# Working Memory URL: https://mechanisticmindset.com/wiki/working-memory Tags: core-framework, computational-lens # Working Memory #core-framework #computational-lens ## What It Is Working memory is the active space in which you hold information while thinking. It works like a computer's **RAM**: a small space that holds what is in use right now. Its capacity is approximately **4–7 items** at once. The limit is a hard constraint. When you try to hold more than fits, some items are pushed out. You lose your place or forget a consideration because keeping the new item active displaces another one. ## Why This Matters A complex problem requires several things to remain available together. You need component A and its properties, component B and its relation to A, the goal, the constraints, and the next step in the reasoning. If that is more than fits, the whole problem cannot be held in mind at once. Thinking becomes a repeated process of loading one part and losing another. A project feels overwhelming because the information needed to reason about it exceeds the space available. ## The Constant Overflow Problem Suppose you begin with component A, then need to consider B. Loading B pushes A out. The next step needs both A and C, so you have to recover A while retaining C. With only a few slots available, keeping the relevant material active requires constant swapping. That extra work is separate from solving the problem. You are maintaining an overflowing buffer, recovering lost pieces and trying to remember where the reasoning stopped. “Just think harder” does not help because effort does not increase its capacity. ## Externalization Solves This Writing down A changes what happens when you move to B. A remains on the page while B occupies your attention. After writing down B, you can work on C and D without needing to keep A and B active too. The page becomes extended memory. You read the relevant pieces, process them, and write the result back. Working memory can do the current operation while the page holds what is needed later. The computing comparison is between an in-place operation, restricted to the existing input data structure, and an operation that uses additional storage. The first can be elegant but limited. The second makes larger operations tractable. [Journaling](/wiki/journaling) provides that extra storage for reasoning that would otherwise have to remain entirely in working memory. ## Why "Hold That Thought" Fails An interruption makes the difference visible. If the thought has only a few parts, you may be able to recover it afterward. A thought with more parts than fit comfortably was being kept active by continuous rehearsal, and the interruption breaks that loop. When you return to it, some of it is missing. [Interruptions](/wiki/focus) therefore cost more than the minutes they occupy. They can take with them a complex thought that was available only because it was being maintained. A written record preserves material that would otherwise depend on the rehearsal continuing unbroken. ## Complex Thoughts Require Development Over Time A complex idea outgrows what you can hold in your head as it develops. The first day's vague intuition may fit, while an attempt to expand it exceeds capacity and loses its own thread. Writing three sentences on Day 1 preserves the current understanding. On Day 2, you can read them and add another piece. By Day 7, the accumulated record reveals connections that were not visible at the beginning. By Day 30, the thought has become three pages. Each session adds an amount small enough to process at once. The journal holds the combined result, allowing the idea to grow beyond the capacity of any one session's working memory. ## [Procrastination](/wiki/procrastination) and Working Memory “Work on project” leaves several questions unresolved. Which project is it? What is its current state? What needs to happen next? What is the first concrete action, what blocks it, and what would count as success? To begin, those questions and their answers must be loaded together. Six questions and their six answers come to twelve items, well past what fits. The task feels impossible in the way a program too large for memory will not load. The reluctance to start follows from the capacity constraint. ### Solution: [The Braindump](/wiki/the-braindump) Ten minutes of writing moves the accumulated material onto a page and clears working memory. Reading it usually reveals one priority, one specific blocker, and one concrete first action. The [activation cost](/wiki/activation-energy) of starting falls because the cognitive work has been externalized. Beginning no longer requires holding the whole unresolved project in mind. ## Pattern Recognition Across Time A month of behavior contains 30 days multiplied by dozens of variables: hundreds of observations. Holding them in working memory to find a pattern is impossible. A journal makes the comparison sequential. Read the entries and mark good days and bad days. Examine the good days for a pattern, then the bad days, then compare the results. Working memory handles each small operation while the record retains the remaining data. Computers process large datasets in the same way: read a chunk, process it, write the result, then read the next chunk. External storage allows the full computation to exceed what fits in active memory. ## [The Braindump](/wiki/the-braindump) as Preprocessing The morning braindump prepares material for that kind of analysis. Writing empties the items occupying working memory, frees the processing space, and gives the complexity a persistent location. Reading the result then becomes possible without having to preserve every other part mentally. The ambiguity that was consuming capacity is externalized and resolved. The activation cost falls, and work can start because the next action is now identifiable. The dump is therefore a preparation step for execution, as well as a record. ## Why Whiteboards Work A whiteboard keeps external memory continuously available. Walking past it cues logging without requiring you to remember to open a tool. Previous entries remain visible, so their accumulation does not require repeated recall. Digital tracking puts the record behind an interface. Opening the application requires loading that information into working memory. A whiteboard leaves it in the physical environment, reducing the amount that needs to be kept active in your head. ## The Same Limit in Ordinary Tasks The same capacity constraint affects several ordinary tasks: - A list of 15 todos cannot be held in mind at once. Grouping it into 3–5 chunks makes it manageable. - Learning 10 new concepts simultaneously exceeds capacity. Learning 3–5, mastering them, then moving to the next 3–5 keeps the current set small enough. - Debugging requires the current function and 3–5 related pieces, rather than the entire codebase at once. - Conversation cannot sustain 10 separate discussion threads, and a large group stops holding a single conversation together. Any system with more parts than can be held at once needs either external storage or chunking. Chunking is also where the comparison with RAM stops: a byte has a fixed size, while an item in working memory is whatever you can treat as one thing. ## Question Design and Working Memory A [question](/wiki/question-theory) must fit within the capacity available to answer it. “How can I be better?” starts a search through every concept, relationship, and possibility in the knowledge graph. The search has no boundary and nothing that tells it when to stop, so it either does not finish or returns whichever result happens to be salient. A bounded question reduces the search to a handful of items. “What's one improvement to work launch sequence?” selects a particular system. “What's the next action on highest-priority task?” narrows the request further to one prioritized item. The constraints make a search possible that would otherwise overflow. The answer also needs a bound. “Tell me about willpower” permits indefinite content. “What are three ways willpower depletes?” supplies both a topic and a stopping condition. [Effective questions](/wiki/question-theory) use these LIMIT clauses to keep the result set within working-memory capacity. ## Related Concepts - [Journaling](/wiki/journaling) — external storage for reasoning across time - [The Braindump](/wiki/the-braindump) — clearing and preparing the current material - [Procrastination](/wiki/procrastination) — tasks whose unresolved complexity exceeds capacity - [Focus](/wiki/focus) — maintaining one thread without fragmenting it - [Tracking](/wiki/tracking) — recording system state instead of recalling it - [Question Theory](/wiki/question-theory) — keeping a question small enough to hold in mind - [Pedagogical Magnification](/wiki/pedagogical-magnification) — matching resolution to capacity - [Effective AI Usage](/wiki/effective-ai-usage) — AI as an extension of working memory ## Key Principle When a problem has more parts than can be held at once, a journal, whiteboard, or task tracker can hold the parts that do not need attention at this moment. That lets working memory perform the next operation without also preserving the entire problem.