Core Primitive
Information you might need later goes into a searchable reference system.
The information you saved but cannot find
In the previous lesson, you learned the three-outcome processing model: every piece of information that enters your life requires a decision — act on it, store it, or discard it. That lesson was about the decision. This lesson is about what happens when the decision is "store."
Specifically, it is about what happens when the information does not require action right now but might be useful later. A research finding you want to cite in a future project. A tax document you will need in April. A recipe your friend recommended. A framework from a book you read that applies to a problem you have not encountered yet. A contact's details. A product specification. A flight confirmation number.
This type of information has a name: reference material. It is static. It does not require you to do anything. It needs to be available when you need it and invisible when you do not. And for most people, it is the category of information that is most frequently saved and most frequently lost — not because they did not store it, but because they stored it in a way that made retrieval functionally impossible.
You have experienced this. Everyone has. You know the information exists. You remember encountering it. You might even remember roughly when and where. But you cannot find it. Your email has forty thousand messages. Your notes app has three hundred unsorted notes. Your desktop has documents named "stuff.docx" and "notes_final_v2.docx" and "misc." Your bookmarks folder has not been opened since the year you created it. The information is saved. The information is lost. These two conditions coexist more often than anyone wants to admit.
This lesson builds the system that eliminates that contradiction. Not a complex system. Not an elegant system. A system optimized for the only thing that matters when you need reference information: the speed and reliability of getting it back.
The retrieval principle
Here is the single most important idea in this lesson, and it contradicts what most people assume about filing: a filing system should be designed for retrieval, not for storage.
This sounds obvious until you watch yourself violate it. When you save a piece of information, your instinct is to categorize it — to put it in the "right" folder, to assign it the "correct" label, to slot it into a taxonomy that reflects the logical structure of the information itself. You file a pricing framework under "Business Strategy" because that is what it is about. You file a recipe under "Cooking" because that is its domain. You file a tax form under "Finance > Taxes > 2026" because that is where it logically belongs.
The problem is that your future self — the person who will actually need this information — does not think in terms of taxonomic categories. Your future self thinks in terms of the problem they are trying to solve. They are not thinking, "I need something from my Business Strategy folder." They are thinking, "What was that pricing framework where you tier based on customer value rather than cost-plus?" They are not browsing a taxonomy. They are running a search with the words that describe their current need.
This gap — between how you file and how you search — is the core failure mode of most reference systems. You file by category. You retrieve by context. Unless those align perfectly, retrieval fails.
David Allen, in "Getting Things Done" (2001), identified this problem and proposed a deliberately simple solution: a single alphabetical reference filing system. No complex folder hierarchies. No nested categories. Just clearly labeled files in alphabetical order, where each label reflects the first word you would think of when looking for that item. Allen's argument was that the friction of deciding where to file something is the primary reason people do not file at all, and that a system so simple it requires almost no filing decisions will actually get used — which matters infinitely more than whether the taxonomy is theoretically optimal.
Allen was writing in the era of physical file cabinets, but the principle he identified is more relevant now than ever. Digital tools have infinite storage and powerful search. The bottleneck is not space. It is not organization. It is naming and describing items in a way that makes them findable when you need them.
The filing cabinet metaphor versus the search engine metaphor
Most people's mental model of filing comes from the physical filing cabinet: folders organized in a hierarchy, each item placed in exactly one location, retrieval accomplished by navigating the tree structure to the correct leaf node. This model worked tolerably well when the number of items was small and the categories were stable. It breaks catastrophically when the number of items grows into the thousands and the categories are ambiguous.
The ambiguity problem is structural, not personal. A pricing framework can legitimately be filed under "Business Strategy," "Pricing," "Frameworks," "Conference Notes," or "Project: Revenue Model." Each category is defensible. None is uniquely correct. And the more categories are available, the less likely you are to remember which one you chose — because the choice was somewhat arbitrary, and arbitrary choices are precisely the ones memory discards first.
The search engine metaphor replaces this entirely. In a search-based system, you do not navigate to the item. You describe the item, and the system finds it. The item does not live in a single location within a hierarchy. It lives in a flat pool of content, and it surfaces when your search terms match its content, title, or tags. The filing question is no longer "where does this go?" but "what words will I use when I need this back?"
This is a profound shift, and it has a concrete operational implication: when you save a reference item, write the title and description in the words your future self will search for. Not the words that describe the item's category. Not the words the author used. The words you will type into a search box six months from now when you need this information and have forgotten nearly everything about it except the problem it solves.
A note titled "Conference Notes Feb 2026" is filed for storage. A note titled "Value-based pricing tier framework — Samir Patel, InfoConf 2026" is filed for retrieval. The second title contains every word you might plausibly search for: the concept (value-based pricing), the specific topic (tier framework), the source (Samir Patel), and the context (InfoConf 2026). Finding it later requires matching on any of those terms. Finding the first title requires remembering the exact month and the fact that you attended a conference, which tells you nothing about the content.
The systems people actually use
Several well-known methodologies have emerged for organizing reference material. Each has a different philosophy, but they converge on the same operational truth: simplicity in structure beats elegance in taxonomy.
David Allen's GTD Reference Filing. Allen recommends a single, alphabetical filing system — physical or digital — where each item gets a plain-language label. No subcategories. No hierarchy deeper than one level. The test is: can you create a new folder and file something in under sixty seconds? If not, the system is too complex and you will stop using it. Allen's system optimizes for the speed of filing, which he argues is the binding constraint. A perfect taxonomy that creates filing friction will accumulate an unfiled backlog. A simple system that accepts items instantly will actually capture everything.
Tiago Forte's PARA Method. Forte, in "Building a Second Brain" (2022), proposes organizing all digital information into four top-level categories: Projects (active, with a deadline), Areas (ongoing responsibilities), Resources (topics of interest for future reference), and Archives (completed or inactive items). Reference material in the PARA model goes primarily into Resources — organized by topic rather than by source — with the principle that items should be filed where they will be most useful, not where they came from. Forte's key insight is that organization should reflect actionability. Projects are most active, Areas are ongoing, Resources are future-use, and Archives are dormant. Information flows through these categories as its relevance changes.
The Johnny Decimal System. Johnny Noble's system assigns structured numerical identifiers to categories: ten broad areas (00-09 through 90-99), each containing up to ten categories. The constraint is deliberate — you cannot have more than one hundred categories total, which forces you to keep the taxonomy shallow and manageable. The numbering system provides a universal address for every item, making it equally navigable in digital and physical environments. The discipline of the numerical limit prevents the category sprawl that kills most hierarchical filing systems.
The Noguchi Filing System. Yukio Noguchi, a Japanese economist, proposed the simplest possible filing method: file everything chronologically. New items go on the left side of the shelf. When you retrieve an item, it returns to the left side. Over time, frequently used items migrate toward the left and rarely used items drift toward the right. The system requires zero categorization decisions. It optimizes entirely for recency and frequency of use. In the digital world, this is roughly equivalent to relying on "recently modified" sorting — a strategy that works surprisingly well for items accessed regularly and fails for items accessed rarely.
Each of these systems works. None of them is universally best. The choice depends on your volume of reference material, your retrieval patterns, and — critically — your willingness to maintain the system over time. The best reference filing system is not the most theoretically sound one. It is the one you will actually use consistently, because an imperfect system that captures everything is infinitely better than a perfect system that sits empty while your reference material scatters across fifteen different locations.
The two-click rule
Regardless of which system you adopt, there is a practical benchmark that separates functional reference systems from decorative ones: any reference item should be retrievable within two actions. One search, one click. Or two clicks through a shallow hierarchy. Or one command and one scan.
If retrieval takes longer than that — if finding a single reference item requires navigating through four levels of folders, remembering which category you used, scrolling through hundreds of items within that category, and then opening several candidates to find the right one — the system is failing its only purpose. You will stop using it. You will resort to re-Googling, re-asking, re-reading, and re-discovering information you already captured, wasting time that the system was supposed to save.
The two-click rule is not arbitrary. It reflects a psychological threshold: the point at which the effort of retrieval exceeds the effort of re-acquisition. If finding something in your reference system takes three minutes, and finding it on Google takes two minutes, your reference system has negative value. It cost you time to file the item and more time to retrieve it than simply finding it fresh. Rational behavior, in that case, is to stop filing entirely — which is exactly what most people do with systems that are elaborate to file into and slow to search.
The operational implication: test your system regularly. Pick a reference item at random and time your retrieval. If it takes more than fifteen seconds, diagnose why. Is the title wrong? Are the tags missing? Is the folder structure too deep? Is the search function inadequate? Each slow retrieval is diagnostic data about your system's fitness.
Reference versus working notes
There is a distinction that most filing advice elides, and getting it wrong will corrupt your reference system: the difference between reference material and working notes.
Reference material is static. It is information you store in its finished form, to be looked up and used as-is. A tax document. A product specification. A citation with its key finding. A contact's details. A recipe. A policy document. You do not modify reference material — you consult it.
Working notes are active. They are your thinking in progress — ideas being developed, drafts being revised, connections being drawn. A project plan that changes weekly. Meeting notes that need to be processed into action items. A research document where you are synthesizing sources into an argument. Working notes are alive. They evolve. They belong in a different space than reference material, because the relationship you have with them is different: you interact with working notes, and you retrieve reference material.
When these two types share the same system without distinction, the system degrades. Your reference searches surface incomplete drafts. Your working notes are buried under static documents. The reference system — which should be clean, stable, and searchable — becomes cluttered with items that are still in flux. And the working space — which should be dynamic and current — becomes weighed down by items that are finished and should be filed away.
The fix is simple: maintain a clear boundary. Reference material goes into the reference system. Working notes live in whatever workspace you use for active projects. When a working note is finished — when the draft is complete, the meeting notes have been processed, the research has been synthesized — the useful output gets filed as reference, and the working note gets archived or deleted. This is the lifecycle: active work produces reference artifacts, and reference artifacts go into the filing system where they become findable for future use.
The paradox of elaborate filing systems
There is a counterintuitive finding that runs through the research on personal information management: people who spend the most time organizing their files do not retrieve them faster than people who spend almost no time organizing.
A 2006 study by Boardman and Sasse at University College London found that the most meticulous filers — people with deep folder hierarchies and detailed categorization — spent significant time on filing decisions but did not retrieve information more quickly than people who used shallow structures and relied on search. In some cases, the meticulous filers were slower, because they had to remember which category they chose — a recall task that gets harder as the number of categories grows.
This finding does not mean organization is pointless. It means that the return on organizational complexity diminishes rapidly. A few broad categories plus good search beats an elaborate taxonomy in almost every practical scenario. The first level of organization — having a single, consistent home for reference material — produces enormous retrieval gains over having no system at all. The second level — basic categorization or tagging — adds incremental value. Everything beyond that adds filing cost without proportionate retrieval benefit.
The practical recommendation is: start simple. Radically simple. One location. Descriptive titles. A few broad tags if your system supports them. Rely on search for retrieval. Add structural complexity only when you have concrete evidence that search alone is failing — when you are regularly unable to find items that you know you filed. In most cases, the evidence never arrives, because search with good titles is remarkably effective.
Building your reference filing system: the operational steps
Here is a concrete protocol for building a reference filing system that actually works, regardless of which tool you choose.
Step one: Choose a single default location. This is where reference material goes unless there is a compelling reason to put it elsewhere. One app. One folder. One location. The specific tool matters far less than the consistency. Notion, Obsidian, Apple Notes, Google Drive, a file system folder, Evernote — any of these work. The fatal choice is distributing reference material across all of them, because then retrieval requires searching six different locations and hoping you guess correctly which one contains the item.
Step two: Establish a naming convention. Every reference item gets a title written for your future searching self. The format should include what the item is about and enough context to distinguish it from similar items. "Value-based pricing framework — Samir Patel" is better than "Pricing notes." "Home insurance policy — State Farm — 2026" is better than "Insurance." Front-load the title with the most searchable terms.
Step three: Use tags sparingly. If your system supports tagging, use a small number of broad tags — ten to twenty at most. Tags are useful for browsing ("show me everything tagged 'health'") but they are not a substitute for good titles. The danger of tagging is that it becomes a second categorization system layered on top of the first, doubling the decision cost without doubling the retrieval benefit. Each tag should represent a context in which you are likely to want multiple items at once.
Step four: Adopt a processing habit. Reference material does not file itself. Decide when you will process incoming reference items — daily, at the end of each work session, during a weekly review — and protect that time. The processing step is quick: give the item a searchable title, assign one or two tags if applicable, and drop it into the default location. Thirty seconds per item. If it takes longer than that, your system is too complex.
Step five: Maintain ruthlessly. Every few months, scan your reference system for items that are outdated, duplicated, or no longer relevant. Delete them. A reference system that grows without pruning eventually drowns its useful items in noise, degrading search quality. The goal is not to keep everything. The goal is to keep what you might actually need and to keep it findable.
The Third Brain: AI as retrieval amplifier
AI transforms reference filing in two specific ways, and both of them operate on the retrieval side rather than the filing side.
First, AI-powered search understands meaning, not just keywords. Traditional search requires you to match the exact words in the document. If you titled a note "value-based pricing tiers" and later search for "customer-value pricing structure," traditional search misses it. Semantic search — the kind built into tools like Obsidian with AI plugins, Notion AI, and others — understands that these phrases describe the same concept and surfaces the match. This does not eliminate the need for good titles, but it provides a safety net for the cases where your filing vocabulary and your retrieval vocabulary diverge.
Second, an AI assistant can serve as a retrieval interface for your entire reference system. Instead of searching through your files yourself, you describe what you are looking for conversationally — "I need that framework about pricing that I picked up at a conference last year, where you tier pricing based on the value the customer receives rather than cost-plus" — and the AI searches your reference system using that description. This is particularly powerful when you remember the concept but not the terminology, which is the exact scenario where traditional keyword search fails.
There is a more structural application as well. When you are building your reference system, you can describe your filing to an AI and ask it to identify retrieval weaknesses. Paste in ten of your note titles and ask: "If I needed each of these items six months from now and could only remember the general topic, would these titles let me find them?" The AI will flag the vague ones, the ambiguous ones, and the ones that use jargon you are likely to forget — improving your naming convention before retrieval failure teaches you the same lessons more painfully.
The sovereignty principle holds: the AI does not decide what is worth filing. It does not determine your reference categories. It amplifies your retrieval capability so that the filing system you built delivers on its promise — making your stored knowledge available when you need it.
The bridge to action
You now have a home for information that does not require action but might be needed later. The reference filing system accepts processed information, stores it in a searchable form, and returns it when you need it. The design principles are straightforward: optimize for retrieval over storage, keep the structure simple, name items for your future searching self, and rely on search rather than taxonomy.
But reference is only one of the two "store" outcomes from Processing means deciding what to do with each item's processing model. The other outcome is information that requires you to do something — an email that needs a reply, an idea that needs to be scheduled, a commitment that needs to be tracked. That information does not belong in your reference system. Putting it there buries it among static items, where it will sit unacted upon because nothing in the reference system is designed to surface items that need attention.
Information that requires action needs a different home — one that is designed not for retrieval on demand but for surfacing at the right time. That home is the action filing system, and it is the subject of the next lesson.
Sources:
- Allen, D. (2001). Getting Things Done: The Art of Stress-Free Productivity. Viking.
- Forte, T. (2022). Building a Second Brain: A Proven Method to Organize Your Digital Life and Unlock Your Creative Potential. Atria Books.
- Noble, J. (2020). Johnny.Decimal: A System to Organize Your Life. johnny.decimal.
- Noguchi, Y. (1993). The "Super" Organization Method (「超」整理法). Chuko Shinsho.
- Boardman, R., & Sasse, M. A. (2004). "Stuff goes into the computer and doesn't come out": A cross-tool study of personal information management. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 583-590.
- Jones, W. (2007). Keeping Found Things Found: The Study and Practice of Personal Information Management. Morgan Kaufmann.
- Barreau, D., & Nardi, B. (1995). Finding and reminding: File organization from the desktop. ACM SIGCHI Bulletin, 27(3), 39-43.
Frequently Asked Questions