The irreducible epistemic atoms underlying the curriculum. 4,828 atoms across 8 types and 2 molecules
Design knowledge systems with standardized interfaces between atomic units to enable recombination without requiring knowledge of internal implementation.
When a concept resists precise naming, treat that resistance as a diagnostic signal revealing incomplete understanding rather than a labeling problem.
Name concepts using complete declarative statements that make falsifiable claims rather than topic labels to enable direct composition and evaluation.
Design note titles to anticipate retrieval queries by matching the natural language your future self would use when searching for that concept.
Establish domain-specific ubiquitous language where each concept maps to exactly one term and each term maps to exactly one concept to eliminate translation overhead.
Store claims and their supporting evidence as separate addressable nodes linked explicitly rather than fused into single annotated statements.
Force explicit link creation between separated claim and evidence nodes to create deliberate evaluation moments that interrupt automatic confirmation bias.
Maintain provenance metadata for evidence nodes specifying methodology, sample characteristics, and study limitations to enable independent quality assessment.
Label observations and interpretations explicitly with distinct markers to make the cognitive boundary between perception and inference visible and auditable.
Apply the camera test to distinguish observations from interpretations by asking whether a recording device could capture the claimed phenomenon without human judgment.
Structure notes to preserve multiple competing interpretations of the same observation rather than collapsing to a single narrative to maintain interpretive flexibility.
Structure captured notes to be interpretable without access to their original source context by embedding provenance, purpose, and relational metadata within each atomic unit.
Select granularity based on the questions your system needs to answer rather than seeking an inherent correct level of detail.
Calibrate information granularity to match working memory capacity and task complexity — too fine produces fragmentation, too coarse produces overload.
Treat questions as first-class knowledge atoms that persist and evolve rather than temporary gaps to be filled and discarded.
Maintain open questions as persistent search filters that automatically detect relevant information across future encounters without conscious effort.
Design tasks to activate curiosity by framing them as investigations with specific knowledge gaps rather than obligations to complete.
Make operational definitions explicit before reasoning from them — trace conclusions back to the specific meanings that bear their weight.
When disagreement persists despite shared facts, trace it to conflicting definitions rather than continuing to debate evidence or conclusions.
Maintain bounded contexts where terms have precise local definitions, with explicit translation layers between contexts where the same word means different things.
When concepts recur across three or more contexts with similar structure, extract the shared pattern into a named abstraction that each instance references.
Defer abstraction until you can identify what varies versus what remains invariant across instances — premature abstraction produces vague unusable generalizations.
Track the evolution of your beliefs over time rather than overwriting previous positions, because the trajectory of revision itself contains knowledge that the current state alone cannot provide.
Design capture and retrieval systems to minimize friction between having a thought and externalizing it, because even small increases in effort create selection bias toward capturing only high-activation thoughts.