Not all knowledge ages the same way
In 1962, economist Fritz Machlup at Princeton coined the term "half-life of knowledge" — the time it takes for half of what is known in a field to be superseded or proven wrong. At the time, an engineering degree had a knowledge half-life of about ten years. An engineer could graduate, work for a decade, and still be operating with mostly current expertise. By 1991, that half-life had dropped to five years. A decade after that, it was two and a half — shorter still if the engineer worked in information technology.
This was not a metaphor. Machlup was measuring a real phenomenon: knowledge decays at rates that are predictable in aggregate, even when any single fact's expiration date is unknowable in advance.
Samuel Arbesman formalized this insight in The Half-Life of Facts (2012). He studied nearly 500 published articles on liver disease and found a strikingly clean decay curve: clinical knowledge about cirrhosis and hepatitis had a half-life of roughly 45 years. Half of what hepatologists believed in 1960 was wrong or obsolete by 2005. A study in a physics journal loses half its influence in about 10 years; in urology, the equivalent figure is 7.1 years.
The mechanism is not that old facts become lies. It is that better measurements replace rough ones, exceptions erode generalizations, and context shifts make once-useful frameworks misleading. The knowledge does not vanish — it gets displaced by knowledge with higher resolution.
This means your information diet is not a single stream. It is a mixture of materials with wildly different shelf lives, and your job is to know which is which.
The decay spectrum: flash, cycle, bedrock
Not every piece of information decays at the same rate. A useful way to think about your inputs is to sort them into three categories based on their functional half-life.
Flash information has a half-life of hours to days. Stock prices, trending topics, most notifications, breaking news, social media discourse, Slack messages about what is happening right now. Flash information is not inherently useless — a stock trader needs it, a journalist needs it — but for most knowledge workers, the ratio of flash to durable input is catastrophically skewed. You consume it because it feels urgent. It is almost never important.
Cycle information has a half-life of months to a few years. Industry trends, current best practices, framework documentation, quarterly analysis, market reports. This is the working layer of professional knowledge. It is necessary and genuinely useful — but it expires. The React documentation you read this year will be partially outdated next year. The management framework your company adopted will be replaced. Cycle information keeps you competent now but does not compound over time.
Bedrock information has a half-life of decades to centuries. First principles, mental models, foundational theories, human psychology, logic, rhetoric, mathematics, physics. Charlie Munger's mental models work because they are drawn from fields whose core insights are old — incentive theory, supply and demand, second-order effects, cognitive biases. These ideas were true before you were born and will be true after you die. Euclid's geometry is 2,300 years old and still taught. Adam Smith's observations about specialization are 250 years old and still govern how companies organize.
The half-life of medical knowledge about hepatitis is 45 years. The half-life of Aristotle's Nicomachean Ethics on practical wisdom is — so far — about 2,400 years and counting. That is not a difference of degree. It is a difference of kind.
The question this creates for your epistemic practice is direct: what percentage of your attention budget goes to each layer?
The Lindy effect: survival as evidence
Nassim Nicholas Taleb gave this pattern a name in Antifragile (2012): the Lindy effect. The concept, originally observed informally among comedians at Lindy's delicatessen in New York in the 1960s, states that for non-perishable things — ideas, books, technologies, cultural practices — the longer something has survived, the longer its remaining life expectancy.
A book that has been in print for 40 years can be expected to remain in print for another 40. If it survives another decade beyond that, its life expectancy extends to 50 more years. This is the opposite of biological aging, where the longer something has lived, the closer it is to death. Ideas age in reverse. Survival is evidence of robustness.
Taleb's reasoning is straightforward: time is equivalent to disorder. Anything that has endured centuries of changing contexts, competing ideas, shifting cultures, and new evidence has demonstrated resistance to disorder. It has been stress-tested by history in a way that no peer review or popularity metric can replicate. A philosophy book from 50 AD that people still read is more likely to contain durable insight than a bestseller from last month — not because old automatically means good, but because time has already done the filtering that you would otherwise have to do yourself.
The practical implication for your information diet is this: when choosing what to read, study, or internalize, the age of the source is informational. Not dispositive — old ideas can be wrong — but informational. A book that has been continuously read for a century has passed a filter that a book published last Tuesday has not.
Nat Eliason captured this hierarchy cleanly: "The more ephemeral a piece of information is, the less likely it is to be valuable." He mapped it across media types. CNN's 24-hour news cycle is maximally ephemeral and typically low-value. A quarterly publication like Foreign Affairs is less ephemeral and correspondingly more curated. A book that remains popular for decades has survived a brutal selection process and is "almost without exception fantastic."
This is not snobbery. It is time management. If you only have two hours to read today, spending them on something Lindy-tested is a better expected-value bet than spending them on something published this morning.
Why your knowledge system needs a depreciation model
Every business depreciates its assets. A laptop purchased this year will be worth less next year and worthless in five. Accountants model this explicitly. They do not pretend that a five-year-old server has the same value as the day it was installed.
Your knowledge system deserves the same honesty. The industry analysis you captured eighteen months ago may now be misleading. The technical tutorial you bookmarked two years ago may reference deprecated APIs. The market thesis you wrote in your notes last year may have been invalidated by events you have not yet accounted for.
Most personal knowledge management systems treat all captured information as equally valid regardless of age. Your bookmarks from 2019 sit alongside your bookmarks from yesterday with no indication that their reliability has changed. Your notes do not signal which claims have been superseded. Your reading list does not distinguish between sources that will be relevant in five years and sources that will not be relevant next month.
This is the knowledge equivalent of never updating your inventory. You accumulate without auditing, and over time your system fills with information that is not just unhelpful but actively misleading — outdated mental models masquerading as current understanding.
Andy Matuschak identified this pattern in his distinction between evergreen and transient notes. "Most people take only transient notes," he writes. "After a year of writing such notes, they'll just have a pile of dissociated notes. The notes won't have added up to anything: they're more like fuel, written and discarded." Transient notes are not bad — they serve working memory — but they are flash-layer material. If your entire knowledge system is built from transient captures, you have a system optimized for recency with no mechanism for durability.
Matuschak's alternative — evergreen notes — are "written and organized to evolve, contribute, and accumulate over time, across projects." The key word is accumulate. Evergreen notes are bedrock-layer material: insights stated precisely enough that they remain useful months or years after you wrote them. They are the knowledge equivalent of assets that appreciate rather than depreciate.
Sönke Ahrens makes the same point in How to Take Smart Notes. Luhmann's Zettelkasten worked because each card was an atomic idea — a single claim, stated clearly, connected to other claims by context rather than category. The system Luhmann built over 40 years (90,000+ notes, 70+ books, 400+ articles) worked because the notes were bedrock. They were not summaries of what Luhmann read on a particular Tuesday. They were distilled principles that stayed relevant as his thinking evolved across decades.
The failure mode is not capturing too much. The failure mode is capturing without discriminating — giving a trending take the same structural weight as a foundational principle, and then wondering why your notes feel less useful over time.
Building a time-aware information practice
Once you see the decay spectrum, you can restructure your attention and capture systems around it.
Audit your inputs by half-life. Look at where your reading time actually goes. Most people discover that 70% or more of their information consumption is flash-layer — news, social media, notifications, chat. This is not because flash information is bad. It is because flash information is engineered to capture attention. Its producers have optimized for urgency, novelty, and emotional activation — all of which trigger consumption regardless of durability. Your job is to notice this and rebalance.
Weight your capture system toward bedrock. When you encounter a bedrock-layer insight — a mental model, a first principle, a pattern that has been validated across contexts — treat it differently from a cycle-layer note. Give it a permanent home. State it in your own words. Connect it to other bedrock ideas. This is what Matuschak means by evergreen notes and what Ahrens means by permanent notes in the Zettelkasten. Not everything you read deserves this treatment. The things that do deserve more time and care than you are probably giving them.
Tag captures with decay estimates. This is the integration step above, and it bears repeating because it is the single highest-leverage change you can make to a knowledge system. When you save something — a highlight, a bookmark, a note — ask: will this still be true and useful in a week? A year? A decade? Tag accordingly. Over time, this practice trains your attention to discriminate automatically. You start noticing, as you read, whether a sentence is flash, cycle, or bedrock. That noticing is the skill.
Depreciate actively. Set a quarterly review to scan your notes and bookmarks for information that has decayed. Technical documentation that has been superseded. Market analyses that have been invalidated. Predictions that have been resolved. Either update these items, archive them, or delete them. A knowledge system that never depreciates is a knowledge system that lies to you with increasing frequency over time.
The Third Brain layer: AI as depreciation engine
This is where AI changes the game for knowledge management at the personal level.
AI systems are increasingly capable of flagging outdated content automatically. In enterprise knowledge management, 70% of organizations now use AI-powered systems to identify and prune redundant, outdated, or trivial content. The same principle applies to your personal knowledge base.
An AI assistant that has access to your notes can do something you cannot do efficiently at scale: compare your captured claims against current evidence. That market analysis from 2024 — is the data still valid? That technical pattern you documented — has the framework deprecated it? That mental model you imported from a book — has subsequent research refined or contradicted it?
This is not about replacing your judgment. It is about extending your judgment across a larger surface area. You can review 20 notes carefully in a sitting. AI can scan 2,000 and surface the ones most likely to have decayed since you captured them.
The deeper application is building a personal knowledge base with built-in temporal awareness. Instead of a flat collection of notes where everything has equal standing, you build a system where bedrock ideas are structurally privileged — more connected, more frequently surfaced, more central to your thinking workflows — while flash and cycle information is explicitly temporary, scheduled for review and depreciation.
The goal is not to eliminate ephemeral information from your life. You need cycle-layer knowledge to do your job. You sometimes need flash-layer information to respond to immediate events. The goal is to stop treating all three layers as if they have the same value, the same shelf life, and the same claim on your attention.
Protocol: the half-life audit
- List your top 10 information sources — the feeds, publications, people, and platforms you consume most regularly.
- Classify each as primarily flash, cycle, or bedrock.
- Calculate your current ratio. If more than 60% of your sources are flash-layer, you are over-indexed on recency.
- Identify two bedrock sources to add — books, longstanding publications, primary texts in your field that have been validated over decades.
- Identify two flash sources to remove or downgrade — move them out of your primary feed, check them once a week instead of daily, or unsubscribe entirely.
- Apply decay tags to your next 20 captures. Every time you save something this week, label it flash, cycle, or bedrock. At the end of the week, review the distribution.
- Schedule a quarterly depreciation review. Pick a date. Put it on your calendar. When it arrives, scan your saved notes and bookmarks for items that have decayed past usefulness. Archive or delete them.
The point is not to become an information ascetic. The point is to match your attention allocation to the actual durability of what you are consuming. Time is the scarcest resource in your epistemic practice. Spending it on information that will be irrelevant by Friday is a choice — and usually an unconscious one.
In the next lesson — Signal compounds and noise dilutes — we examine what happens when you consistently prioritize durable information over ephemeral information: the compounding effect that makes bedrock knowledge exponentially more valuable over time, and the dilution effect that makes noise consumption actively corrosive to clear thinking.
Sources
- Fritz Machlup, The Production and Distribution of Knowledge in the United States (Princeton University Press, 1962). Coined "half-life of knowledge" and measured decay rates across professions.
- Samuel Arbesman, The Half-Life of Facts: Why Everything We Know Has an Expiration Date (Current/Penguin, 2012). Systematic study of knowledge decay rates across scientific disciplines.
- Nassim Nicholas Taleb, Antifragile: Things That Gain from Disorder (Random House, 2012). Formalized the Lindy effect — the principle that non-perishable things gain life expectancy with age.
- Andy Matuschak, "Evergreen Notes" and "Most People Take Only Transient Notes," notes.andymatuschak.org. Distinction between transient and evergreen knowledge in personal systems.
- Sönke Ahrens, How to Take Smart Notes (2017, 2nd ed. 2022). Zettelkasten method and atomic notes as bedrock-layer knowledge practice.
- Nat Eliason, "Ephemerality vs. Value for Information, Social Media, Life," nateliason.com. Framework mapping information ephemerality to expected value.
- IEEE Spectrum, "An Engineering Career: Only a Young Person's Game?" (2012). Data on shrinking half-life of engineering knowledge from 35 years (1930) to under 5 years (2000s).