Learn it → Apply it → Save as Skill
Working memory can hold only a few new things at once — everything beyond that is wasted effort, not learning. Whenever someone reads your documentation, sits through your training, uses your interface, or follows your instructions, they are working inside that narrow limit. How you organize and present information decides whether their effort builds understanding or just drains the tank before anything sticks.
The Dr. Wallace's Favorite Mental Models is designed to turn useful ways of thinking into reusable tools. Each model gives you a fast visual overview, deeper interactive learning, practical application exercises and an AI-ready Markdown skill you can save for future use.
Instead of merely reading about mental models, you can gradually build your own AI skills library: a collection of portable reasoning skills that can help ChatGPT, Claude, Grok and other AI assistants apply the mental models you value.
Before you write, teach, design, or instruct, ask how much of the audience's limited working memory your material actually needs for the ideas themselves — and how much it's wasting on clutter, confusion, or poor sequencing that has nothing to do with the subject.
Important: working memory holds roughly three to five new chunks at a time. Most slides, manuals, and onboarding flows are built as if it holds dozens.
Most useful whenever you're designing something another person (or an AI system following instructions) has to read, learn, or follow for the first time.
Cognitive load theory names three kinds of mental effort. These four moves turn the theory into something you can actually do to a document, a lesson, an interface, or a prompt.
Separate the three loads: intrinsic (the subject is genuinely hard), extraneous (the presentation is making it harder than it needs to be), and germane (productive effort that's actually building understanding).
Remove extraneous load first — it's the only kind that's pure waste. Cut redundant text that repeats an image, merge split information that forces the eye to jump around, and delete decoration that isn't carrying meaning.
Break intrinsic complexity into smaller pieces sequenced from simple to complex, so no single step demands more than working memory can hold at once.
Protect germane load: use worked examples before independent practice, then fade the scaffolding as real expertise grows, so effort keeps going toward durable understanding instead of disappearing into confusion.
Open a use case to see what the model changes about how you'd approach it.
A reference page that forces readers to hold five undefined terms in mind while parsing a sixth is spending their working memory on your prose, not your product. Define terms before you need them, put code next to the text that explains it instead of in a separate panel, and cut anything that doesn't change what the reader does next.
Giving a new hire or new user everything on day one feels thorough and teaches almost nothing. Sequence the load: the two or three things they need to survive today, then the next layer once the first is automatic, not all of it at once because it's convenient to say once.
Every extra visible option, label, or notification competes for the same small working-memory budget as the task the user actually came to do. Progressive disclosure — showing advanced controls only once someone needs them — is cognitive load management with a different name.
Packing a session with every fact you know about a topic maximizes what you said, not what anyone absorbed. Teach the smallest version that's still true, let it become automatic, and only then add the exception that makes it more complete.
A dense slide read aloud word-for-word forces the audience to choose between reading and listening — the classic split-attention problem. One idea per slide, said once, in whichever format (spoken or written) carries it best, leaves more capacity for actually thinking about it.
A system prompt or skill file that buries the one rule that matters inside ten paragraphs of context is asking a model to do its own extraneous-load stripping under time pressure. Lead with the decision that matters, chunk the rest, and the failure rate on exactly the instructions you cared most about tends to drop.
This is an illustrative example designed to teach the method, not a claim about any real company or study.
Expand for who developed it, the three load types, where the capacity numbers come from, and how the theory has been challenged since.
John Sweller developed cognitive load theory in the late 1980s through his research on problem-solving, aiming — in his own framing — to provide guidelines for presenting information in a way that helps learners' intellectual effort go toward understanding rather than toward fighting the format it arrived in.
Intrinsic load is the inherent difficulty of the material itself — calculating 2+2 versus solving a differential equation — and instructors can't remove it, only break it into smaller schemas. Extraneous load comes from how the material is presented and is squarely within the designer's control — describing a shape in words instead of simply showing it is a classic example. Germane load is the effort that goes into actually building durable mental schemas from the material; it rises when intrinsic load is high and extraneous load is kept low, which is the combination that supports real learning.
It traces to psychologist George A. Miller's influential 1956 paper describing short-term memory capacity as "seven, plus or minus two" chunks. Later research has generally revised that estimate downward for working memory specifically — commonly cited at closer to three or four chunks without active rehearsal — but the underlying point Miller established still holds: the limit is real, small, and the same for everyone doing genuinely new mental work.
Yes. Subsequent research has questioned whether the three load types simply add together the way early versions of the theory assumed, suggesting instead that they may influence each other rather than combine additively. There's also ongoing debate over whether germane load is really an independent category or mostly just a byproduct of how intrinsic and extraneous load are managed. A 2023 review in Educational Psychology Review, "The Development of Cognitive Load Theory: Replication Crises and Incorporation of Other Theories Can Lead to Theory Expansion," traces this debate and argues the theory has had to expand and absorb ideas from other frameworks in response, rather than standing unchanged since the 1980s.
Germane load is the productive, effortful thinking that actually builds understanding. Stripping it out along with the genuine clutter leaves something easy to skim and hard to retain.
Narrating a slide word-for-word while the audience reads the same words is the textbook redundancy effect — it adds load, it doesn't add clarity.
Heavy scaffolding that helps a novice can actively slow down an expert, who now has to wade through explanations they don't need to get to the part they're looking for.
Cutting words that were actually carrying meaning doesn't reduce extraneous load — it just removes content, often pushing more of the real difficulty back onto the reader to reconstruct.
A single document, course, or interface rarely serves true beginners and experienced users equally well — what minimizes load for one group often maximizes it for the other.
Some subjects are genuinely hard — that's intrinsic load, and no amount of clever formatting removes it. The model manages that difficulty; it doesn't make it disappear.
heavy scaffolding and step-by-step chunking aimed at novices can slow experts down (the expertise-reversal effect) — match the support to the actual experience level in front of you.
brainstorming, creative work, and genuine discovery often benefit from some unstructured difficulty and ambiguity that efficient-learning design would otherwise strip away.
applying a formal load-reduction process to a one-line instruction or a task nobody finds difficult is overhead without a problem to solve.
This browser tool structures your thinking. It does not send your information anywhere or perform AI analysis.
Save as Skill converts the model into portable Markdown instructions that teach an AI when and how to apply Cognitive Load Theory, what questions to ask, what evidence to seek, what mistakes to avoid, and how to structure its analysis.
Save the skills you find useful and organize them into your own reusable library. The Markdown is intentionally model-neutral, so it can be used wherever an AI system accepts uploaded files, project knowledge, custom instructions, or reusable text context. Platform capabilities differ, so this is a portable skill file rather than a claim of one-click installation.
Sweller's original work was about instructional design, but the same working-memory limit governs anyone trying to absorb anything new — a customer reading a help article, a user following a setup wizard, or an AI system parsing a long system prompt. Scale the analysis to wherever someone has to hold new information in mind to act on it.
No — length and load aren't the same thing. A longer document broken into well-sequenced chunks with worked examples can carry less real load than a short, dense paragraph that crams several unrelated ideas together. The goal is matching structure to capacity, not minimizing word count.
Finding that the difficulty people are struggling with is actually intrinsic — a genuine property of the subject that no amount of restructuring removes — rather than extraneous friction created by poor presentation. Reformatting won't fix a problem that was never a formatting problem.