The irreducible epistemic atoms underlying the curriculum. 4,828 atoms across 8 types and 2 molecules
Lay out multiple notes simultaneously in parallel visual access rather than reading them sequentially, because synthesis requires holding multiple ideas in working memory or visual space at the same time.
During synthesis, look for structural parallels and shared dynamics across inputs rather than topical overlaps, because topical overlaps produce aggregation while structural parallels produce synthesis.
Test synthesis output by asking whether it makes a claim that could not be found in any individual source - if it could be produced by listing inputs, you have aggregated rather than synthesized.
When you cannot explain a concept in simple terms to someone with zero shared context, treat this as diagnostic evidence that your processing is incomplete rather than a labeling problem.
When information backlog in any queue exceeds your daily processing rate by more than 10%, declare that queue a bankruptcy candidate requiring immediate archive-and-reset rather than incremental catch-up.
Start new information processing habits with a minimal version (processing 5 items, 2-3 minutes) for five consecutive days before expanding to full sessions, because initiation barriers determine habit formation more than execution quality.
Operate information processing on four nested cadences: daily triage (20-30 min), weekly review (catch drift), monthly audit (source quality), and quarterly purge (accumulated sediment).
Choose one information processing tool by defining 5 minimum requirements, selecting the first that meets them, and committing for 90 days before re-evaluation, rather than maximizing across all available options.
Track daily whether you completed your information processing session (yes/no) rather than optimizing configuration, because consistent execution of a mediocre tool outperforms sporadic use of a perfect one.
Measure information pipeline health by counting processing days in the last 30 (target: 28+), not by note count, graph density, or tool sophistication, because daily execution is the only variable that compounds.
When information pipeline maintenance consumes more cognitive resources than the pipeline saves, the system economics have inverted and simplification is required, not optimization.
Annotate each output type in your taxonomy with its approximate half-life (hours, weeks, months, years) to calibrate appropriate investment in durability, structure, and polish.
For each output type, define both 'good enough' criteria and explicit over-investment thresholds to prevent perfectionism from consuming effort that belongs to higher-stakes work.
Position quality verification between 'I think this is done' and delivery as a mandatory pause point with an explicit checklist, not as optional review when time permits.
Limit output checklists to five-to-nine items targeting errors that are both frequent and consequential, ordering from most catastrophic to least to enable partial execution under time pressure.
Build checklists from your documented error history rather than hypothetical failures, promoting an error to checklist status only after it has occurred in actual deliverables.
Embed checklists at the physical or digital location where delivery occurs, making them harder to skip than to use by integrating them into templates, pull request forms, or pre-send screens.
Schedule creation and editing as two separate calendar blocks with a minimum gap of thirty minutes (preferably overnight) to allow defamiliarization and enable fresh-eyes error detection.
For outputs requiring high polish, structure editing as three sequential passes—structural edit (organization), line edit (sentence clarity), copy edit (mechanics)—to prevent simultaneous optimization across incompatible dimensions.
Build default two-pass policies by output type—emails under five sentences single-pass, documents over one page mandatory two-pass, strategic outputs three-pass with overnight incubation—to prevent case-by-case deliberation.
Route attention and effort disproportionately to outputs with long half-lives (months-to-years) and high audience reach over outputs with short half-lives (hours-to-days) and narrow reach when both compete for finite production time.
Before consuming any information during learning sessions, define the specific output that consumption will feed into—if no output can be named, the learning session does not belong on the calendar.
For recurring operational outputs (status reports, bug reports, meeting minutes), design templates with nearly complete structure including section headings, formatting, and boilerplate text, requiring only variable content insertion.
Extract templates from your three best past examples of an output type by identifying common structure across those instances, rather than designing templates from theory.