Learn it → Apply it → Save as Skill
Memory fades fastest right after learning, then the loss slows down. Without reinforcement, most newly learned material is already gone within a day or two — and the drop is steep, not gradual. Knowing the shape of that curve tells you when to step in before the loss happens, not after.
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.
When someone learns something new and never revisits it, most of the loss happens in the first day or two — not spread evenly across the following weeks. After that initial drop, what remains fades much more slowly.
Important: the exact percentages vary by person, material and difficulty. The shape — fast early loss, then a slow tail — is the reliable part, not any specific number.
The model is most useful whenever people are taught something once and are expected to still have it days, weeks or months later.
The forgetting curve describes what happens by default. The useful move is interrupting it on purpose, before the biggest drop.
Something is taught, read or practiced for the first time and encoded into memory.
Without any review, expect the sharpest loss in the first day or two, then a slower, longer tail.
Schedule a review just before the predicted drop, not long after it — then space later reviews further apart.
Ask people to recall or perform the material instead of passively re-reading or re-watching it.
Open a use case to see what the model changes.
A single long onboarding session feels complete on day one but tests poorly by day seven. Expect that, and build short refreshers into the first two weeks instead of relying on the original session to hold.
A once-a-year session decays long before the next one. Shorter, more frequent refreshers timed around the steepest part of the curve retain more than one long annual event, even at the same total training time.
A single sales-kickoff briefing fades fast under the pressure of calls and quotas. A short quiz or role-play at day two and day seven preserves far more than a slide deck emailed afterward.
Procedural steps taught in a single session should be re-tested, not assumed retained, especially before the skill is used on a real patient or case. Schedule a supervised re-check rather than relying on a certificate date.
Usage of a newly trained workflow often drops within a week of training, not because agents are careless but because the memory trace was never reinforced. A short check-in at day three catches this before it becomes a habit.
A single onboarding tour fades like any other unreinforced memory. A brief in-product nudge timed a few days later often restores usage that a longer initial tutorial could not.
This is an illustrative example designed to teach the method, not a universal statistic.
Expand for the origin of the curve and what later research confirmed or complicated.
Between 1880 and 1885, Ebbinghaus ran a self-experiment: he memorized lists of meaningless three-letter “nonsense syllables” (such as WID or ZOF) so that prior knowledge of real words would not help him, then tested his own recall after different time intervals and measured the “savings” — how much faster he could relearn the list compared to learning it fresh. He published the results in 1885 in “Memory: A Contribution to Experimental Psychology,” describing it himself as a limited, incomplete study on a single subject.
In 2015, researchers Murre and Dros attempted a direct replication using a single subject and nonsense-syllable lists similar to Ebbinghaus’s original method, and found an overall curve shape closely matching his data. Simple exponential formulas tend not to fit the data perfectly across the full time range, which is one reason the curve is best treated as a general pattern rather than a precise equation.
No. The rate of forgetting changes with the difficulty and meaningfulness of the material, how well it connects to things already known, physiological state such as sleep and stress, and how actively it is rehearsed. Emotionally significant memories and well-practiced procedural skills are widely discussed as following a different, typically slower, decay pattern than rote verbal material like nonsense syllables.
A single training session, video or briefing is not a finished job if the goal is retention past the first week.
Since the steepest loss happens earliest, the first review should come soon after learning, with later reviews spaced progressively further apart — not spread in even weekly intervals from day one.
Passive re-exposure, such as re-reading or re-watching, is weaker at building durable memory than active retrieval, such as a quiz, a practice run or teaching it to someone else.
Meaningful, well-connected material and well-rehearsed procedural skills decay far more slowly than isolated facts. The same refresh schedule will not fit both.
The original data came from one person memorizing nonsense syllables. Use the shape — fast early loss, slower tail — as a heuristic, not a lookup table for your own content.
Fatigue, competing new learning and poor sleep can all accelerate loss well beyond what the “expected” curve predicts.
the material is highly meaningful, emotionally significant or tightly connected to existing knowledge — these tend to decay far more slowly than the rapid-early-loss shape described here.
you are training a physical or procedural skill, such as a manual technique or operating equipment. Motor and procedural memory commonly follow a different, often slower, decay pattern than verbal recall.
what looks like “forgetting” is really a sign the original teaching was unclear or disorganized. No refresher schedule fixes a first explanation that never created a real memory trace.
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 the Forgetting Curve, what questions to ask, what evidence to seek, what mistakes to avoid and how to structure its recommendations.
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.
Ebbinghaus proposed a logarithmic-style equation for his own savings data, and later analyses have found that no single simple exponential formula fits forgetting data perfectly across all time ranges. The reliable, well-replicated part is the shape — fast initial loss followed by a long, slow tail — not a specific formula you should expect to apply exactly to new material.
Each additional review appears to slow the rate of subsequent forgetting, which is the basic logic behind spaced repetition systems: later reviews can be spaced further apart than earlier ones because the memory has become more durable, not because less reinforcement is suddenly needed.
Not necessarily. Some apparent forgetting reflects weaker retrieval access rather than full erasure, which is part of why relearning something is often faster than learning it the first time, and why a small retrieval cue can bring back material that seemed lost on an unprompted test.