Personal Context Management: Why Your Highlights Are the New Second Brain

The End of PKM as We Knew It

Personal knowledge management had a good run. For over a decade, the PKM movement taught millions of people to capture, organize, and retrieve information using tools like Notion, Obsidian, Roam Research, and Evernote. Tiago Forte's "Building a Second Brain" methodology became the gold standard, with over 25,000 learners adopting the CODE framework (Capture, Organize, Distill, Express).

It worked. Sort of.

The dirty secret of PKM is that most people never got past the capture phase. A 2023 survey of Notion users found that 68% of their databases hadn't been opened in over 30 days. Readwise reported that the median user reviews saved highlights fewer than twice per month. People built elaborate second brains, then rarely consulted them.

The reason is simple: human retrieval is slow and effortful. You need to remember that you saved something, recall roughly where it is, navigate to it, read through it, and synthesize it with whatever you're working on. That process takes minutes, sometimes longer. In practice, most people just Google it again or ask a colleague.

Then large language models arrived. And the entire calculus changed.

AI doesn't need a perfectly organized folder structure. It doesn't need tags, color codes, or carefully nested hierarchies. What it needs is context: raw material about who you are, what you've read, what you found important, and how you think. Feed it that, and it can do in seconds what PKM promised but never quite delivered.

This is the shift that Tiago Forte himself acknowledged in 2025 when he introduced the concept of "Personal Context Management." The bottleneck is no longer organizing knowledge for your own retrieval. It's curating context for AI to use on your behalf.

If you've been building a second brain, you're not starting over. You're upgrading.

---

What Is Personal Context Management?

Personal Context Management (PCM) is the practice of deliberately curating, structuring, and maintaining a personal context layer that AI systems can draw on to produce relevant, personalized output.

Here's the distinction. PKM asks: "How do I organize what I know so I can find it later?" PCM asks: "How do I give AI enough about me so its output is actually useful?"

The difference matters because AI has inverted the retrieval problem. Finding information is now trivially easy. Any LLM can retrieve facts, summarize research, or explain concepts. What AI can't do, without your help, is know which facts matter to your specific situation. It can't know that you're writing a strategy memo for a fintech startup, that you've been researching behavioral economics for six months, or that you disagree with the prevailing view on subscription pricing.

That's context. And it's personal.

PCM has three layers:

1. Identity context: Your role, expertise, interests, goals, and preferences. This is relatively stable and changes slowly.
2. Knowledge context: The specific ideas, sources, and insights you've accumulated over time. This is your highlight library, your annotations, your reading history.
3. Task context: What you're working on right now. The project, the deadline, the audience, the constraints.

Traditional PKM handled layer 2 (poorly, for most people) and ignored layers 1 and 3. PCM treats all three as inputs to a system where AI does the heavy lifting of retrieval, synthesis, and even first-draft creation.

---

Why Context Beats Knowledge in the AI Era

Ethan Mollick's "Co-Intelligence" (2024) introduced a useful framework: AI is a "co-intelligence" that amplifies human capability, but only when given sufficient direction. In controlled experiments at Wharton, consultants who provided detailed context to AI tools produced work that was 40% higher quality than those who used generic prompts.

The variable wasn't intelligence. It was context.

Consider two prompts:

Generic: "Write a blog post about customer retention."

Context-rich: "I run a B2B SaaS company selling to mid-market HR teams. Our churn rate is 8% monthly, mostly from companies with under 50 employees. I've highlighted research from Bain showing that a 5% retention increase yields 25-95% profit gains. Write a post for our customer success team about reducing churn in the SMB segment."

The second prompt produces dramatically better output. Not because the AI is smarter, but because you gave it your context.

Now scale that up. Instead of manually writing context into every prompt, imagine an AI that already has access to:

• Every article you've highlighted in the past year
• Your annotations and notes on those highlights
• The topics you return to most frequently
• The sources you trust (and those you've ignored)
• Your writing style, vocabulary preferences, and typical arguments

That's what a PCM system provides. Your highlights become a persistent context layer that makes every AI interaction more useful.

Research supports this. Sparrow et al. (2011) demonstrated the "Google effect," showing that people's memory for information drops significantly when they know it's stored externally. But the follow-up finding was more interesting: people become excellent at remembering where information is stored. We're natural indexers, not natural hard drives. PCM leans into this strength rather than fighting it.

---

Highlights: The Atomic Unit of Personal Context

Not all captured information makes good context. Bookmarks are too coarse. Full articles are too long. Tags are too abstract. But highlights hit a sweet spot.

A highlight is a passage you chose to save because it mattered to you. That act of selection is itself information. It tells AI: "This person found this specific passage worth preserving, out of everything they read that day."

Highlights have several properties that make them ideal for PCM:

• Pre-filtered: You've already done the editorial work. Every highlight represents a judgment call about relevance.
• Source-attributed: Unlike raw notes, highlights carry their original source URL, author, and publication date. This gives AI provenance and credibility signals.
• Compact: A typical highlight is 1-3 sentences. You can feed hundreds of highlights into an AI context window without hitting token limits.
• Time-stamped: The sequence of your highlights tells a story about how your thinking evolved. AI can weight recent highlights more heavily.
• Annotatable: When you add a note to a highlight, you're adding your interpretation. That's the richest form of personal context.

Roediger and Karpicke's research on retrieval practice (2006) showed that the act of actively engaging with material, selecting key passages, annotating them, reviewing them later, strengthens memory formation by 50-80% compared to passive re-reading. Highlighting isn't just good for AI. It's good for your own learning.

The key is that highlighting must be frictionless. If it takes more than a few seconds, you won't do it consistently. This is where Glasp's web highlighter changes the equation: highlight any passage on any webpage with a single click, and it's saved, organized, and available as context for AI.

---

From Capture to Context: How the Workflow Changes

The traditional PKM workflow looked like this:

1. Read something interesting
2. Capture it (copy-paste, bookmark, screenshot)
3. Organize it (file into folders, add tags)
4. Review it periodically
5. Use it when the moment arises

Most people completed steps 1-2 and abandoned the rest. The organizing step was too labor-intensive, and the review step required discipline that few maintained.

The PCM workflow is different:

1. Read something interesting
2. Highlight the specific passages that resonate (this is both capture and curation in one step)
3. Annotate when you have a reaction worth preserving
4. Let AI organize by topic, theme, and relevance
5. Chat with your highlights when you need to synthesize, write, or think through a problem

Notice what disappeared: manual organization and periodic review. AI handles both. You focus on the parts humans do best, reading critically and making judgment calls about what matters, and AI handles the parts humans do worst, organizing, retrieving, and synthesizing across large volumes of saved material.

This isn't theoretical. Glasp's AI chat already enables this workflow. You can ask questions across your entire highlight library: "What have I saved about pricing psychology?" or "Summarize the arguments for and against remote work from my highlights." The AI draws on your curated context, not the entire internet, to produce answers grounded in sources you've already vetted.

---

Video as a Context Source

Text isn't the only medium worth mining for context. YouTube alone hosts over 800 million videos, and video content now accounts for 82% of all internet traffic (Cisco, 2023). For many people, video lectures, conference talks, and podcast interviews are their primary learning medium.

The problem is that video is terrible for traditional PKM. You can't highlight a video. You can't skim it. You can't search within it efficiently. A 45-minute conference talk might contain three genuinely valuable insights buried in 42 minutes of setup, examples, and Q&A.

This is where AI-powered video summarization changes the game. YouTube Summary generates structured summaries of any YouTube video, extracting the key points, arguments, and timestamps. You can then highlight the parts of the summary that matter to you, effectively converting a 45-minute video into 3-5 targeted highlights that feed into your context layer.

The efficiency gain is substantial. Reading a summary takes 2-3 minutes versus watching a full video. And the highlights you save from that summary carry the same contextual value as highlights from written articles: they're pre-filtered, attributed, and ready for AI to use.

For researchers, students, and knowledge workers who consume significant amounts of video content, this closes what was previously an enormous gap in PKM systems. Your learning from video no longer disappears the moment the browser tab closes.

---

The Social Layer: Learning from Others' Context

Here's where PCM diverges most sharply from traditional note-taking. Your highlights exist not just for you and your AI, but as a signal to others about what you found valuable.

When you read an article on Glasp, you can see what other readers highlighted on the same page. This is a form of distributed cognition. Research on collective intelligence (Woolley et al., 2010) shows that groups consistently outperform individuals at identifying important information, not because any single person is smarter, but because diverse perspectives catch what any one reader would miss.

Consider reading a dense research paper on climate economics. You might highlight the key finding and the methodology section. But another reader, an economist, highlights the assumption buried in footnote 14 that undermines the paper's central claim. A third reader, a policy analyst, highlights the implementation challenges discussed in the conclusion. Each person's highlights reflect their expertise and priorities.

In a PCM system, this social layer becomes part of your extended context. You're not just drawing on your own highlights. You're drawing on the collective reading intelligence of a community of thoughtful readers. This is what separates learning in public from learning in isolation.

The social dimension also creates accountability. When your highlights are visible to others, you're more likely to read carefully and highlight thoughtfully. It's the same principle behind "working with the garage door up," as Andy Matuschak puts it. Public learning is more rigorous learning.

---

Building Your PCM System

A practical PCM system doesn't require a complex setup. It requires consistency in three habits and clarity about one principle.

The principle: Save context, not content. You don't need to capture entire articles. You need to capture the specific passages that intersected with your thinking. Quality over quantity, every time.

Habit 1: Highlight as you read. Install a web highlighter and use it on every article, blog post, and research paper you read. Don't overthink color coding or tagging at this stage. Just highlight what resonates. Aim for 3-7 highlights per article. If you're highlighting more than that, you're probably not being selective enough.

Habit 2: Annotate when a highlight triggers a thought. Not every highlight needs a note. But when a passage makes you think "this connects to X" or "I disagree because Y," capture that reaction in a brief annotation. These annotations are the highest-value context you can create, because they encode your unique perspective, not just what the author said.

Habit 3: Chat with your highlights weekly. Set aside 15-20 minutes each week to ask your AI questions about your recent highlights. "What themes have I been reading about this week?" "How does this week's reading connect to what I saved last month?" "Draft an outline for a memo on [topic] using my highlights as sources." This is where the compound value of PCM appears.

For your reading across books and e-readers, Kindle highlights can be imported and merged with your web highlights, creating a unified context layer across all your reading.

A sample weekly workflow:

The total additional time investment is roughly 30-45 minutes per week. The return is a continuously growing context layer that makes every AI interaction more personalized and every piece of writing more grounded in real sources.

---

What This Means for the Future of Learning

The implications of PCM extend beyond productivity. They reshape how we think about learning itself.

Ebbinghaus demonstrated that we forget 90% of new information within a week. For over a century, the primary countermeasure was spaced repetition: reviewing material at increasing intervals to strengthen memory traces. It works, but it's effortful. Most people don't sustain it.

PCM offers an alternative path. Instead of fighting the forgetting curve with brute-force repetition, you externalize your learning into a context layer and let AI handle retrieval. You still benefit from the initial engagement (highlighting forces active reading, which strengthens encoding). But you no longer need to maintain everything in biological memory. Your AI remembers for you, and it can synthesize across hundreds of sources in ways your brain never could.

This isn't cognitive offloading in the dangerous sense that Mollick and others warn about. The critical distinction, explored in depth in our piece on AI's impact on learning, is between outsourcing thinking and outsourcing retrieval. PCM outsources retrieval while keeping the thinking, the selection, the annotation, the synthesis, firmly human.

There's a second implication. As AI models improve, the value of your personal context increases, not decreases. A model that's 10x better at reasoning will produce 10x better output when given rich personal context versus generic prompts. Your highlight library is an appreciating asset.

And there's a third. Personal knowledge management was always somewhat solitary. You built your system, you used your system. PCM, especially with a social layer, creates network effects. Every person who highlights an article adds context that benefits others. The more people participate, the richer the collective context becomes. Knowledge compounds individually and collectively.

---

Frequently Asked Questions

How is Personal Context Management different from just using ChatGPT?

ChatGPT (or any LLM) without your context produces generic output based on its training data. PCM gives the AI your specific highlights, annotations, and reading history as input. The difference is like asking a stranger for advice versus asking a colleague who's read the same research you have. Same AI, radically different output quality.

Do I need to change my existing PKM setup?

No. PCM builds on top of existing PKM practices. If you already take notes in Obsidian or Notion, those notes can become part of your context layer. The main shift is in mindset: instead of organizing notes for your own retrieval, you're curating context for AI-assisted retrieval and synthesis.

How many highlights do I need before PCM becomes useful?

There's no strict minimum, but the compound effect becomes noticeable around 200-300 highlights (roughly 6-8 weeks of consistent reading and highlighting). At that volume, AI can start identifying patterns in your interests, surfacing connections between sources, and producing outputs that feel genuinely personalized.

Won't AI just make us lazier about reading?

The research is mixed but instructive. Passive AI use (asking for summaries without reading) does reduce comprehension and critical thinking. Active AI use, where you read first, highlight what matters, and then use AI to synthesize, actually improves learning outcomes. The key is that PCM keeps you reading and making judgments. The AI amplifies your engagement rather than replacing it.

Is my highlight data private?

On Glasp, you control the visibility of your highlights. Public highlights contribute to the social layer and help other readers. Private highlights remain yours alone. Either way, your highlight data is your data, and it can be exported at any time.

How does this relate to the "Second Brain" concept?

Think of PCM as Second Brain 2.0. The original Second Brain was designed for human retrieval: you organize information so you can find it. PCM is designed for AI-assisted retrieval: you curate context so AI can find, synthesize, and generate from it. The capture habits are similar. The downstream workflow is fundamentally different.

---

Conclusion: Your Highlights Are Your Competitive Advantage

The shift from PKM to Personal Context Management is not a trend. It's a structural change in how humans interact with information and AI.

Every highlight you save today becomes part of a context layer that makes your AI interactions more useful tomorrow. Every annotation adds your unique perspective to that context. Every week of consistent highlighting compounds into a richer, more personalized knowledge base that no one else has.

Generic AI output is a commodity. Personalized AI output, grounded in your reading, your thinking, your curated sources, is a competitive advantage. And the raw material for that advantage is remarkably simple: the passages you chose to highlight.

You don't need a complex system. You need a highlighting habit, a tool that makes it frictionless, and an AI layer that can draw on your context when you need it.

Start today. Read an article. Highlight what matters. Add a note when something sparks a thought. Do it again tomorrow. Within weeks, you'll have the foundation of a PCM system that grows more valuable with every piece you read.

Glasp makes this workflow seamless: highlight the web, summarize videos, chat with your knowledge, and learn from what others find valuable. Your highlights aren't just memories. They're the context that makes AI truly yours.

Export your highlights anytime. Build on any platform. The context you create is always yours.