Encoding sets the ceiling — the structure built at first contact decides what exists to remember — and retrieval is a builder, not an audit: each recall strengthens, reorganizes, and stress-tests what encoding made. The enemy is passive review: re-reading and recognition feel like learning and aren’t. So commit to answers before seeing them, build relationships rather than collections, work concepts before labels, and let daily use supply retrieval reps. When recall keeps failing, suspect the encoding before the schedule. Around the loop sits regulation — felt difficulty as the gauge, calibration over scores — and a session machine that spends no willpower on choices.

1. Encoding

• Commit before content. A real attempt before the answer forces connection-seeking and exposes the precise gap; establishing the problem first flips practice transfer from roughly 1:1 toward 10:1 (Problem-First Learning, Prestudy).
• Relationships, not collections. Encoding is grouping, connecting, and ordering — schema construction, not fact accumulation (Schema Construction, Importance-Based Chunking).
• The working loop is Aim → Shoot → Skin: questions first, answers hunted through sources, the map cleaned into retrievable structure (Bear Hunter System; faster variant Hipshot).
• Concept first, label last. Meet the behavior and importance of a thing before its name, or the name becomes a hook with nothing on it (Reverse Explanation).
• Generation over reception. Representations you construct are the encoding; visible activity that bypasses the thinking is the trap (Thinking on Paper, The Shortcut Problem).
• Get the first pass right. One error costs ~4x the time of doing the step correctly; speed comes from changing the process, never from rushing (Accuracy Before Speed).
• Meaning builds encoding; attention gates it. No technique survives divided attention (Rules of Effective Memorization).

2. Retrieval and spacing

• Recall, not recognition. Recognition feels like knowledge and isn’t; re-reading’s fluency is the illusion the whole trap is built from (Retrieval, Spaced Interleaved Retrieval).
• Space so forgetting helps. Widening intervals exploit partial forgetting to reveal true gaps; a session with zero failed recalls was confirmation, not testing (SIR, Retrieval).
• Rotate the angle. Knowledge practiced in one direction is brittle; vary the retrieval form and the similarity window across sessions (Interleaving: Multiple Angles).
• Reconstruction sits above recall: form and judge new relationships among known nodes — curveball immunity lives there, not in replayed answers (Reconstruction).
• Teach it back. Explaining whole → parts → whole, off pure memory, tests integration at every level at once (WPW).
• Repair gaps, don’t re-rep them. When an item keeps failing, suspect the encoding: rebuild the structure or the connection rather than rescheduling the same rep (SIR gap repair, Retrieval).

3. Application

• Daily use supplies retrieval reps. Retro-label each real-world rep, patch what real use missed, and schedule study around upcoming work instead of running parallel review (Opportunistic Retrieval).
• For languages: meet material cold at the comprehensible edge. No pre-memorized lists — reconstruct meaning from context, then learn exactly the missing pieces; words learned to solve a live problem outlast memorized ones (Aim-Shoot-Skin for Language Learning).
• For scripts: stagger the attributes. Sound and meaning first, form attached sessions later, overlapping small cohorts; recall routes run through component cross-matching (Pipelining, IME Method).
• Hunt shared structure. Components recombine; periodic passes over the whole collection find overlaps that per-item review never surfaces (Sound Series, How Chinese Characters Work).

4. Regulation and calibration

• Felt uncertainty is the gauge. None means comfort-zone drift; the reversion response — retreating to offloading when it spikes — is the signature failure (The Learning Zone).
• Train the radar for the shift from active processing into passive consumption; it happens silently and often (Building the Radar).
• Calibrate confidence, not scores. The two-question Mt-Stupid test plus a conditions check; judge multi-week trends, never one session’s results (Confidence Calibration).
• The technique is only as good as the thinking it produces — a correct-looking artifact with the wrong cognition behind it is the failure wearing a costume (The Technique Is Only as Good).

5. The machine around the learning

• Zero in-session choices. Repeated what-next decisions bleed time and focus; reduce vague choices to closed questions before the block starts (Decisional Delays, Choice Throttling).
• Rest ≈ work ÷ 3, taken at the first dullness — not when the block collapses (OFF-Rest Timing).
• Protect one to three priorities absolutely; the plan must survive a bad day, which means slack and concrete next actions, not optimism (Priority 0+1, Building a Schedule That Survives).
• Train the return, not the streak. Drift inside a work block is normal; what compounds is how fast attention comes back (Focus Management).
• Small bursts are real sessions. Prime–map–explore–dive–consolidate works in 5–30 minute fragments, with one consolidated retrieval mission to repair the seams (Microlearning System).

6. The filter under everything

• Mindset decides what difficulty means — fuel or evidence against you — and it trains behaviorally, through smaller cheaper mistakes, never through affirmations (Mindset, Fixed vs Growth Mindset).
• Spend attention only on controllable process; outcomes are symptoms (Locus of Control).

Omitted deliberately: decision-making doctrine (lives under Decision Making), environment and ownership (under Minimalism as Systems Design), agentic engineering (its own hub), and the standing stance toward doing intended things at all — motivation architecture, defusion, mistake relationship — which lives in Mindset, Condensed; the boundary is stance there, in-session steering here, and this page’s §4 and §6 are that stance applied at session scale. Every line above traces to a published page — a doctrine without an owner page is a gap, and three are flagged for the maintenance loop: recall-test hygiene (cueing, verdicts), session-scope rules, and the strong form of the lapse doctrine (when a failed item indicts encoding versus scheduling) have no dedicated owners yet.