Problem-First Learning

Churning through practice problems without a conceptual map pays back exactly one question type per question practiced. Formula-heavy subjects invite the trap: skim a sliver of theory, jump straight to the equations, start grinding — and because applying early is genuinely good for learning, the grind feels productive. But the loop never cycles back to the knowledge. Each attempt is trial and error against a memorised procedure, success looks random, and the applied, in-context questions that decide outcomes stay out of reach — they were never part of the practiced pattern. Flipping the order fixes the ratio: establish the problem the topic exists to solve, build the big picture of every way it applies, and only then learn the equations. Practice volume stays roughly fixed, but it starts to transfer — ten questions can unlock a hundred question types, a hundred can cover ten thousand — because each question now tests an understanding rather than adding one more routine.

The order is the whole cognitive difference: learn in order to solve a problem already in hand. Meeting applications after the theory, as worked examples, files them as more material to memorise — every new application adds to the pile. Meeting the problem first makes each equation arrive as the tool for something you already wanted to do.

Building the map before touching equations

• Inventory the keywords. Pull the topic’s main concepts, equation types, and keywords from summary notes, encyclopedia entries, search, and the curriculum, laid out as one list. The urge to dive into the equations and the overwhelm at the list’s size are both normal; the pass starts slow and accelerates, because early answers pre-cover the relationships hiding in later items.
• Interrogate every keyword in a fixed order. What is it, why is it important, how does it relate to the rest — asked before any equation is opened. The second and third questions carry the weight in formula subjects: the why supplies relevance, and the relationships are what the map is made of.
• Hold the question while searching. Skip any resource that answers slowly or in excessive detail — lingering on a mismatched source is pure waste. Once answered, stop and move on; fine detail gets its turn later.

The three-to-five-minute cycle

• Alternate aim and shoot in tight loops. Keep the keyword list visible and return to it every 3–5 minutes: take whatever was just learned and ask how it relates to everything on the list — same, different, cause, consequence. The list works as a standing set of aims; the reading is the shooting.
• Map relationships as simply as the page allows. Represent each answered question in the simplest structure that captures the relationship, adding detail only as it earns a place. The artifact is a logical framework of the topic, never a transcript.
• Stop when purpose is locked and loaded. The endpoint is knowing, for every equation not yet learned, exactly what it will be for once it is. About ninety minutes built that state on a topic relearned from scratch — confident in every application while still unable to run a single calculation. From there the equations install fast, and practice turns into hypothesis testing: predict the approach, check, and let each wrong answer correct the model rather than just that one question.

Authoring the hardest questions

• Write your own questions, maximally hard and always applied. Once answering feels comfortable, compose questions wrapped in realistic scenarios that demand several concepts in combination — intimidating enough to look scholarship-grade.
• Read the failure correctly. Answering found questions while being unable to invent hard ones flags, with high probability, a conceptual gap: combining and comparing concepts sits a level above applying any single one, and that level only exists if the map does.
• Repair at the map. The time-efficient fix is returning to the keyword map, rebuilding how everything fits together and what each piece is for, then attempting to author again. Grinding more practice problems leaves the gap untouched.
• The pair is protective. Practicing found questions plus authoring your own covers the whole space of what an exam can ask; surveyed across multiple year levels and curricula, no question fell outside it. Trading authored challenges with a partner tests both maps at once.

Where the procedure migrates to AI

The same split now governs coding and agentic skill acquisition, with the procedural half increasingly absorbed by AI. An agent can hold the syntax, API details, and step-by-step execution — the equivalent of the equations — leaving the human exactly what this page builds: the map of what each tool is, why it exists, and how the pieces relate, with purpose locked before implementation starts. The diagnostic transfers intact. Someone who can operate the tools but cannot author a maximally hard, realistic problem for the system being built is missing the map; the repair is the same — rebuild the conceptual layer before running more reps through the tool.

Links into the system

Specialises the Bear Hunter System aim–shoot–skin cycle for formula-heavy subjects; the problem-before-information order comes from Cave Theory, its relationship map is what Survive and Thrive keeps in memory, and the big-picture-before-details sequence parallels Layers of Learning. The procedural/declarative split is defined in Declarative, Procedural, and Conditional Knowledge, and the AI-absorbs-the-procedure angle feeds Agentic Engineering.