The Content Flood: Why More Isn't Better
Every day, the internet gets bigger. WordPress sites alone publish roughly 7.5 million blog posts per day. YouTube receives over 500 hours of new video every minute. That was the baseline before generative AI.
Since ChatGPT's launch in late 2022, AI-generated content has surged. NewsGuard identified over 1,000 AI-generated news sites operating by mid-2024. Originality.ai estimated that 14% of all web content published in 2025 was fully AI-generated, up from under 2% in 2022. Amazon's Kindle store saw a 400% increase in self-published titles between 2022 and 2025, with a significant portion produced using LLMs.
This isn't a quality problem alone. It's a signal-to-noise problem. When content creation costs approach zero, the supply of mediocre content becomes effectively infinite. The scarce resource is no longer the ability to produce; it's the ability to filter.
Herbert Simon predicted this in 1971: "A wealth of information creates a poverty of attention." He was right fifty years early. The difference now is that the wealth of information has gone from "a lot" to "incomprehensible." Attention hasn't scaled with it.
For anyone who cares about learning, this shift changes the game entirely. The question isn't "can I find information about X?" That's trivially easy. The question is "out of 10,000 pieces on X, which 3 are worth my time?" That's the curation question. And right now, most people are outsourcing it to algorithms that don't have their interests in mind.
If you're thinking about how to manage this flood deliberately, our piece on building a healthy information diet covers the tactical side.
What Taste Actually Means (and Why It Matters Professionally)
The word "taste" sounds subjective, maybe even elitist. But in practice, taste is a specific cognitive skill: the ability to evaluate quality quickly and accurately based on deep pattern recognition.
Steve Jobs described it simply in a 1996 interview with Wired: "The only problem with Microsoft is they just have no taste. They have absolutely no taste. And I don't mean that in a small way, I mean that in a big way, in the sense that they don't think of original ideas, and they don't bring much culture into their products."
Jobs wasn't being abstract. He was describing a concrete competitive advantage. Apple's success came from knowing what to leave out as much as what to include. Every product decision was a curation decision.
Patrick Collison, CEO of Stripe, is known for sharing reading lists and intellectual influences publicly. His personal website has curated book recommendations organized by theme, with annotations explaining why each one matters. Collison doesn't just read widely. He selects carefully and contextualizes what he selects. That's taste in action.
Maria Popova built Brain Pickings (now The Marginalian) into one of the most respected intellectual publications on the internet. She reads roughly 12 books a week, but publishes about 3 essays. Her value isn't volume. It's the 75% she discards and the connections she draws between the 25% she keeps.
Taste, understood this way, has three components:
- Breadth of exposure: You can't judge quality if you haven't seen enough examples. Popova reads 12 books to write about 3. Jobs studied calligraphy, architecture, Zen Buddhism.
- Evaluative frameworks: Experienced curators develop internal models for what "good" looks like. These frameworks are often tacit, built through repetition rather than explicit rules.
- The courage to exclude: Curation requires saying no. Algorithms can't do this because they optimize for engagement, not value. Humans can.
This is learnable. Researchers at the University of Chicago found in 2023 that aesthetic judgment in domains like visual art, music, and writing improves measurably with deliberate practice and feedback. Taste isn't innate; it's trained through sustained, reflective engagement with material.
Algorithmic Recommendation vs. Human Curation
These two approaches look similar on the surface. Both present you with content. But the underlying logic is fundamentally different.
| Dimension | Algorithmic Recommendation | Human Curation |
|---|---|---|
| Optimizes for | Engagement (clicks, time on site, shares) | Meaning (insight, usefulness, quality) |
| Selection method | Statistical patterns in behavioral data | Judgment based on expertise and context |
| Accountability | Opaque; no one "chose" what you see | Transparent; a specific person made the selection |
| Serendipity | Low; reinforces past behavior | High; curators surface unexpected connections |
| Context | Absent; items are ranked by relevance scores | Present; curators explain why something matters |
| Bias type | Popularity bias, recency bias, engagement bias | Personal bias (but identifiable and correctable) |
| Scales how | Infinitely, at near-zero marginal cost | Slowly, limited by human attention |
| Trust model | "The system knows you" | "I trust this person's judgment" |
Netflix's recommendation engine drives 80% of viewer choices (Netflix Tech Blog, 2023). YouTube's algorithm determines 70% of all watch time. Spotify's Discover Weekly playlist, built on collaborative filtering, accounts for a significant share of music discovery. These systems work. But "works" means "keeps users engaged," not "makes users smarter or better informed."
The distinction matters because engagement and value diverge regularly. YouTube's algorithm will happily send you down a 4-hour rabbit hole of increasingly extreme content. A good human curator would hand you one 20-minute video and say, "This is the best explanation I've found. Watch this and you're done."
Human curation introduces something algorithms structurally can't: editorial judgment. When a curator says "read this, skip that," they're importing values, expertise, and context that no engagement metric captures. That's why newsletters from individuals (Ben Thompson's Stratechery, Matt Levine's Money Stuff, Emily Oster's ParentData) often outperform algorithmic feeds for depth and satisfaction. Someone chose what to include.
Filter Bubbles, Echo Chambers, and the Cost of Automation
Eli Pariser's 2011 book The Filter Bubble warned that algorithmic personalization would trap users in information cocoons. At the time, many researchers pushed back, arguing the effects were small.
Fourteen years later, the evidence is stronger. A 2025 study published in Nature Human Behaviour by Bhadani, Roth, and colleagues examined how LLM-powered content recommendations affected information diversity. Over a six-month period, users exposed to AI-curated feeds showed a 34% reduction in the diversity of sources they consumed, compared to a control group using chronological feeds. The effect was strongest among users who interacted most with the system, a feedback loop where engagement deepened narrowing.
This matters beyond politics. In professional contexts, filter bubbles mean knowledge workers see only ideas that match their existing frameworks. An engineer who only reads content the algorithm serves will develop increasingly narrow expertise. A founder who relies on algorithmic news feeds will miss signals outside their industry.
The mechanism is straightforward. Recommendation algorithms learn from behavior. If you click on articles about React, the algorithm serves more React articles. Over time, you stop seeing articles about Rust, Elixir, or system design. Your feed becomes a mirror of your past interests, not a window into what you might need to learn next.
A 2024 working paper from the MIT Sloan School found that 62% of knowledge workers described their primary information feeds as "repetitive," and 41% said they felt "stuck in a content loop." Only 18% reported regularly encountering perspectives that challenged their assumptions through algorithmic feeds.
The cost is real. Diverse information intake correlates with better decision-making. A 2023 study in Management Science by Aggarwal and Woolley found that teams whose members consumed information from diverse sources outperformed homogeneous-information teams by 29% on complex problem-solving tasks. Algorithms that narrow your inputs are quietly degrading your outputs.
The Anti-Algorithm Movement
A growing number of readers, creators, and knowledge workers are actively opting out of algorithmic feeds. This isn't a fringe trend. It's measurable.
Substack reported that paid newsletter subscriptions grew 50% year-over-year in 2025, reaching over 35 million active paid subscriptions. These readers are paying real money for one thing algorithms can't provide: a specific human's judgment about what matters.
RSS reader usage, long declared dead, has been climbing steadily since 2023. Feedly reported a 38% increase in active users between 2023 and 2025. Readwise Reader, Omnivore (before its acquisition), and Feedbin all reported growth. People are rebuilding their own information pipelines.
Curated community platforms are gaining traction too. Glasp's community feed shows what real people are highlighting and reading across the web, a social signal that's far more useful than what an algorithm thinks you'll click. When you can see what thoughtful readers are paying attention to, you get serendipity without manipulation.
| Approach | Example | What You Get | What You Give Up |
|---|---|---|---|
| Algorithmic feed | Twitter/X, TikTok, YouTube Home | Infinite content, zero effort | Control, diversity, depth |
| Paid newsletters | Stratechery, Money Stuff, The Diff | Expert judgment, consistent quality | Breadth, free access |
| RSS / reader apps | Feedly, Readwise Reader | Full control, no algorithm | Discovery (you choose sources manually) |
| Community curation | Glasp, Hacker News, Are.na | Diverse perspectives, social proof | Scale (smaller communities) |
| Personal curation | Notion databases, Obsidian vaults | Perfect fit, deep context | Time-intensive to maintain |
The common thread is intentionality. Each of these approaches puts the human back in the loop. You're not passively receiving whatever the algorithm serves. You're actively choosing sources, evaluating quality, and building your own information architecture.
This connects directly to the idea of learning in public. When you curate openly, sharing what you read and why, you create value for others while strengthening your own judgment.
Curation as a Professional Skill
Curation isn't just a personal practice. It's becoming a core professional competency across industries.
In venture capital, the ability to curate deal flow separates top-performing firms from average ones. Marc Andreessen has said that the best investors are essentially "curators of people and ideas." Y Combinator's success comes not from capital (which is abundant) but from selecting the right 1-2% of applicants.
In product management, the job is fundamentally curatorial. A PM doesn't build features. They decide which features to build and which to kill. Good PMs maintain a deep, curated understanding of user needs, competitive dynamics, and technical constraints. They select the right problems to solve.
In education, the shift from "sage on the stage" to "guide on the side" is really a shift toward curation. The best teachers don't just present information. They select the right readings, sequence them effectively, and contextualize each piece within a broader framework.
Knowledge workers broadly are finding that curation skill correlates with career advancement. A 2024 McKinsey Global Survey on workforce productivity found that employees who maintain structured personal knowledge bases (curated notes, annotated bookmarks, organized highlights) are 23% more productive than those who rely on search-and-retrieve methods alone.
This makes sense when you think about it. The average knowledge worker spends 9.3 hours per week searching for information, according to a 2023 IDC study. If you've already curated the best resources on your core topics, that search time drops dramatically. Curation is an investment in future efficiency.
For researchers, curation is especially powerful. As we covered in our piece on collective intelligence, when individuals curate and share their findings publicly, they create a compounding knowledge resource. One person's curated highlight becomes another person's starting point.
Social Annotation: The Curation Layer for the Internet
Here's a thought experiment. Imagine if every time someone read an article and found a key insight, that insight was visible to the next reader. Not as a comment at the bottom. Not as a like or share. As a highlighted passage, right in the text, with the reader's note explaining why it mattered.
This is what social annotation does. It creates a curation layer on top of the existing internet.
The concept goes back to Vannevar Bush's 1945 essay "As We May Think," where he described the "memex," a device that would let users create trails of linked annotations across documents. Bush imagined that these associative trails, created by thinkers with good judgment, would become more valuable than the documents themselves.
Eighty years later, the technology exists. Glasp's web highlighter lets you highlight passages on any webpage and share those highlights publicly. Your highlights show up in a community feed. Other users can see what resonated with you. Over time, this builds a rich, human-curated map of what's most valuable across millions of web pages.
This is fundamentally different from algorithmic recommendation. An algorithm says "people who clicked this also clicked that." A highlighted passage says "a real person read this carefully and found this specific part valuable." The informational content is richer, more trustworthy, and more useful.
Social annotation also solves a problem that affects how AI is reshaping learning: the passivity trap. When you highlight text and write a note about why it matters, you're engaging in active processing. You're not just consuming. You're evaluating, selecting, and contextualizing. That's curation in its purest form.
And when those annotations are shared, they create what network theorist Clay Shirky called "cognitive surplus turned productive." Every annotation is a tiny act of curation that benefits the entire community. Import your Kindle highlights into Glasp, and your years of reading become a public resource. Use YouTube Summary to capture key points from video content, and you've curated the essential 10% from an hour-long talk.
The Economics of Curation in the AI Era
There's a simple economic argument for why curation becomes more valuable as AI improves: when the supply of content goes to infinity, the supply of good judgment doesn't.
This is a classic case of complementary goods. AI makes content production nearly free. But the value of any individual piece of content drops toward zero as supply increases. What remains scarce, and therefore valuable, is the ability to distinguish the 1% that's excellent from the 99% that's noise.
Li Jin, co-founder of Variant Fund, has argued that the creator economy is shifting toward a "curator economy." Her analysis shows that in saturated content markets, audiences increasingly pay for filtration rather than creation. A newsletter that writes original analysis is valuable. A newsletter that finds and contextualizes the best existing content is often more valuable, because the reader's problem isn't lack of content. It's lack of filtering.
The wage data supports this. Glassdoor data from 2025 shows that content curator roles at media companies command salaries 15-30% higher than content creator roles at equivalent levels. Research librarians, who are professional curators, have seen a 22% increase in median compensation since 2020 as organizations struggle with information overload.
At the organizational level, the return on curation investment is even clearer:
- Consulting firms charge clients primarily for curated knowledge. McKinsey's value isn't generating original data. It's selecting the right frameworks and evidence for a specific client's situation.
- Venture capital returns are driven by selection quality. The best VCs don't see more deals than average VCs. They pick better.
- Academic publishing is a curation business. Peer review, journal selection, and editorial judgment are all forms of curation that determine which research gets attention.
As AI drives content creation costs toward zero, every business will increasingly be a curation business. The organizations and individuals who develop strong curation capabilities now will have a structural advantage.
Building a Curation Practice
Curation is a skill, which means it improves with deliberate practice. Here's a practical framework for building a daily curation habit.
Step 1: Diversify your inputs. Subscribe to 5-10 sources outside your primary field. If you're an engineer, add a philosophy blog, an economics newsletter, and a design publication. The goal is breadth of exposure. Use RSS readers or tools like Glasp's community feed to discover what smart people in other fields are reading.
Step 2: Highlight actively. Don't just read. Mark what resonates. Use Glasp's web highlighter to capture passages as you encounter them. The physical act of selecting text forces you to make judgment calls: is this worth saving? This trains your evaluative instinct.
Step 3: Annotate with context. A highlight without a note is half a curation. Add a sentence explaining why you saved it. "This contradicts the standard view on X." "Best explanation of Y I've found." "Connects to Z idea." These notes are where taste becomes explicit and shareable.
Step 4: Review and connect. Once a week, review your recent highlights. Look for patterns, contradictions, and connections across sources. Use Glasp's AI chat to ask questions about your collected highlights and surface relationships you might have missed. This synthesis step is where curation becomes knowledge creation.
Step 5: Share selectively. Publish your best finds. Write a weekly roundup. Share a reading list. The act of curating for an audience raises your standards. You don't share everything you find; you share only what clears a quality bar. This is taste under pressure.
Step 6: Iterate on your sources. Every month, audit your information inputs. Drop sources that have declined in quality. Add new ones that fill gaps. Your curation practice should evolve as your knowledge and interests develop. A static reading list is a sign of stagnation.
This is very close to what we described in our piece on building a healthy information diet, but with an important addition: the sharing layer. Curation practiced in isolation helps you. Curation practiced in public helps everyone.
Frequently Asked Questions
Isn't AI curation getting good enough to replace human curators?
AI is excellent at finding content that matches your past preferences. It's poor at the things that make curation valuable: surfacing unexpected connections, applying values-based judgment, and providing context that requires lived experience. A 2025 survey by the Reuters Institute found that 67% of respondents trusted content recommended by a specific person they follow more than content recommended by an algorithm. AI can assist human curators (by surfacing candidates, summarizing content, finding related pieces), but the editorial judgment layer remains distinctly human.
How do I develop taste if I'm just starting out?
Volume first, then selectivity. Read widely for 3-6 months without worrying about curation quality. Highlight generously. Over time, you'll notice that your early highlights were less discriminating than your recent ones. That gap is taste developing. The University of Chicago research on aesthetic judgment (2023) found that improvement requires roughly 200 hours of deliberate, reflective engagement with material in a given domain. There's no shortcut, but the trajectory is reliable.
Can curation be a career, not just a side practice?
It already is, for many people. Newsletter writers like Ben Thompson (Stratechery, estimated $3M+ annual revenue) and Matt Levine (Bloomberg's Money Stuff) are professional curators. Research librarians, museum curators, talent scouts, venture capitalists, and academic journal editors are all curators by profession. As content production costs drop, demand for skilled curators will grow across industries. Li Jin's "curator economy" thesis predicts this will become a larger employment category over the next decade.
How is Glasp different from bookmarking or saving articles?
Bookmarking saves a URL. Curation saves the specific insight and the reason it matters. When you highlight a passage in Glasp, you're isolating the signal from the noise within a piece. Your highlights become searchable, shareable, and connected to your other highlights. Over time, this builds a personal knowledge graph that reflects your intellectual development. The social layer means other users benefit from your curation, and you benefit from theirs, creating a collective intelligence effect that no private bookmarking tool can match.
Won't AI-generated content make curation even harder?
Yes, in the short term. The volume of content will increase faster than the quality of filtering tools. But this is precisely why human curation becomes more valuable, not less. In a world of infinite AI-generated articles, the person who can say "these three are worth reading and here's why" provides enormous value. The flood makes the filter more important, not less.
Conclusion: Curate or Be Curated
There are really only two options now. You can let algorithms decide what you see, read, and think about. Or you can develop the skill to choose for yourself and help others choose better.
The evidence is clear: algorithmic curation optimizes for engagement, narrows your worldview, and degrades the diversity of your information intake. Human curation, practiced with intention and shared openly, does the opposite. It surfaces unexpected ideas, builds genuine understanding, and creates compounding value for everyone in the network.
Taste isn't a luxury. It's a survival skill for the information age. And like any skill, it develops through practice: reading widely, highlighting deliberately, annotating with context, and sharing with others.
Start building your curation practice today. Install Glasp's web highlighter, start highlighting what matters, and share your finds with the community. Your taste, developed through thousands of small curation decisions, will become one of your most valuable professional assets.
The age of AI is also the age of the human curator. The question is whether you'll be one, or whether you'll let someone else's algorithm curate your reality for you.