Concepts
The irreducible epistemic atoms underlying the curriculum. 4,828 atoms across 8 types
When content appears in 3+ contexts, extract a single…
When a contained item appears in three or more contexts requiring synchronization, extract it into a referenced shared source with a single canonical version rather than maintaining multiple copies.
Put critical instructions in the first 20% of system…
Position your system prompt's most important instructions in the first 20% of tokens, because transformer attention mechanisms allocate disproportionate processing to early sequence positions through positional encoding and causal masking.
Limit knowledge graph relationship types to 5-7 — more…
When building knowledge graphs, limit relationship type taxonomies to 5-7 types rather than attempting comprehensive ontological coverage, because classification overhead beyond this threshold produces diminishing informational returns while increasing maintenance cost.
Typed edges enable AI reasoning; embeddings only find…
When AI systems traverse your knowledge graph, maintain typed relationship labels with explicit predicates rather than relying on semantic similarity alone, because typed edges enable logical reasoning while embeddings only surface associative proximity.
Distinguish real hubs (genuine cross-domain centrality)…
Distinguish real hubs (concepts earning centrality through genuine cross-domain relationships) from artificial hubs (index pages linking to everything in a category) by testing whether removing the hub would sever meaningful conceptual pathways or merely convenience pathways.
Use Shneiderman's overview-first protocol for graph…
Apply Shneiderman's three-phase visualization protocol to graph exploration: overview first without reading labels, zoom and filter to examine specific clusters, then details-on-demand for individual nodes.
Separate reference material (static lookup) from working…
Separate reference material (static information for lookup) from working notes (evolving thinking) into distinct systems, as mixing them corrupts both retrieval and development.