6.7 KiB
Notebook Implementation Plan
This note turns the Notebook operating model into a concrete Didactopus implementation sequence.
It assumes the conclusions in notebook-operating-model.md:
- the Notebook is the durable knowledge layer
- learner products should derive from it
- distinctions, qualifications, constraints, and source roles are first-class
- public rendering must stay source-grounded and paraphrastic by default
Goal
Make Didactopus operate as a system with:
- source-grounded ingestion and review
- a durable Notebook concept network
- learner, workbench, and public products derived from that network
The key implementation mistake to avoid is treating the Notebook as a static page generator. It should instead be the stable intermediate representation that other Didactopus products depend on.
Phase 1. Strengthen the source-grounded substrate
Primary concern:
- preserve the information needed to build a useful Notebook later
Required additions at this layer:
- source role classification:
- overview
- mechanism
- nuance
- controversy
- argumentation
- distinction candidates:
A vs BA does not imply BB can occur without A
- learner-significance cues:
- why this distinction matters
- what misconception it prevents
- what explanatory work it does
- first-class secondary products:
- definitions
- qualifications
- constraints
- quote candidates
Expected implementation targets:
GroundRecallreview/export payloadsCiteGeistbibliography support and claim-support outputs- source-adapter metadata from
docliftand related import flows
Completion indicators:
- concept review payloads expose source roles
- distinction-like claims are tagged explicitly
- secondary products are inspectable without custom post-processing
Phase 2. Build Notebook-native concept structure
Primary concern:
- organize knowledge around explanatory hubs rather than narrow labels
Required additions:
- explicit hub-concept representation
- first-ring and second-ring neighborhood representation
- support for concept aliases without collapsing meaningful distinctions
- preferred-source selection by source role
- Notebook-level summaries built from grounded claims plus secondary products
Expected Didactopus targets:
notebook_pagepayload shape- hub/neighborhood builder logic
- pack emission for Notebook-facing artifacts
Completion indicators:
- a Notebook page can identify:
- the primary hub
- first-ring neighbors
- key distinctions
- preferred overview/mechanism/nuance sources
- bibliography topics no longer have to serve as the primary Notebook center
Phase 3. Make distinctions learner-facing
Primary concern:
- learning works better when concepts are contrasted, qualified, and scoped
Required derived-product features:
- distinction panels in learner workbench views
- definitions, constraints, and qualifications surfaced beside explanations
- “why this matters” cues for important conceptual differences
- quote/source-trail views for argumentation workflows
Expected Didactopus targets:
- learner workbench UI and backend payloads
- mentor/practice/evaluator session grounding
- lesson and activity generation
Completion indicators:
- learner-facing explanations can say not only what something is, but also:
- what it is not
- what it does not imply
- what nearby concepts it is often confused with
- practice prompts can target misconceptions and contrastive understanding
Phase 4. Separate rendering contracts
Primary concern:
- the same Notebook knowledge should support different output modes without blurring their rules
Three rendering contracts should be explicit:
Notebook contract
- preserve concept structure
- preserve source trails
- preserve review context
- preserve distinctions and caveats
Workbench contract
- surface definitions, constraints, qualifications, and quote candidates
- prefer inspectability over polish
- retain enough detail for argumentation and source checking
Public exposition contract
- prefer paraphrase over copied wording
- mark all quotations explicitly
- attach source citation in display
- never present unmarked source wording as original Didactopus prose
Expected Didactopus targets:
- notebook-page rendering
- workbench payloads and views
- public publication/export paths
Completion indicators:
- public pages can be audited for quote marking and citation display
- workbench pages expose more raw source-oriented structure than public pages
Phase 5. Source-role-aware retrieval and ranking
Primary concern:
- different tasks need different kinds of support
Retrieval should be able to prefer:
- overview sources for first-pass orientation
- mechanism sources for explanatory detail
- nuance sources for qualifications and constraints
- controversy sources for dispute framing
- argumentation sources for rebuttal and debate workflows
This should influence:
- Notebook page support lists
- learner-session grounding
- workbench recommendation ordering
- quote-candidate selection
Completion indicators:
- the same concept can yield different ranked source sets for
learn,review,argue, andpublishcontexts
Phase 6. Pack and publication alignment
Primary concern:
- the Notebook must become a stable export surface for other Didactopus flows
Needed outputs:
- Notebook-aware pack format additions
- workbench-friendly secondary-product exports
- publication-safe Notebook/public page export rules
Expected Didactopus targets:
- pack emission
- backend API payloads
- public export and frontend consumption
Completion indicators:
- packs can carry hub concepts, neighborhood structure, secondary products, and source-role metadata
- public and workbench consumers can rely on one shared substrate while rendering differently
Near-term priority sequence
The next practical Didactopus sequence should be:
- make distinctions first-class in review/export payloads
- add source-role metadata to Notebook-facing outputs
- upgrade
notebook_pageto summarize hub, neighborhood, and secondary lanes - expose secondary-product and distinction views in learner/workbench flows
- enforce explicit public citation and quote-marking contracts
What the pilot changed
Before the pilot, it was reasonable to think of the Notebook as a side product assembled from reviewed concepts.
After the pilot, the stronger view is:
- the Notebook should be the stable center of Didactopus knowledge organization
- learner products are better when built from Notebook structure
- short web captures and longer textbook sources should play different roles
- definitions, qualifications, constraints, and quotes are not optional extras
That change should guide both schema evolution and product decisions.