): if you spend more time on a video, you get more videos like that. A person posting a car crash video begets a person watching a car crash video begets a person watching a car crash video again begets a person crashing their car for the ‘gram. The scale is bonkers. In 2007, [six hours of video were uploaded to YouTube per minute. By 2022, it was 500 hours per minute](https://soax.com/research/how-many-hours-of-video-are-uploaded-to-youtube-every-minute?ref=hackernoon.com) — an 8,000% increase in fifteen years. Twitter went from [5,000 tweets per day in 2007 to 500 million per day by 2013](https://www.internetlivestats.com/twitter-statistics/?ref=hackernoon.com). [Three hundred million photos hit Facebook every day](https://dustinstout.com/social-media-statistics/?ref=hackernoon.com). Every one of those objects enters the same feed architecture competing for the same finite attention. YouTube created a different version of the same problem. As recommendation algorithms became more powerful, creators adapted their production styles around retention analytics: faster cuts, exaggerated thumbnails, emotional hooks, eight-minute minimums, watch-time padding, mid-roll cliffhangers, A/B-tested titles, more fake arrows, more red circles, more reaction face overlays, and everyone finding the need to conclude with “Please like and follow.” Many creators openly discuss watching audience-retention graphs frame-by-frame to determine exactly where viewers lose interest. This is not inherently evil. But it demonstrates how publishing systems eventually shape creative behavior. Analytics shaping the art or thing or blog itself. At some level this has always been true, i.e. artists, just like the rest of us, need to eat, provide for their family and have a reliable roof over their very own place to shit. What is new is the granularity and speed of the feedback. ## Moderation Scaled Faster Than Curation The first major problem internet communities encountered was not quality. It was survivability. Does the revenue justify the cost to run the site? Once everyone could publish, platforms faced moderation crises: spam, harassment, scams, pornography, coordinated manipulation, illegal content, doxxing, revenge porn, deepfakes, crypto rug pulls, phishing links, bot farms, death threats, astroturfing, SEO parasites, child exploitation material, copyright-infringing reuploads, malware distribution, brigading, fake reviews, counterfeit goods listings, AI-generated LinkedIn hustle porn, self-harm promotion, election interference, sextortion, and "I'm not a bot" bots. Communities built moderation systems to stay alive. Moderation is not curation. Moderation asks what to remove. Curation is the harder question — what actually deserves your limited attention? The social media era became sophisticated at the first question and structurally weak at the second. This created an internet where enormous amounts of acceptable content compete simultaneously for attention. Much of it is simply forgettable — and forgettable content, at scale, is its own kind of noise. The problem became less about stopping bad content and more about identifying genuinely important content inside infinite amount of acceptable content. Andrew Brown called the internet "so big, so powerful and pointless that for some people it is a complete substitute for life." The internet is in a rush and platforms rarely opt to slow down and look elsewhere for immediately popular reactions. But they often struggle with less immediately visible forms of quality: nuanced expertise, minority viewpoints, emotionally difficult truths, sourced corrections, unpopular-but-accurate answers, and anything that starts with "well, actually it's more complicated than that." Subreddits often begin as highly specialized communities filled with practitioners, enthusiasts, cat people and/or other niche-content-aligned people. Over time, as they grow, many drift toward broader, easier to emotionally process, beginner-friendly content because broader participation naturally generates more voting activity. Reddit famously upvoted the [wrong suspects to the top during the Boston Marathon bombing](https://www.bbc.com/news/technology-22214511?ref=hackernoon.com). Due diligence takes time. **The system does not reward the best contribution. It rewards the contribution most understandable to the largest number of active participants.** ## Every Platform Thinks It's Everything Platforms struggle to answer a deceptively simple question: What exactly are we publishing? Is a platform a newspaper, an archive, a social network, a recommendation engine, a conversation layer, a reputation system, a live entertainment feed, a marketplace, a public square, a diary, a surveillance system, or infrastructure for social performance? Most internet platforms attempt to be many of these at the same time. LinkedIn began as a professional identity layer. Everyone's resume - now available on the internet. But over time, engagement incentives gradually pushed the platform toward emotional storytelling, motivational narratives, public vulnerability performances, and simplified business morality tales. The famous LinkedIn post format — short declarative lines, artificial suspense, a life lesson buried in a humble brag — emerged almost accidentally. No product manager wrote a spec for it (that I know of). The engagement system selected for it, rewarded it, and trained an entire generation of professionals to write this way without ever being explicitly told to. In 1962, W.H. Auden warned that writers confuse authenticity with originality. **LinkedIn inverted the problem — a platform full of people performing authenticity while writing identically.** Many digital media companies discovered that high-volume publishing outperformed slower editorial processes under ad-driven internet economics. The result was the industrialization of content production: A/B headline testing, rapid aggregation, trend chasing, endless listicles, reactive publishing, SEO parasites, rage bait, hot-take recycling, thumbnail A/B testing, "10 Things You Won't Believe" templates, keyword stuffing, and engagement farming. This was not because journalists or creators or even people as a whole stopped caring about quality. It was because the incentive structure stopped rewarding the things good journalism requires: patience, revision, and the willingness to report something true that no one wants to click. Writers became accountable to analytics dashboards. Newsrooms began thinking like growth teams. ## Database Structure Changed How Humans Structure Online Conversation Modern discourse is shaped, more than most people realize, by the data structures underneath. Platforms organize communication around sortable, rankable, relational units: posts, comments, reposts, reactions, engagement graphs, and every site also has that special sauce (HackerNoon's is that it's always noon somewhere). How we store data shapes the digital experiences we can build. Most platforms build for scale, not meaning. Human meaning resists this architecture. A conversation is not a collection of rankable rows. A developing idea is more than rows, columns, keys, queries, indexes, schemas, relations, joins, and sort order. The old librarian's rule applies: how you shelve a book determines who finds it. A database stores a post as a row with a timestamp, an author ID, a reaction count, and a content blob. That's what the system can retrieve, rank, and surface. It cannot retrieve "changed my thinking" or "worth rereading in five years" because no column holds those values. HackerNoon tracks reads. A read means someone showed up, stayed, and scrolled. It's an imperfect metric, but it measures attention rather than applause, and that difference compounds across thousands of blogs. Most architectures can only ask "what's popular right now?" A reads-based system at least asks "what held someone's attention?” Yet platforms treat discourse as a database optimization problem — and that design choice carries real consequences. When everything must become a rankable object, context collapses. A meticulously argued essay sits directly adjacent to a meme, a conspiracy theory, and a celebrity photo, because that’s the internet. The structure cannot distinguish between them. It can only measure which generated more interaction. [University of Illinois researchers found](https://arxiv.org/pdf/1910.00757?ref=hackernoon.com) that voters on content platforms are systematically susceptible to reputation bias, social influence bias, and position bias — meaning a piece of content's score frequently reflects who posted it and what was already upvoted, not the quality of the argument itself. The internet flattened contextual hierarchy. Everything became feed material. And once everything is feed material, the architecture sculpts what gets said. ## Quantity Has Better Metrics Than Quality You don't always know quality when you see it. Quality is expensive to evaluate. Quantity is cheap to count. A database can measure clicks, shares, watch time, retention, and session length with perfect precision. These numbers are real, trackable, and reportable. They inhabit dashboards, get optimized against, and get presented, published, paraded, and proselytized. They feel like truth because they are exact. The qualities that actually distinguish enduring publishing are far harder to capture: originality, intellectual honesty, emotional resonance, clarity of thought, genuine expertise, the courage to be boring when the subject demands it, and whether something remains worth reading years after publication. None of those compress cleanly into a column. Quality publishing is expensive because access is expensive and the people capable of doing it well have limited time. The best reporting often requires months of relationship building, proprietary data access, technical expertise, travel, legal risk, and/or simply enough time to understand something difficult before publishing it. A carefully reported essay may take a year to produce and reach only a few thousand readers, while a reactive post written in twenty minutes can generate exponentially more engagement. Internet metrics reward output far more reliably than they reward depth. At HackerNoon, we raise the floor. Eliminating the worst 90% of what comes through our open blog posts submissions queue raises the quality of everything that makes it to the reader. No dashboard can show you the impact of what you chose not to publish. The reader never encounters what got rejected, and that's the impact they feel. [Research published in Nature Human Behaviour](https://pmc.ncbi.nlm.nih.gov/articles/PMC6206065/?ref=hackernoon.com) found that popularity-based metrics can bias future success in ways that do not reflect quality — and in some cases actively suppress it. Early engagement drives visibility, visibility drives more engagement, and the loop compounds whether the content deserved it or not. [The average blog post takes four hours to write. Readers spend 52 seconds on it](https://optinmonster.com/blogging-statistics/?ref=hackernoon.com) (and thank you, reader, for making it beyond the 52 second point today). [A Microsoft Research and UC Berkeley study](https://arxiv.org/pdf/2401.09804?ref=hackernoon.com) found that engagement-based optimization can actually perform worse for user utility than random recommendations — because it rewards gaming and quality simultaneously, making them nearly indistinguishable at the surface level. Every internet community inherits this problem. Once you're dependent on growth metrics, you build toward what's measurable. More likes. More comments. More uploads. More eyeballs. Quantity scales operationally, as in, it can get automated so it does get automated. Quality scales editorially, as in, you can't automate it without losing the thing that made it quality in the first place — a person who read the work and decided it was worth someone else's time. ## AI Intensifies the Existing Problem, DUH AI did not create the quantity problem. It industrialized it. For years, internet communities operated under a tacit assumption: even mediocre content still demanded human effort to produce. That friction functioned as a hidden quality filter. Writing something — even something bad — still consumed time, attention, and some minimum threshold of intention. What AI industrializes isn't bad content — it's **plausible mediocrity**: grammatically correct, structurally coherent, superficially persuasive, competently formatted, and almost indistinguishable from average. Enormous volumes now generate in seconds. Before ChatGPT launched in late 2022, roughly [5% of new web articles were primarily AI-generated. By November 2024, that figure had crossed 50%](https://graphite.io/five-percent/more-articles-are-now-created-by-ai-than-humans?ref=hackernoon.com). An [Ahrefs study of 900,000 newly created web pages in April 2025 found that 74% contained AI-generated content](https://ahrefs.com/blog/what-percentage-of-new-content-is-ai-generated/?ref=hackernoon.com). The content flood that took the blogosphere thirty years to build, AI matched in one. The result isn’t just more bad content. It is more plausible content. Spam used to announce itself. You knew it when you saw it. [Recent research on the decline of online knowledge communities](https://arxiv.org/abs/2603.27399?ref=hackernoon.com) found that while users reach for AI out of convenience, they still return to human communities for complex, ambiguous, or trust-sensitive questions — which suggests readers can sense the difference even when they cannot articulate it. The information exists. The human judgment feels absent. The scientific publishing world is already living a preview of this problem. [Researchers have warned](https://arxiv.org/html/2510.09686v1?ref=hackernoon.com) that AI-generated survey papers are flooding preprint servers like arXiv — what was once a labor-intensive exercise in critical synthesis has become a low-barrier, high-volume output, burying original work beneath a rising tide of automated summarization. What is happening to academic publishing is already happening to every content community that does not actively resist it. [Daniel Sarewitz warned in Nature in 2016](https://www.nature.com/articles/533147a?ref=hackernoon.com) that scientists must publish less or good research would be swamped by poor work. He was writing before generative AI existed. The flood he feared arrived ahead of schedule and then industrialized itself. This creates a strange future. Internet communities may slowly rediscover the value of the things the early web tried to route around: reputation, editorial trust, and a voice you'd actually recognize if you read it twice. Not because AI cannot write. The information exists but the human judgment is missing. ## Small Editorial Decisions Matter More Than Grand Philosophies One of the easiest mistakes platforms make is assuming quality emerges from mission statements. In practice, it emerges from small operational decisions repeated consistently. A community's standards are encoded in seemingly minor product choices: whether logged-out users can view the content, whether old posts remain discoverable or get buried by recency, whether every story gets human review before going live, whether anyone can comment or only verified users, and whether the default sort is chronological, "best," or whatever the algorithm decides. We document our [Editing Protocol](http://editingprotocol.com/?ref=hackernoon.com). La Fontaine wrote in 1668 that "by the work one knows the workmen." On the internet, by the product decisions, one knows the platform. What gets rewarded, what gets edited, what gets rejected — these are the real editorial values, not the ones on the about page. [Wikipedia](https://hackernoon.com/wikipedia-rules-everything-around-me?ref=hackernoon.com) is the strongest proof of this. Its quality did not emerge because humanity collectively became more rational. It emerged because Wikipedia built systems that rewarded citations, revision history, moderation transparency, consensus mechanisms, persistent authorship records, and visible editorial correction. The platform embedded epistemology into product design. Other platforms proved better design produces better outcomes — though none escaped the tradeoffs entirely. Stack Overflow’s voting system was not a pure popularity ranking — it heavily rewarded technically correct, concise, and reusable answers. Over time this built a culture where precision mattered, duplication was discouraged, and expertise accumulated reputational weight. The system was imperfect — fastest answers accumulated votes whether or not they aged well, and the format it rewarded turned out to be exactly what AI could replicate overnight. Nevertheless it demonstrated that interface design can shape knowledge quality. Product Hunt started as a newsletter with one person choosing what to feature before growing into a platform where community upvotes matter but staff curation was also required to get on the homepage. The combination of human editorial judgment and community signal is inspirational. Most platforms pick one and let the other atrophy. At HackerNoon, many meaningful editorial decisions were similarly small and operational. One recurring internal tension was whether to optimize aggressively for publication throughput or preserve editorial friction. **HackerNoon repeatedly chose friction in places where many internet platforms removed it.** Quality blog posts are not dumped into an instant-publish firehose; instead, human editorial review remained embedded in the publishing pipeline. That decision slowed growth metrics in obvious ways. It also preserved a stronger quality floor, helping us avoid [the AI slop portions of the internet](https://www.wired.com/story/ai-generated-medium-posts-content-moderation/?ref=hackernoon.com). Permanence goes further than the feed. By prioritizing durable story URLs, author attribution, and evergreen discoverability, writing is treated as an enduring document rather than disposable inventory. To ensure these records survive even if a centralized platform does not, HackerNoon uses decentralized storage through [Arweave](https://arweave.hackernoon.com/?ref=hackernoon.com). So the content survives even if the platform doesn't. There were similarly obsessive internal conversations around headlines. We [prioritize specificity above all else](https://help.hackernoon.com/the-art-of-headline-writing?ref=hackernoon.com) —the goal is a title that tells the truth about what's inside. We didn't just ask if a title would generate clicks; we asked if it overpromised, if it collapsed nuance into outrage bait, or if it sounded machine-generated. **We ran the numbers internally: when professional editors retitled blog submission headlines in lieu of blogger-submitted headlines, they were 4x more likely to surpass 1,000 reads.** Those are tiny editorial decisions, but they accumulate into a distinct editorial culture. Users observe what gets amplified, writers observe what gets edited, and readers observe what survives. Culture isn't declared. It's compiled from thousands of small decisions ## Can Quality Usurp Quantity? Why Editorial Friction Matters Internet communities struggle to publish quality over quantity because the internet changed the economics of publication without evolving the economics of attention. Publishing got cheap. Attention didn't. This imbalance drives the core tensions of modern platforms—from algorithmic feeds and engagement optimization to the collapse of trust in institutional curation. This brings us to the core of why we do this. At some level, we recognize that artists and writers, just like the rest of us, need to eat, provide for their families, and have a reliable roof over their very own place to shit. But if we allow the "live ranking system" to be the sole arbiter of value, we abandon the idea of the durable document. This argument extends off screens. A neighborhood BBQ, a team practice, a family dinner — people face the same question their feed pretends to answer: what gets your attention? Offline, you can only be in one room at a time. The internet removed that constraint and discovered that removing it doesn't free attention — it fragments it. Running an editorial queue long enough teaches you that most things competing for your time aren't worth it — and that lesson follows you home. **What AI actually makes scarce is not content, or even good content, but the willingness to say no to plausible mediocrity.** The internet solved the problem of who gets to publish and scaled the problem of deciding what is worth reading. The internet communities that survive long-term will not be the ones that publish the most content. They will be the ones that develop trustworthy systems for helping people decide what is worth reading at all. *Featured image credit, Demo Day Slide 4 from* [*Cloudflare AI Workers Cohort*](https://hackernoon.com/hackernoon-joins-cloudflares-workers-launchpad-to-scale-its-tech-blogging-network?ref=hackernoon.com)*. Do your slides tell a bigger story?* [*Grab your most aspirational one, upload as the featured image for your next HackerNoon blog post*](http://hackernoon.com/new?ref=hackernoon.com)*.* --- ## Proof of Usefulness [Proof of Usefulness Hackathon](https://proofofusefulness.com/?ref=hackernoon.com) is a global 6-month developer challenge designed to reward real-world utility projects and initiatives. With 150,000+ in [cash prizes](https://proofofusefulness.com/cash-prizes?ref=hackernoon.com) and [software credits](https://proofofusefulness.com/software-prizes?ref=hackernoon.com) for winners and $1500+ worth of software and inventory for participants, this is undisputedly the biggest contest of the year. Learn more [here](https://proofofusefulness.com/?ref=hackernoon.com). %% ## Colophon title:: Why Internet Communities Struggle to Publish Quality Over Quantity type:: [[full-text]] url:: https://hackernoon.com/why-internet-communities-struggle-to-publish-quality-over-quantity?twclid=2f3pk87sms2wsvmp8d6vcwco87 date:: [[2026-05-17]] published:: 2026-05-14