eduKateSG’s Ecosystem Atlas

Release, Contradiction and Change-Control Protocol

The federation is allowed to grow quickly, but broad growth cannot bypass ownership and verification. New work should be treated as a release into a live knowledge system, not as an isolated article.

1. Resolve the job. Identify whether the reader needs discovery, public reference, specialist depth, learner diagnosis, recovery, local implementation or routing.
2. Check the owner. Search the live estate and boundary ledger before creating a new broad URL.
3. Classify the claim. Separate durable mechanism from time-sensitive fact, local implementation and interpretation.
4. Verify the evidence. Prefer the strongest current source for claims that can change; record uncertainty when the evidence does not justify precision.
5. Publish into the graph. Add the new page beneath the correct hub and link both directions where the relationship is durable.
6. Run collision control. Check whether another page now answers the same job; narrow, support or hand off rather than allowing silent owner duplication.
7. Propagate important changes. When a canonical owner changes a rule, route or definition, update the dependent handoffs that would otherwise send readers into stale guidance.
8. Observe the system. Use search data, reader behaviour and internal audits to find blind spots, not to replace editorial judgement.
9. Preserve correction history. Correct the live owner and keep the reason for the change intelligible; do not create a second competing owner merely because the first needed repair.

Release control line: need → owner check → evidence check → build or strengthen → connect → collision check → publish → observe → correct → propagate. This protocol is the operating counterpart to Project Atlas’s public trust and evidence covenant.

AI and Educational Technology Governance

AI is a federation-wide capability layer, not a ninth owner. It can improve access, explanation, practice, feedback, planning, translation, simulation and verification, but every AI-assisted route must preserve the underlying learning job, the canonical subject owner, authorship and the ability to test independent performance.

• Public literacy, tool choice and integrity: eduKateSingapore.
• Learner-runtime use: eduKateSengkang, especially How AI-Assisted Study Works.
• Tutor implementation and human judgement: The Tutor System.
• Mathematics technology: Bukit Timah Tutor Technology Atlas.
• English validation: Singapore English Tuition Centre.
• Science validation: eduKateSingapore Science World.
• Systems, authorship and institutional mechanism: eduKateSG.

AI release test: What learning job is being assisted? What is the tool allowed to do? Which source or subject owner verifies the result? What must the learner still explain or perform alone? What evidence shows the learner is stronger after the tool is removed? If those questions cannot be answered, the AI workflow is not yet an education workflow.

Access, Inclusion and Learner Variation

The federation should not treat difference as a reason to create a parallel, lower-expectation education system. Access changes the route, representation, support, interface, timing or response mode while preserving the underlying learning goal wherever that goal remains appropriate.

• Inclusive-education system design: eduKateSG owns the mechanism-level questions of needs identification, support planning, access, participation and system design.
• Adaptive teaching and learner runtime: eduKateSengkang owns changing the next teaching move from learner evidence without silently lowering the target.
• Parent-facing learning-difference routes: eduKateSingapore owns public guidance and tutor-selection context for learning differences.
• Mathematics accessibility technology: Bukit Timah Tutor owns accessible representation and technology in Mathematics.
• Language access: English-language structure and communication routes remain with Singapore English Tuition Centre; multilingual explanation may support access but should not distort the underlying English target.
• Routing: eduKateAI / Orchard resolves unclear access needs to the correct learning, specialist, recovery or public-reference owner.

Inclusion test: What is the real learning goal? What barrier is preventing access or expression? What adaptation changes the barrier rather than the construct? What evidence will show that the learner can use the capability with the intended level of independence? Different access does not automatically mean a different intellectual destination.

Lifelong Learning, Career and Adult Capability

The federation does not end at school graduation. Adult learning has different constraints—work, family, time, prior experience, professional standards, career transitions and the need to turn learning into usable capability—but it still belongs to the same education graph.

• Ongoing career and adulthood learning: eduKateSG Career & Adulthood Hub owns continuing education, professional growth, career education, lifelong learning and the transition from formal education into adult capability.
• Public knowledge and research: eduKateSingapore World Knowledge Research Library supplies reference, research literacy and wider knowledge beyond the school syllabus.
• Career change and capability rebuilding: eduKateYishun owns skills conversion, career change and rebuilding when an existing route no longer works.
• Advanced specialist learning: Mathematics continues through Bukit Timah Tutor; English, communication and language performance continue through Singapore English Tuition Centre.
• Tutor discovery: eduKateSingapore provides public routes for university, adult and professional tutoring when direct instruction is the appropriate intervention.
• Human routing: eduKateAI / Orchard helps resolve ambiguous life, work, study and family questions to the correct wing.

Lifelong-learning control line: current capability → desired role or purpose → gap → owner → learning route → practice or work product → evidence of capability → transfer into real use → review → next stage. Credentials may be useful evidence, but the federation should also ask what the learner can now understand, do, explain, build, decide or contribute.

Educators, Tutors and Education-System Practice

The federation serves educators as well as learners. Teaching quality depends on more than subject knowledge: diagnosis, curriculum coherence, assessment, feedback, professional learning, implementation, classroom conditions and institutional design all affect what learners can actually do.

• Teacher, school and education-system design: eduKateSG Education Hub owns the broad system layer, including How Education Works, teacher professional learning, curriculum, assessment and institutional mechanisms.
• Tutor professional practice: The Tutor Handbook owns evidence-informed tutoring decisions, tutor coaching, implementation, programme quality and the professional operating manual for tutoring.
• Mathematics pedagogy: Bukit Timah Tutor owns specialist Mathematics teaching, representation, reasoning, technology and advanced mathematical routes.
• English pedagogy: Singapore English Tuition Centre owns specialist English teaching through its English system and English Tutor Handbook series.
• Public reference and resources: eduKateSingapore supplies curriculum, evidence, research literacy and public learning resources.
• Research-to-practice: eduKateSG owns the mechanism and Sengkang/Tutor System owns the implementation evidence loop.

Educator control line: learner evidence → instructional problem → curriculum/subject owner → teaching move → implementation → observation → feedback → professional review → learner retest. A teaching method is not successful because it was delivered; it is successful only when the intended learner capability becomes more reliable and transferable.

Capability Graph Governance

The federation uses one cross-subject capability overlay so different wings can talk about learner performance without flattening subjects into generic skills. The canonical learner-runtime version lives in the Sengkang Learning Atlas.

• Subjects keep their own knowledge structures. English meaning, mathematical reasoning and scientific explanation are not interchangeable just because they may all involve representation or verification.
• Capabilities are evidence questions. Orientation, representation, meaning recovery, retrieval, reasoning, strategy selection, execution, communication, verification, transfer, self-monitoring and independence describe what a learner can currently do with the relevant knowledge.
• States and capabilities are different. A capability can be blocked, supported, stable, transfer-ready or examination-ready; the state tells us how reliably it survives.
• Interventions target the first limiting capability. Do not reteach an entire subject when one upstream capability is the actual bottleneck.
• Progress requires changed-condition evidence. Familiar success is not enough; transfer and independence must eventually be tested with support reduced.
• Adult learning uses the same overlay. The capability graph continues into university, professional practice, reskilling and lifelong learning even when the subject or workplace context changes.

Capability control line: subject knowledge → target capability → current state → evidence → intervention → retest under changed conditions → transfer → independence. The graph lets the federation connect school learning, tutoring, examinations, adult capability and civilisation without pretending they are the same curriculum.

Assessment, Measurement and Capability Evidence

The federation separates score, evidence, capability, state, opportunity and outcome. A mark is useful information, but it is not the learner.

• Assessment-evidence runtime: eduKateSengkang owns the generic question of what an assessment can tell us, what it cannot tell us and how evidence should change the next teaching move.
• Public formats, objectives, cohorts and rules: eduKateSingapore Curriculum & Examination Library.
• Subject-specific assessment: Bukit Timah Tutor for Mathematics; Singapore English Tuition Centre for English; Science World plus Sengkang runtime for Science knowledge/performance interpretation.
• Independent-performance test: success with prompts, homework help, visible models or AI assistance is not automatically evidence of independent mastery.
• Opportunity check: missing evidence is not the same as missing capability; a blank, inaccessible task or unobserved skill may mean the learner did not have a fair opportunity to demonstrate the construct.
• Construct coverage: a small sample should not support a broad mastery claim unless it represents enough of the intended capability.
• Recovery: when evidence shows that the current route has broken down, translate the result into a recovery plan through eduKateYishun.

Measurement control line: task → evidence → construct → context/support conditions → interpretation → capability/state estimate → next action → retest. Grades and credentials may summarise performance for a purpose; they do not replace the underlying evidence graph.

Student Data, Privacy, Consent and Safeguarding

Learning evidence can be useful only if it is handled responsibly. The federation therefore treats student data, safeguarding and consent as operating constraints, not optional footnotes added after the learning system is designed.

• System-level student-data governance: eduKateSG owns necessity, proportionality, access, retention, security and accountability for education data.
• Safeguarding systems: eduKateSG owns the system-level questions of prevention, reporting, referral and accountability.
• Tutor data practice: the Tutor Handbook owns day-to-day learner-progress evidence, data ownership, retention, anonymisation and sharing decisions.
• Purpose limitation: collect or retain learning evidence because it serves a clear instructional, safeguarding, administrative or legal purpose—not because storage is easy.
• Minimum necessary access: parents, learners, tutors, staff, AI tools and external services should see only the information required for the legitimate job they are performing.
• AI caution: identifiable learner work, personal data, school records or sensitive information should not be sent into tools merely for convenience when the same learning job can be done with less data.
• Human escalation: when a situation becomes a safeguarding, welfare, legal or emergency concern, the normal learning/tutoring workflow yields to the appropriate responsible adult, institution or professional route.

Data control line: purpose → minimum data → authorised access → secure use → instructional value → review → anonymise, retain or delete. Safeguarding control line: concern → immediate safety → responsible adult/institution → record only what is necessary → appropriate referral → follow-up. The learning graph must never make efficient routing more important than learner safety or privacy.

Publication Taxonomy | Every Page Has a Job

The federation does not treat every URL as the same kind of content. Each page should have a declared role in the graph so readers, editors and search systems can tell whether it is defining, routing, teaching, practising, measuring, updating or documenting.

Minimum page contract: every serious page should be able to answer four questions: What job do I own? Which broader owner am I under? What evidence or learning action should happen here? Where should the reader go next? Time-sensitive pages add a fifth: When must this be reviewed?

Content Lifecycle | Published Is Not a Sufficient State

A large education estate needs more than draft and published. A page may be correct, useful and indexable while no longer being the canonical owner. Lifecycle status therefore describes the page’s current job, not whether its URL still exists.

Lifecycle rule: never create a replacement page merely because an older owner is imperfect. First decide whether the old page should be repaired, narrowed, superseded, archived or retired. The URL history, backlinks, reader expectations and internal graph are part of the decision.

Content Build Standard | Owner, Route, Facet or Do Not Build

Scale is useful only when every new page has a distinct job. A competitor page, a high-volume keyword or an empty slot in a spreadsheet is not by itself a reason to publish. Search can reveal demand; the federation still decides whether that demand deserves a new owner, a supporting route, an update to an existing page or no new page at all.

• Canonical owner: create or strengthen when the job is broad, durable, important and not already owned. Major owners should be deep enough to define the system, connect mechanisms, show examples, diagnose failure, support transfer and route onward. Depth is earned by coverage, not padding.
• Supporting route: use for one mechanism, question, skill, example, paper component, learner state or implementation problem. It must point clearly to its owner.
• Local facet: use only when locality changes the reader’s decision, school context, travel, implementation, programme or community information. Do not rewrite generic subject knowledge with a postcode.
• Time-sensitive page: use when date, cohort, policy, schedule or current evidence materially changes the answer. It needs a review path and current-source discipline.
• Atlas / directory: maps owners, states and routes. It is not an excuse to create a second canonical explanation of every node it links to.
• Do not build: when an existing owner can satisfy the same job after a surgical update or handoff.

Editorial standard: major owner pages are normally comprehensive longforms when the subject genuinely requires it; narrow jobs may remain narrow. Search language should appear naturally where it helps readers and machines understand the job, especially early in the page, but competitor wording never sets the intellectual ceiling. Titles name the reader job and owner clearly rather than stamping one repeated commercial suffix across the estate.

Ecosystem Observability | How We Know the System Is Healthy

Traffic is only one signal. The federation should be monitored as a knowledge and learning system, with health measured across ownership, evidence, freshness, routing, accessibility and real educational usefulness.

• Owner clarity: major intents resolve to one canonical owner without two master pages claiming the same job.
• Collision rate: overlapping pages are detected early and narrowed, connected or consolidated before they become permanent architecture.
• Route integrity: important pages have a clear parent, owner, next route and return path; dead ends and orphan pages are exceptions.
• Freshness debt: time-sensitive claims, official rules, dates, formats and pathways are reviewed before they become stale guidance.
• Evidence debt: consequential claims remain traceable to appropriate sources, with uncertainty visible where evidence is incomplete.
• Handoff health: specialist, local, runtime, recovery and public-reference pages link to the correct owner rather than to whichever page was published most recently.
• Learning-path coverage: readers can move from explanation to practice, diagnosis, transfer, examination, recovery or advanced study without leaving a conceptual gap.
• Accessibility and inclusion: routes remain usable across different access needs, response modes and learner conditions without lowering the target invisibly.
• Machine consistency: titles, metadata, structured data and internal anchors describe the same site role and do not redefine a property accidentally.
• Search telemetry: impressions, ranking changes, CTR and query-page relationships identify discovery gaps, snippet weakness and cannibalisation; they trigger diagnosis, not automatic rewriting.
• Quality over volume: growth in page count is not treated as progress unless coverage, navigation, ownership or learner usefulness also improves.

Health loop: observe → detect drift → locate owner → inspect evidence → repair the smallest broken layer → recheck routing → remeasure. The preferred fix is surgical: correct the layer that is wrong without destabilising layers that are already working.

Learning Object Contract | What Every Instructional Page Must Carry

Whether the object is a lesson, guide, Learning Manual, worksheet, diagnostic, casebook or specialist explanation, it should make the learning job legible. The exact format may differ, but the underlying contract should remain recoverable.

Instructional rule: exposure is not the same as learning. A strong learning object should make it possible to move from explanation to learner action, from learner action to evidence, and from evidence to transfer or repair. Objects that only inform may still be valuable reference pages, but they should not claim an instructional job they do not perform.

Multilingual Learning, Translation and Localisation

Global growth requires more than translating words. The federation must preserve concept, evidence, learner job, cultural context and owner relationships as content crosses languages.

• Translation and localisation mechanisms: eduKateSG owns how meaning, terminology, context, localisation, QA and translation systems work.
• English target-language structure and mediation: Singapore English Tuition Centre owns how multilingual speakers move meaning into, through and around English without losing the English learning target.
• Public translation studies and reference: eduKateSingapore owns the public overview of Translation Studies, interpreting, localisation, technology and human judgement.
• Bilingual scaffolding inside learning: eduKateSengkang may use Chinese or other first-language explanation to support English learning, but the scaffold does not become the target capability unless the page is explicitly teaching translation or mediation.
• Concept ownership survives language change: a translated page normally remains under the same canonical concept owner. A second language does not create a second broad owner.
• Cultural adaptation must be explicit: examples, idioms, institutional references and audience assumptions may need localisation; the underlying claim and source trail should remain traceable.
• Machine translation requires verification: fluency is not proof of preserved meaning. Important educational, technical, legal, scientific or assessment claims require subject-aware checking.

Multilingual control line: canonical concept → source meaning → target learner/audience → translation or mediation → subject/language verification → learner-use test → cross-language handoff. Translate the route without duplicating the owner.

Qualifications, Pathways and International Comparability

Different education systems should be compared carefully, not stacked into one imaginary global ladder. Programme names, school years, qualifications and examinations can look similar while measuring different content, depth, assessment modes and purposes.

• Comparability mechanisms: eduKateSG owns qualifications frameworks, level descriptors, ISCED, credential recognition and the mechanisms that make different systems comparable without pretending they are identical.
• Current Singapore public pathways: eduKateSingapore owns the parent-facing map of the Singapore education system, supported by the Curriculum & Examination Library for current rules and cohorts.
• Mathematics crosswalks: Bukit Timah Tutor owns world Mathematics curriculum and examination comparisons, including Singapore, IB, IGCSE, UK, US and other routes.
• English examinations: Singapore English Tuition Centre owns international English examination and language-performance comparisons.
• Pathway recovery: eduKateYishun owns retakes, second-chance education and rebuilding when the current qualification route breaks down.
• Owner resolution: eduKateAI / Orchard routes ambiguous school, qualification, career and cross-system questions before a reader commits to the wrong specialist path.

Crosswalk rule: compare purpose → entry level → curriculum scope → assessment construct → depth → progression rights → recognition context → learner goal. Similar titles do not prove equivalence, and comparability does not automatically mean recognition or transfer credit.

Locality Protocol | Place Must Change the Job

Local search is useful when place changes the reader’s decision, constraints or implementation. A neighbourhood name alone is not educational value. The federation therefore separates local owner, local discovery facet and local knowledge.

• Actual local owner: a property may own family intake, consultation, teaching or delivery in its real operating geography. eduKatePunggol owns the Punggol family/local implementation layer; Bukit Timah Tutor owns Bukit Timah Mathematics implementation; other wings keep only the local jobs they genuinely perform.
• Local discovery facet: a page may help a reader find the correct subject/year/route from a place query, but it must route to the canonical subject and local owner instead of pretending the location creates a new curriculum.
• Local knowledge: schools, transport, libraries, study spaces, town history, community institutions and place-based learning may be valuable in their own right. These pages should teach something real about the place rather than exist merely to carry a locality keyword.
• No false-presence implication: a local page must not imply that eduKate has a physical branch, tutor, class or facility in a neighbourhood unless that is actually true.
• No postcode substitution: generic subject explanation should not be copied repeatedly with only the estate, MRT station or neighbourhood name changed.
• Local facts require freshness: schools, transport, facilities, operating hours, programmes and neighbourhood conditions may change and therefore need current-source discipline.
• Local pages inherit subject ownership: a Punggol A-Math page still routes mathematical depth to BTT; a local English page still routes deep English to SETC; a local Science page still routes underlying Science to Science World.
• Legacy locality is preserved intelligently: old local pages with history, backlinks or useful information may remain as supporting or historical facets, but they should not silently outrank the current owner inside the internal graph.

Local-page admission test: What does this place materially change? What local evidence or decision does the page add? Is there a real local owner? Which canonical subject/system owner sits underneath it? If removing the place name leaves essentially the same article, the page probably does not deserve a new local URL.

Family Decision Architecture

Parents often have to make decisions before they have complete information. The federation separates the different evidence streams so one school result, one tutor opinion, one local convenience or one search result does not become the whole decision.

• Decision framework and pathways: eduKateSG Parents & Pathways Hub owns the broad parent decision process across PSLE, Full SBB, subject levels, pathways, support and family choices.
• Public facts and current rules: eduKateSingapore owns curriculum, examination, tutor-discovery and Singapore education-reference information.
• Learner evidence: eduKateSengkang supplies evidence about the learner’s current state, first weak link, capability and response to support.
• Recovery options: eduKateYishun owns the decision when the existing route has failed and the family needs to recover, retake, rebuild or choose a second-chance pathway.
• Local implementation: eduKatePunggol owns Punggol family intake, consultation and local tuition implementation; specialist/local wings keep only the implementation they genuinely perform.
• Subject depth: BTT and SETC answer Mathematics and English specialist questions after the family decision has been narrowed to the subject job.
• Ambiguous situations: eduKateAI / Orchard resolves questions that cross learning, school, family, money, work or life before routing deeper.

Family decision line: family question → current facts → learner evidence → constraints and values → options → likely consequences → smallest reversible next step where possible → evidence after action → review. A parent decision should remain revisable when new evidence arrives.

Learner Agency and Independent Performance

The federation should gradually transfer control toward the learner. Help is useful when it builds capability that survives after the help is removed; it becomes a problem when every difficult step remains owned by the parent, tutor, answer key or AI tool.

• Broad student doorway: eduKateSG Student & Learner Hub helps students find the right subject, study, exam, recovery or pathway route.
• Learner runtime: eduKateSengkang owns capability state, first weak link, practice, transfer, self-monitoring and independence.
• Independent-performance test: Sengkang owns the question of what the learner can still initiate, choose, execute, monitor, check and adapt when unnecessary support is removed.
• Public knowledge: eduKateSingapore supplies the reference and learning material the learner may legitimately use.
• Subject depth: BTT, SETC and Science World supply specialist knowledge without taking over the learner’s decisions permanently.
• Recovery: Yishun owns the route when independence collapses because the underlying capability or pathway needs rebuilding.

Learner-agency line: understand the job → attempt → notice uncertainty → seek the smallest useful help → act → check → explain → transfer → reduce support → own the next attempt. The long-term destination is not a student who never needs resources or people; it is a student who can use them deliberately without surrendering the thinking.

Federation Relationship Registry | How the Graph Is Encoded

The operating Atlas uses the eduKateSG Crosswalk Registry & ID System as its canonical relationship grammar. The federation extension keeps the original XWALK law: connect systems without collapsing them into one another.

• Wing identity: every public wing has a stable FED.WING.* ID.
• Job identity: broad canonical jobs use FED.JOB.* IDs.
• Relationship identity: cross-site edges use XWALK.FED.{SOURCE}.{TARGET}.{RELATION}.v{VERSION}.
• Controlled relations: use registered terms such as OWNS, ROUTES_TO, RETURNS_TO, PUBLIC_REFERENCE, SPECIALIST_DEPTH, RUNTIME_DIAGNOSIS, RECOVERY_HANDOFF, LOCAL_IMPLEMENTATION, SUPPORTS, LOCAL_FACET_OF, TRANSLATION_OF, PRACTICE_FOR, ASSESSMENT_OF, EVIDENCE_FOR, SUPERSEDES, HISTORICAL_OF and CROSSWALK_TO.
• Direction matters: a handoff from Sengkang to Yishun is not the same edge as the return from recovery into the learner runtime.
• Identity is not equivalence: two wings can be related, affiliated or mutually routing without being the same entity. Do not misuse alternate names, sameAs or redirects to erase a real boundary.

Graph rule: visible links explain the relationship to humans; registry IDs make the relationship stable for governance, audits and future machine-readable layers. The page remains readable without the ID, but the ID prevents the meaning of the edge from drifting over time.

Freshness, Version and Review Protocol

Not every page ages at the same rate. A durable definition may remain useful for years; an examination date, school programme, fee, local facility or policy can become wrong quickly. The federation therefore separates concept version, relationship version and fact freshness.

Version rules

• Major version: change when the owner, construct, meaning, public job or core route changes materially.
• Minor version: change when the owner remains the same but the map, evidence model, relationship set or important coverage expands.
• Relationship version: crosswalk edges carry their own version because a page can remain stable while its destination or handoff changes.
• As-of date: time-sensitive claims should expose the date or cohort to which the claim applies; evergreen pages should not be forced into dated URLs merely to prove freshness.
• Review date is not truth: a recent edit timestamp does not prove accuracy. The underlying source and owner must still support the claim.

Freshness control line: claim class → owner → source → as-of date → review interval → trigger monitoring → update or archive → propagate changed handoffs. The system should spend maintenance effort where truth decays fastest rather than rewriting stable pages simply because they are old.

Canonical Owner Quality Gate | When a Page Is Strong Enough to Own the Job

A canonical owner is not declared by title, age, traffic or word count. It earns ownership by covering the job deeply enough that narrower pages can safely depend on it. Major owner pages should normally be substantial longforms when the subject genuinely requires that breadth, but length is a consequence of coverage—not a target to pad.

Owner-promotion rule: a page should not be promoted to canonical owner until it passes the gates relevant to its job. Supporting pages may be narrower by design. A reference table does not need a full diagnostic programme; a flagship learning owner usually does. The standard is functional completeness for the role, not identical length or structure across the estate.

Federation Page Contract | The Minimum Metadata of a Serious Page

Every important page should be able to resolve the fields below, whether they are stored in visible copy, structured data, internal registry data or editorial metadata. The contract makes pages easier to route, audit, maintain and connect without forcing every article into the same visual template.

PAGE_CONTRACT:
  canonical_id:
  page_role:
  canonical_owner:
  job:
  audience:
  subject_or_domain:
  stage_or_level:
  geography:
  time_scope:
  target_capability:
  prerequisites:
  evidence_expected:
  practice_or_action:
  assessment_or_verification:
  parent_owner:
  supporting_routes:
  next_routes:
  return_route:
  relationship_ids:
  source_class:
  freshness_class:
  as_of_date:
  lifecycle_status:
  version:
  review_trigger:

• Role before format: a page can be a flagship, Atlas, manual, lesson, assessment, local facet or historical record; its job determines which fields matter most.
• Scope must be explicit: school level, geography, curriculum, examination, language, audience and date should be stated when they materially constrain the answer.
• Capability must be testable: learning pages should identify what the learner should become able to understand, retrieve, decide, explain, perform or transfer.
• Evidence must match the claim: the page should not claim mastery, readiness, effectiveness or equivalence from evidence that cannot support that conclusion.
• Routes are part of the content: parent owner, specialist handoff, practice, assessment, recovery and return path are not footer decoration; they are part of how the learning system works.
• Maintenance is designed at publication: freshness class, as-of date and review trigger should be decided when the page enters the graph, not years later when someone notices it is stale.

Contract rule: if an important page cannot say what it owns, what capability or decision it serves, what evidence matters, where it sits in the graph and when it needs review, it is not yet fully integrated into the federation.

Change Ledger and Release Receipts

A long-lived knowledge system needs institutional memory. Important changes should therefore leave a compact release receipt that records what changed, why it changed, which owner made the change and which dependent routes were checked. This is not a public changelog for every typo; it is a governance record for changes that affect meaning, ownership, routing, evidence or currentness.

CHANGE_RECEIPT:
  change_id:
  date:
  owner_id:
  page_or_registry_id:
  change_type:
  previous_state:
  new_state:
  reason:
  evidence_or_trigger:
  affected_relationship_ids:
  affected_pages:
  machine_identity_check:
  internal_route_check:
  freshness_check:
  rollback_or_recovery_note:
  status:

Propagation rule: an owner change is not complete when one page is edited. Recheck the parent hub, major inbound routes, local facets, specialist handoffs, structured data, metadata, registry edges and any current page that names the old owner. The release receipt closes only when the graph is coherent again.

Correction rule: prefer an intelligible repaired history over silent duplication. When a page needed correction, strengthen the live owner and preserve enough context that a later editor can understand why the system changed.

Release Receipt | Federation Governance Reset — 26 September 2026

change_id: CHANGE.20260926.FEDERATION.001
owner_id: FED.WING.EDUKATESG
change_type: OWNER_CHANGE + CROSSWALK_CHANGE + SCHEMA_IDENTITY
previous_state:
  - several wings were being hardened as if they were independent SEO properties
  - Orchard public routing still reflected an older five-wing map
  - some site identities over-emphasised tuition delivery format
  - eduKateSG machine identity used "eduKate Singapore" as an alternate name
  - cross-site ownership rules contained older A-Math, civilisation and locality assumptions
new_state:
  - eight-wing federation with distinct canonical jobs
  - Project Atlas = public master cartography
  - Ecosystem Atlas = operating governance and build ledger
  - Orchard = reception / owner resolution / return path
  - Sengkang = learner runtime
  - Yishun = understanding-first interpretation / Well Being / recovery / rebuilding / transformation
  - BTT = Mathematics specialist
  - SETC = English specialist
  - eduKateSingapore = public knowledge / evidence / reference
  - eduKatePunggol = Punggol family / town and place-based learning / local implementation
  - eduKateSG = systems / research / discovery / civilisation mechanisms
  - stable federation IDs and relationship vocabulary registered
reason:
  - semantic drift was detected when tuition/3-pax language began defining whole-domain identity
  - whole-estate review showed the deeper Atlas architecture was stronger than the SEO-facing apex language
evidence_or_trigger:
  - live WordPress identity / Atlas review
  - GSC cannibalisation and query-page relationships used as telemetry
  - existing Crosswalk Registry and EducationOS architecture
affected_pages:
  - Project Atlas
  - Ecosystem Atlas
  - Orchard ecosystem guide / runtime
  - Punggol homepage machine layer
  - Sengkang Learning Atlas
  - Yishun Recovery Atlas
  - BTT Mathematics and A-Math owners
  - SETC English owners
  - subject / tutor / examination / vocabulary / science handoff pages
machine_identity_check:
  - Project Atlas eight-wing ItemList has stable website IDs
  - eduKateSG alternate-name collision with eduKateSingapore removed
internal_route_check:
  - major English, Mathematics, Science, Vocabulary, tutor, examination, adult, AI, inclusion, family and learner routes reconnected
rollback_or_recovery_note:
  - useful SEO/contact/schema fixes were preserved
  - only the identity drift and over-templated ownership signals were rolled back or narrowed
status: CLOSED / VERIFIED

This receipt is deliberately architectural. It records the point at which eduKate stopped treating SEO hardening as a collection of site-by-site optimisations and formalised the estate as one federated education system with distinct owners and explicit handoffs.

Algorithms, Computing and Computer Science

The federation separates the field from the learning runtime. Bare algorithm and computer-science questions are not automatically the same job as teaching a learner to trace, reconstruct, design, implement and evaluate an algorithm.

• Public computing knowledge: eduKateSingapore Engineering, Computing and Technology Library owns the public reference layer for computing, engineering and technology; What Is Computer Science? carries the broad discipline definition.
• Algorithms/computing learning runtime: eduKateSengkang owns how a learner builds computational capability through tracing, reconstruction, design, implementation, testing, complexity reasoning, transfer and professional verification.
• Systems crosswalk: eduKateSG owns the mechanism-level handoff between computing, Science, Technology, Engineering and Mathematics.
• Mathematical depth: algorithms that depend heavily on proof, optimisation, numerical analysis or advanced Mathematics may hand off to Bukit Timah Tutor for the mathematical layer without moving the whole computing job there.
• AI: AI remains a tool/technology layer inside computing and education; it does not automatically become the owner of the underlying algorithm, subject or learning capability.
• Search telemetry: if a Sengkang “How to Learn X” page ranks for the bare name of algorithm X with weak CTR, first diagnose query-intent mismatch. Do not automatically rename the learner-runtime page into a generic encyclopedia entry. Strengthen the public concept/reference route where a real gap exists.

Computing control line: field/concept → public reference → learner goal → algorithm/computing runtime → implementation evidence → transfer → specialist mathematical/system handoff where needed. The estate may be deep without becoming semantically duplicated.