Bear Hunter System

The Bear Hunter System is an encoding workflow in which the structure of a topic gets built while the topic is being learned, not assembled afterward. It replaces passive reading, linear notes, and every version of “I’ll organise it later” — collecting first and integrating later wastes the work, because nothing was integrated to begin with.

The name borrows the hunting sequence it runs on — sight the target before firing, fire, then process what you brought down — and the page walks those three passes: Aim, Shoot, Skin. It is the default encoding workflow of this vault, for any subject, and it sits inside Deep Processing, the dimension that accounts for what processing buys over exposure.

Aim — the questions come before the source

Aim is writing what the material must answer before opening it. The pass starts with keyword collection, and collection is deliberately cheap: most topics — a couple of textbook chapters, a few weeks of class — resolve to 15–30 major concepts, not pages of them, and catching about 80% of them in about 25 minutes beats catching all of them in two hours. Collect them out of order on purpose, to break the framing bias of the source’s own headings; invent your own keywords for concepts and diagrams that carry no label; for a wholly unfamiliar term, take a one-paragraph gloss and move on. The extraction itself can go to a tool — that is the easiest part, and the difficulty worth keeping is the thinking that follows. Then group what you have into three to seven hypothetical chunks, hypothetical being the operative word: committing early to a structure is the sunk-cost trap, and if in doubt, draft the flow earlier rather than later so the map can carry the load your head no longer has to — load the workflow first raises on purpose, because the higher-order work lives up there, and only then offloads.

Each concept then gets its questions, and there are two for a reason. Why is this important? forces the big-picture judgment. How does it relate to ___? surfaces the relationships that the importance question misses, because importance is answered through whatever bias you already hold. The questions that work are why- and how-shaped, genuinely unknown to you, and better still if they are emotional, personal, or genuinely curious; plain what is… questions can be dropped entirely, since higher-order answers complete the lower-order ones automatically. Cast the net shallow and wide rather than deep and narrow — questions pay on the content they target and almost nowhere else, so coverage of the topic, not depth on a corner, is the variable that decides whether Aim pays.

What importance looks like varies by subject, and the patterns are startable — never a substitute for asking why is this important? each time, which is where the real chunks come from:

Subject	Importance usually runs through
Biology	Function; structure against function; scale; groups of causes meeting groups of effects
Chemistry	Reaction mechanisms; applications; common reagents
Physics & maths	Real-world applications; the concepts a concept depends on
English	The author’s background, motivations, and values; the significance of the story
History	The movements of the period; cause and effect; the motivations of key players
Law	The principles of justice underneath; the landmark cases that exposed their gaps
Languages	Common usage; situational usage; the effect a form has on the message

Aim goes as early as possible — before the lecture, the class, the reading, the video — and it scales down honestly: five or ten minutes buys a superficial Aim of the largest chunks and major relationships, and even that pays. What it replaces is the hour of past papers: an hour answering questions you wrote is an hour spent entirely on your own weak points, because you only wrote questions you could not already answer, where an hour of ready-made questions surfaces one or two isolated gaps and trains none of the skill of generating curiosity in dull material — and connected questions revise several points at once, which isolated ones never do. Past papers are a thinning bet besides: exam styles shift and curveball questions grow more common, so the ready-made hour trains last year’s exam. Going slightly out of scope is faster, not wasteful — five or ten minutes on something non-examinable that unlocks the examinable content pays for itself in saved relearning. And the metric to distrust is pages covered: studying is not learning, and Aim is not planning for learning — it is the first part of the learning. The step’s own page is Aim, and the grouping discipline it feeds is Importance-Based Chunking.

Shoot — answers go onto the map

Shoot is working the sources to answer your own questions, each answer landing on a growing map rather than in the order the source was written. The answers worth keeping are logical and interconnected — logical meaning you know where the answer comes from, so it needs no memorising; interconnected meaning it joins the nodes already there, and the count of connections is what tells the brain to keep it. The calibration for what that feels like: a violinist knows the bow’s length without ever memorising it, and a footballer knows every teammate’s position in any situation — detailed knowledge nobody sat down to memorise, because it was learned organically and it made sense. Most learning should look like that.

A Shoot session looks chaotic from outside: read, flip back and forth, write a few notes, connect them, find the next question, return to the resource — the first chunks spaced far apart on the page, because sub-chunks will need somewhere to land. From inside it is the opposite — following someone else’s prescribed order is the true chaos, because the brain is solving meaning, structure, and relevance all at once against a sequence it did not choose. The knowledge itself was never linear: it got flattened into a line to be delivered, and the map rebuilds it toward the shape it had in its author’s head. Nor does encoding owe the exam its format — encode in the shape that learns best, then practise retrieving in the shape the exam will ask. The material lands in layers, in order: the backbone and its logic first, then the concepts hung off it, then the concrete details that anchor them — and the fourth layer, arbitrary details, does not belong on the map at all; those go to flashcards later. Not every question gets answered perfectly, and that is fine — the learning occurs in the attempt. The loop also iterates: Aim again as the map raises new questions, Shoot again to answer them, sometimes several cycles before a map is ready to Skin. One diagnostic runs the whole pass: if you aimed well, shooting is easy; difficulty in Shoot usually means old linear habits are steering again. The mechanics live on Shoot.

Skin — the map becomes the artifact

Skin rebuilds the working map into a structure simple enough to redraw with every source closed. It is not tidying: every chunk, sub-chunk, and re-chunk decision is a judgment about what matters most and why, which is the highest order of the work — Skin is where much of the learning happens. The finished form has a name — GRINDE — and a six-part spec. Grouped: heavily chunked. Reflective: the layers visibly sorted, backbone to concepts to important details, so the map shows how you actually worked it out — the marker of an advanced learner. Interconnected: lateral links inside and across layers — real cause-and-effect only, never decorative arrows, each labelled with what the relationship is, the rationale written rather than a bare line drawn. Non-verbal: as few words as possible, doodles and a personal shorthand doing the carrying, both because words hide the structure and because abstract visuals recruit more of the brain. Directional: a visible flow, which is the test of whether the logic is real, since logic is cause and effect. Emphasised: the main line made visually dominant — by size, weight, or colour — which forces an explicit judgment about which connections matter most.

The sizing rule is the 2–4 Rule: two branches off a node is the target, four the ceiling, and fewer than two the signal to merge. Worked once, with neutral content:

A chunk carrying six points splits into two sub-chunks of three — and the split itself
creates a lateral relationship between them that would otherwise have surfaced later as
crossing arrows. A node with one line in and one line out — a single-node chain — gets
removed, its neighbours joined directly, the upstream chunk relabelled so the surviving
relationship reads straight. Later, when siblings appear, the removed idea returns as a
sub-chunk of its own.

Skin runs the same day when possible, or as one weekly pass over the week’s material — what matters more is a rhythm you can sustain. The checks at the end are short: can the topic be explained from the map with the sources closed; do the chunks reflect importance or the source’s order; does any node carry more than four branches; do any single-node chains remain; does every arrow earn its place; did the pass make the map simpler. A map that is beautiful but hard to retrieve is not done. And one expectation is directly falsifiable: material that would have produced around fifty flashcards should produce ten to twenty after a correct cycle, because most of it now follows from the logic and concepts on the map. The refinement passes live on Skin.

Why the sequence is the sequence

A topic separates into four layers — the layers of learning: the logic that holds it together — cause and effect, why it happens and why it matters; the concepts — how it happens and what it is; the important details, which make a concept concrete; and the arbitrary details, learned only because an assessment demands them. The sort between the last two is one question: does knowing this detail help me understand the concept it is attached to? Roughly 30–50% of details pass it. Studied front-to-back with no structure, a subject runs about 10% conceptual, 20% important detail, and 70% arbitrary — seven-tenths of the time spent memorising things with nothing to hang on. Learned in layers, innermost first, the split flips to roughly 30% on logic and concepts and the rest on details that now carry meaning. And the layers are not static — that is the lever. How relevant a detail is depends on whether structure exists to give it relevance, so the trick to shrinking the arbitrary layer is overinvesting in the first two: a tree with ten thousand leaves has five or six main branches. Out of order, the same content grows more complicated and more disconnected and collapses back into raw memorisation. What the levels of mastery mean — and why “required mastery” is computable at all — is Knowledge Mastery: From Recognition to Usable Knowledge‘s subject.

The passes climb a ladder. Aim forces the brain to levels three through five — application, analysis, evaluation — from the start, by making problems before the content is understood; Shoot cycles applying and analysing; Skin lifts the work to evaluating, the level most learners never reach, while level three is the efficiency floor a working session never drops under. The engine underneath is counterintuitive and it is the reason questions come before reading: trying to memorise is hard; trying to understand makes you memorise; trying to apply makes you understand more deeply — jump to application first, and the lower levels fill in underneath faster than they would if addressed directly. Retention itself works the same way. The brain invests in what carries traffic: a station is large because connections run through it, never the reverse — so to make any single fact memorable, stop working on the fact and work on its relationships; the remembering arrives as a byproduct. What you think equals what you learn — not what you write.

Which is why the technique is supposed to feel a particular way, and the feelings have meanings:

How it feels	What it means
Overwhelming	The number of relationships surfacing while answering “why is this important” — a network of possibilities to refine
Confusing	Many viable non-linear structures exist, and deciding between them is prioritisation and evaluation — the highest order of the work
Time-consuming	Each keyword is getting deep examination; the depth is the consolidation, and the feeling fades as speed grows

When it is working, the feeling is as specific as the table: the topic seems to organise itself while being learned — the map making your thinking more structured, not merely recording it. The inverse also holds: finding it easy is the alarm. A learner who feels no difficulty is usually not applying the technique to standard, and study so easy it makes you drowsy is a clear signal nothing is happening — learning is a highly engaged state. The technique also withholds certainty for a long time; the certainty that traditional methods hand out early is an illusion that does not reflect the knowledge actually built, which is where inconsistent results, frustration, and performance anxiety come from. Reading that effort correctly — as signal, not cost — is Cognitive Load & What Mental Effort Is Trying to Cue‘s subject, and steering by it is Self-Regulation‘s.

Running it

Prestudy is the timing decision that pairs with the workflow: studying a topic before the class that formally covers it, so the live session becomes a second pass rather than a first — a decision about when learning starts, where the Bear Hunter System is how. The mechanism is load: it spreads the cognitive cost across sessions instead of concentrating it in one intensive event, where overload is hard to recover from and compounds week over week. Done correctly it ends with a clear big picture across the whole topic, the major ideas seen in relation, some hypotheses about how they connect, and no details memorised yet. Its ceiling is real — a learner with moderately efficient technique can frame four to six weeks of content in a single weekend — and so is its floor: a ten-minute skim before class still pays. The full treatment is Prestudy.

The schedules come in three tiers, with the system’s own grades. Standard — efficiency 5/10, difficulty low: Aim at least one week’s material at the end of the previous week; Shoot before, during, and after each learning event; Skin daily and re-Skin the week, or Skin the week at once — whichever rhythm survives contact with your actual schedule. Accelerated — 7/10, moderate: one weekend Aims an entire topic and Shoots 60–70% of it from self-study material; the remaining 30–40% happens before, during, and after the week’s events; classes stop being first exposure and become recheck, re-Skin, and easy retrieval sessions; Skin closes the week — and a different subject each weekend keeps you weeks ahead of the curriculum. Ultra-accelerated — 10/10, high: Aim, Shoot, and Skin all before the main events, entire topics in a session, curriculum time freed for retrieval, targeting roughly 90% coverage at 90% retention — explicitly not 100% — at about a day per topic; by the system’s own admission, essentially impossible without substantial cognitive training. Under all three tiers, weekly and monthly interleaved retrieval keeps running underneath — the tiers change encoding, never the retrieval floor. When time is genuinely tight, the minimum viable version is a deliberate reduction: the five to ten most important concepts — a cut-down of the usual 15–30 — the why-question on each, three to seven chunks, only the relationships needed to explain the topic, and one cleanup item after learning. What it may never lose, at any speed: relationships stay first, no collapse into pure memorisation, the layers stay ordered, and the order stays yours — order control, the skill the whole workflow trains.

Speed itself is a skill with a name — Hipshot, shooting from the hip rather than sighting down the barrel. It runs Aim and Shoot as one fast alternating pass — keywords filtered to the most important, the questions asked mentally instead of written, answers going straight onto the map. It drastically cuts Aim time and the iterations before a map can be Skinned, and it carries a hard prerequisite: it is only possible near unconscious competence with the standard cycle, it often arrives on its own as a byproduct of that competence, and the retreat rule is explicit — struggle with it, and you return to the full loop. The beginner’s version fails in a known shape: stuck answering questions, no longer relating them, the map decaying into a one-directional information dump.

A model fits into this at specific joints. Hand it the tedious low-effort work — keyword extraction; consolidating several sources into one, which prevents split attention; generating alternate Aim questions after your own attempt, never before it; asking what relationships the map might be missing; generating test questions once the map exists — and challenge everything it returns, since it has no inherent grip on truth however it reads. Keep it away from the thinking the workflow exists to force: chunking and organising are the higher-order learning, so outsourcing them removes the learning, and the question-writing skill behind Aim is never outsourced either, and a beginner cannot yet feel the passivity from inside — which is why the honest advice is not to use a model for structuring at all until the difference between active and passive is a felt thing. The decay this guards against is The Shortcut Problem‘s subject.

When it goes wrong

Fault	Cause	Repair
Passive Aim	Questions are mostly what is…	Rewrite as why/how/relates questions; drop the what-questions entirely
Jeopardy question	A question rebuilt from an answer already held — it looks like inquiry and generates none	Ask only what is genuinely unknown to you
Importance checklisting	”Why does it matter” answered with one reason, then moving on	Most things matter for several reasons; produce them and weigh them
Copying during Shoot	Notes following the source’s order	Back to the Aim questions; answers land under chunks
Reverse causality	Importance justified by what a thing produces rather than what produced it — the relationship now rests on rote	State “A is important because B,” then ask: would I only know that by having memorised it?
Waterfalling	Every connection radiating outward or down; zero lateral links	Ask how the chunks affect each other; the inter-chunk relationships are the backbone
Spiderwebbing	Chunks formed first, relationships drawn in afterward	Regroup around the real relationships — never just redraw arrows
Wheel-and-spokes	Sound chunks, every link radiating from one hub	Connect the chunks to each other, not only to the centre
Overloaded chunk	More than four branches off one node	Split into sub-chunks
Single-node chain	One in, one out	Merge it; a terminal node merges into its parent
Pretty but unretrievable	The map satisfies the eye and not the memory	Brain-dump it closed-book and repair what would not come back
Skipped Skin	Rough notes never become an artifact	Schedule the pass; it can itself serve as a spaced session

The prevention for the three structural faults is one procedure: textbook hierarchy → flat list → relationships identified → chunks formed from those relationships → map. The list step exists to strip the source’s grouping bias before any chunking happens. Reverse causality, worked once: mRNA is important because it creates protein — an arrow pointing at what mRNA produces. That relationship rests on rote, because the map gives no reason mRNA rather than anything else does the creating; it adds an item to memorise instead of removing one. Flip the question — what produces mRNA, what problem the cell solves by making it — and the arrow carries logic that regenerates the protein fact for free. Intuitive chunking prevents the fault; importance checklisting and plain question-and-answer drilling breed it. And when retrieval later fails, the failure carries its diagnosis: no approach to a problem means a gap in how concepts relate, common where procedural drilling crowded out the conceptual network; a high volume forgotten is almost always a chunk-structure fault — check the map for waterfalling and spiderwebbing, and whether the chunks are intuitive — never a plain effort fault; recurring execution errors mean too little practice, sometimes downstream of a conceptual misunderstanding; failure only in new contexts means missing relationships. Difficulty matters more than correctness — struggling marks a gap even when the answer came out right — and an answer sheet checked before producing your own manufactures the illusion of knowledge. The general form of every row above is one page: The Technique Is Only as Good as the Thinking It Produces.

What you are left holding

A structure simple enough to redraw with the sources closed, that answers questions the sources never asked — knowledge left easier to reconstruct, use, adapt, and improve, built during the learning, which was the whole claim. From there the work changes hands, and the artifacts convert directly: Aim questions become the retrieval prompts, the map their answer key, the conversion run after each Skin — so Spaced Interleaved Retrieval tests whether the encoding held and keeps it reachable long-term, with failures reading back as structural diagnosis rather than as a demand for more hours; the full pipeline runs on Prestudy, BHS, and SIR: Turning Information into Usable Structure. The honest limits ride along. The cost is front-loaded and this is where quitters quit — slow at first, then compounding as each new piece meets structure already built, the loss made back even in early competence when the whole arc is measured; the load never falls, you get better at carrying it, until passive study becomes the thing that feels intolerable. Mapping’s measured benefit outside this vault is a range, small to large depending on use and comparison, not a point. And the system names its own hardest terrain: material that is not inherently logical — languages above all — resists importance-based chunking, and the importance question does less work there. Everything else on this page is the during-not-after claim, kept.

Open questions

• Where does the boundary actually fall between an important and an arbitrary detail in a given curriculum?
• How much of the importance question survives on material that is not inherently logical?

Sources

• Prequestion meta-analyses (2023, 2025 — Psychonomic Bulletin & Review; Educational Psychology Review): questions asked before study lift performance on the content they target (g ≈ .66) and almost nothing else (g ≈ .01) — the support for Aim, and the constraint that makes coverage the variable.
• Dunlosky et al. (2013), PSPI — elaborative interrogation at moderate utility; Donoghue & Hattie (2021), Frontiers in Education — ten-technique synthesis, 242 studies.
• Cowan (2001), Behavioral and Brain Sciences — the working-memory capacity of about four chunks; the honest grounding for the branch ceiling as a heuristic, not a mapping law.
• Nesbit & Adesope (2006), Review of Educational Research — concept and knowledge maps, 55 studies: benefits small to large depending on use and comparison.
• Hattan, Alexander & Lupo (2024), RER; Simonsmeier et al. (2022), Educational Psychologist — prior-knowledge activation is real and conditional; prior knowledge correlates with knowledge gain at a mean near zero.
• The workflow itself — the three passes, the layer model, the schedules, and every operating default on this page (concept counts, chunk counts, branch limits, tier grades, coverage shares) — is the learning system this wiki organises, stated as its own operating doctrine throughout.