Repository navigation
Description
Outcome and recommended next step
Make the dictionary usable as a living encyclopedia: read sourced meanings, discover cultural material, and publish explainable, revisable examples selected by curators. Start with a rich mixed-content reader, CineGraph film selection and GIPHY discovery, supplemented by sourced art and quotations.
Confirm the completed #84 correctness gate against the implementation branch, then ship one complete rich-reader and contribution workflow. Do not wait for every provider, general entity import or AI persona before putting a useful reader online. Equally, do not put an attractive submission interface on relationships whose approval can become stale without warning.
This issue is the product rollout and acceptance tracker. #84 remains the execution contract for connection correctness and minimal authoring. #65 owns quote provenance; #66/#67 own the reader/evidence-wall experience; #72 owns relevant source/deployment work; #75 and historical #17 cover curator bots. Link focused implementation PRs/issues here; do not launch all of these as one concurrent rewrite or declare their broader requirements complete through this tracker.
Research baseline: dictionary main at ab0cc0a, archived Rails tag ruby-rails-archive, and local CineGraph source at cc030834, inspected 2026-09-11. Recheck current code before implementation. No legacy data migration is required.
Product sequence updated 2026-09-12: bring clearly labeled automatic discovery into the first reader milestone. The earlier review-first rollout is replaced by the sequence below. #84 received an A in the latest local audit; confirm its merge/check status and reuse its completed implementation rather than scheduling its repairs again. That audit is not evidence that deployment gates or provider integrations are complete.
What we used to have, and what to restore
This is a source/template inspection of the old website, not a claim that its old deployment or every flow still runs.
| Archived capability | Useful product behavior to recover | What must change |
|---|---|---|
| Quote search and submission, citation and disputed/misattributed fields | Find a quotation, inspect it, attach it to a topic | Select an exact meaning; retain exact text, passage/source and separate attribution. Never infer the author from the first name-search result. |
| Artsy and GIPHY search/select controllers | Preview provider results and choose an item | Save one reusable artifact with separate contextual claims, using current provider policies. |
| Topic page automatically searched Artsy/GIPHY by topic title | Immediate discovery suggestions | Search strings are candidate discovery, not proof that an item illustrates this particular meaning. |
| PoetryDB search interface/controller | Find poems by title/author | Verify and finish the complete selection/publication flow; some archived paths were stubs. |
| Other provider code: TMDB, Unsplash, Genius, OpenLibrary, knowledge graph | Future adapter ideas | Inventory, not evidence of complete working integrations. Do not port everything. |
| Emoji associations | Lightweight expressive participation | Not found in inspected archive; treat as a proposed feature, not a verified restoration. |
See archived quote submission, quote search, Artsy picker, GIPHY picker, and topic discovery.
Do not restore the old name-based author matching, URL-only quote deduplication, or delete-the-shared-artifact behavior when removing an attachment. Several quotations can come from one page; one artifact can illustrate several meanings.
The product model: three separate jobs
- Lexical backbone: stable word identities, source-specific meanings/definitions, factual source relationships and revisions. Import appropriately licensed lexical/reference datasets through existing adapters. A word is not owned by the source that first supplies it; a definition remains attributable to its source.
- Cultural objects: a film, quotation, poem, artwork, GIF, person or publication exists independently. Discover through a provider or enter locally; persist selected objects with identifiers, permissible metadata, attribution and provenance. We do not need a copy of the provider's entire catalog.
- Contextual connections: “this film illustrates this meaning of war” is its own attributed claim, with rationale, evidence and review. A provider tag, a curator interpretation and a sourced factual relationship are distinguishable.
Museum metadata is useful evidence about an artwork. It does not make an interpretation of the artwork objective. Likewise, popularity on GIPHY or a high film rating does not establish relevance to a meaning.
Licensed lexical sources ──> WORD ──> SOURCE MEANINGS / DEFINITIONS
↑
│ reviewed contextual connection
│ (who, why, evidence, version)
CineGraph / quote / art / GIF ──> SELECTED CULTURAL OBJECT
│ ↑
└── search results ──> discovery records ──> labeled reader previews
│
human / bot proposal
│
human review ──> selected connection
Reuse #84's object, content, claim and review mechanisms. Add only the missing provider references, discovery state and editorial selection structures after mapping them to existing schemas. No second generic graph, separate media identity system or recreation of the Rails topic tables.
What readers see by default
“Canonical” here means the current editorial selection for a particular meaning, not an invented authoritative definition or a universal truth judgment.
- Source definitions remain individually labeled; specify their display order and keep alternate senses accessible.
- Curators can feature/order accepted examples per meaning and language, with recorded decisions and revisions.
- Accepted examples, pending proposals and machine-discovered results are different states. Eligible provider discoveries may appear in the mixed wall with a visible discovery label; pending user claims do not silently become published interpretations.
- Relevance votes later help order eligible examples; they do not validate attribution, change definitions or override withdrawal.
- Missing content stays empty. Do not fill every word with weak matches to make the page look complete.
- Refresh, rejection, rights withdrawal, merges and splits invalidate/reconcile the selection through Make encyclopedia connections reliable: identity changes, exact evidence, general entities and authoring #84's rules. A pinned card cannot bypass visibility checks.
NARCISSISM Meaning: self-absorption [change]
A clear definition · source · other meanings
DEFINITIONS FROM DIFFERENT SOURCES
[Dictionary definition] [Historical commentary, if available]
[Community definition, if available]
IN CULTURE [All | Art | Film | Quotes | GIFs]
┌ Artwork ───────────────────┐ ┌ Film ───────────────────────┐
│ Image │ │ Poster │
│ Echo and Narcissus │ │ Title · year │
│ Why it relates │ │ Why it appeared │
│ Selected by Alex │ │ Discovered via provider │
│ Source · details │ │ Source · details │
└────────────────────────────┘ └─────────────────────────────┘
┌ Quotation ─────────────────┐ ┌ GIF ────────────────────────┐
│ Exact text and author │ │ Preview │
│ Original source │ │ Search association │
│ Selection / attribution │ │ Provider attribution │
└────────────────────────────┘ └─────────────────────────────┘
Related meanings · people · works
[Suggest a connection] [Explore more]
On mobile, use one readable column with the same source/status information. Reuse the branding/mockup work already underway; do not start a competing redesign. Keep the primary definition immediately readable, with cultural material below it. Historical/community definitions appear only when genuine material exists: never invent a quotation or force every word into three definition tiers. Keep technical revision IDs and internal review vocabulary in details rather than the main reader experience.
Three presentation levels, one reusable reader
- Discovered: the provider returned the item for a recorded query or tag. Explain why it appeared and identify the provider. This is retrieval, not endorsement or an approved semantic claim.
- Selected: an editor chose the item as an illustration of this exact meaning, with rationale and the existing review guarantees.
- Community interpretation: a named contributor proposes a contextual connection and supplies their explanation/evidence. Publication follows the existing review policy; this label is not a bypass for pending or rejected claims.
These are presentation distinctions mapped to existing discovery/claim/review state, not a requirement for a new competing status system. Harmless, policy-compliant provider previews do not each require manual approval before being shown as discoveries. Adverse real-person interpretations do require explicit attribution and review; provider keywords never become factual classifications.
Start ranking with query/tag relevance, duplicate removal, content-type variety and eligible editorial selections. Provider popularity may help within that provider; GIF views and film ratings are not one universal relevance score. No elaborate scoring model, embeddings or fine-tuning prerequisite.
Definitions render independently of providers. Serve allowed cached previews promptly and refresh in bounded background work; never synchronously fan out to every API on each page view. Use lightweight thumbnails and load permitted embeds on intent rather than autoplaying a wall of media. Missing content gets a deliberate empty state, not forced matches.
Provider strategy and actual constraints
| Source/type | Access pattern | First implementation / gate |
|---|---|---|
| Current lexical/reference sources | Existing bounded/resumable imports and refresh | Preserve source meanings; expand coverage separately under #72/#84. ConceptNet relationships are not automatic cultural-example approvals. |
| CineGraph films | Search/lookup on demand, save selected references | First external integration. Known-film selection works with current API shape; thematic discovery needs additional work described below. |
| Quotes and poems | Manual sourced entry first; bounded provider search next | Exact text, creator attribution, publication/locator, language, disputed attribution and display rights. Wikiquote/PoetryDB search results require verification; discovery providers are not automatically primary evidence. |
| GIPHY | First-slice on-demand search/select and bounded discovery | Confirm production access/quota, required attribution, rating controls and permitted caching/display. Retain provider reference; do not assume permission to rehost assets. Current docs. |
| Artsy | Conditional integration or reference links | Public API retirement is announced. Confirm partner/supported access before investing in an adapter; it must not block launch. Official notice. |
| Open museum collections | Alternative first art provider | Evaluate Met or Art Institute of Chicago, including record-specific image/public-domain flags. Metadata access is not blanket image permission. |
| Urban Dictionary | Permission-dependent, attributed access | No bulk absorption in this plan. Its terms require express permission for API access. Confirm access and permitted storage/display first; otherwise use an external reference link. The word/local meaning can exist without this integration. |
| Emoji | App-local Unicode association picker | Later lightweight reader expression. “Readers associate this emoji” is not a definition/factual claim. Animated provider emoji are a separate integration. |
A new adapter must document search, lookup, normalization, external identity, refresh/withdrawal, attribution/display policy, capability limits and request budgets. Credentials remain server-side. Provider failure must not break reading existing definitions or allowed stored examples. Cache/storage rules are provider-specific.
Start with the film service we control
Working assumption: “Synograph” means CineGraph, the project found locally. Confirm intended service and deployed endpoint before integration; no production API was exercised during this research.
Verified in its schema, movie resolver, and router: authenticated /api/graphql, individual film lookup, batch lookup and search_movies(query, year, limit). Movie search currently matches titles, not themes. The schema also supplies useful film metadata and canonical links.
Deliver in two increments:
- Curator searches a known film title, previews it, selects the exact film, and supplies the reason it illustrates a meaning. Persist a namespaced identifier and supported cross-identifiers; reconcile evidenced existing entities without name-only merging.
- Add/verify a bounded theme/genre/keyword discovery capability in CineGraph, coordinated as provider work. Map our selected meaning to provider terms explicitly. Do not implement “films about war” by pretending title substring search is semantic search. Film ratings/cultural scores must remain separate from relationship relevance.
ADD AN EXAMPLE
Meaning: war — armed conflict [change]
Type: [Film] [Quote] [Poem] [Art] [GIF] [Link/local item]
Provider: CineGraph Search title: [................]
[Film results: title, year, creator, source link] [Select]
Why does this illustrate the meaning? [................]
Evidence / scene / passage / source: [................]
Interpretation by: [me / cited claimant]
[Preview] [Submit for review]
REVIEW
[Object identity] [Exact meaning] [Rationale] [Evidence versions]
[Accept] [Request changes] [Reject]
After acceptance: [Feature on this meaning's page]
Delivery order and dependencies
0. Confirm and reuse the completed #84 safeguards
- Confirm checkpoint 1's visibility, identity-change, exact-version approval and ownership guarantees are present on the implementation branch, including the repaired audit cases in the ordinary suite. Do not reimplement completed repairs.
- Reuse kind-neutral explicit selection from checkpoint 2 and the create/connect/review/revise path from checkpoint 3; identify any actual integration gap before changing the schema.
- Prove the above on the first film/quote workflow. Broader general-entity ingestion and extension proofs remain tracked in Make encyclopedia connections reliable: identity changes, exact evidence, general entities and authoring #84; do not silently waive or close them.
Read-only provider feasibility work and fixture-based UI design may proceed alongside these fixes. Production publication depends on the safeguards. This avoids waiting for unrelated breadth while preserving correctness.
1. Ship a rich reader with a bounded invited-contributor workflow
- Build the reusable “In culture” mixed-content section above, including type filters, source links, match explanations and visible discovery/selection distinctions (Design: Definition page — three-tier definitions + community evidence wall (wireframes, UI, UX) #66/Evidence Wall — masonry feed of mixed culture (tweets, lyrics, reels, TikToks, YouTube) with type filters #67).
- Connect CineGraph first for known-title search, preview and exact-film selection; GIPHY second for real search/discovery previews. Verify authenticated live access separately from fixtures. CineGraph thematic discovery remains a capability gate, not an assumed feature of title search.
- Add a small set of sourced art and quotations through the existing authoring path. They provide immediate variety without requiring four live integrations before the reader ships.
- Invited contributors can connect a selected object to an exact meaning, explain why, and submit for review. Reviewers can select/feature, hide and correct the target meaning through the existing revision/review mechanisms.
- Reuse one artifact across meanings with independent rationale/review. Removing one connection does not delete the artifact elsewhere.
- Deliver the three showcase cases below using the same templates, adapters and relationship model as every other word. Do not add special routes or runtime behavior for them.
- Check additional unprepared entries, then expand to a varied pilot of 20 entries; at least 10 have two useful reviewed examples. Discovery cards do not count as reviewed examples, and no requirement forces every media type onto every word.
This is the first product milestone worth showing people: mixed cards, two real integrations, and a small amount of carefully sourced editorial content. It does not require bots, Urban Dictionary, an entire art catalog or automatic thematic matching.
Run #86's bounded Artsy access probe alongside this work. Its announced retirement makes feasibility research time-sensitive, but Artsy must not become the reader's critical dependency. Evaluate an open museum collection as the durable alternative; check record-specific display rights.
2. Publish a small, curated public reader
- Complete the release checks under the relevant deployment work: protect
/ops/importsand other mutation routes, verify account/role boundaries, secrets, production configuration, backups/restore and basic monitoring. - Exclude audit/demo records from the launch corpus and discovery.
- Public visitors read/search the encyclopedia, curated examples and eligible labeled discoveries; contributions initially remain invitation-only. Display provenance, report/challenge access and deliberate empty states.
- Record mobile/desktop walkthroughs, provider failure behavior and representative query performance. This issue plans release; actual deployment is a separate execution action.
Do not postpone this milestone until every idea below exists. Conversely, a green local test suite alone is not a public-release gate.
3. Expand media through the same workflow
- Add a reusable adapter for one viable art provider and broaden quote/poem search with verified attribution and citation handling (Quote layer with provenance scoring #65). CineGraph and GIPHY are already in stage 1; do not schedule them again here.
- Reuse the picker/review/card lifecycle; document any genuine schema extension. Add URL/manual fallbacks where permitted.
- Add basic emoji associations only after participation semantics are defined. Defer full social feeds, every media provider and elaborate provenance scoring.
4. Broaden and measure the discovery introduced in stage 1
- Store reviewed mappings from a meaning/language to provider tags, terms or queries. Preserve whether the match came from a native tag, text search or inferred theme.
- Run bounded asynchronous discovery for chosen entries; do not fan out to every API on each page view or crawl the whole vocabulary. Specify per-provider request/result budgets, retries, refresh and deduplication.
- Expand eligible public discovery previews using the stage 1 labels and provider display policies. Route proposed interpretations to reviewers; do not require manual approval of every harmless search preview.
- Evaluate candidate relevance on the pilot before expanding: precision among the first five results per meaning, wrong-sense matches, duplicate rate, attribution errors, review acceptance and cost. Report denominators and failures; define the rollout threshold before evaluating.
- Automatic acceptance/featuring of semantic claims, if later desired, requires a separate narrow policy and measured evidence. Labeled retrieval previews are already allowed in stage 1. Popularity alone is insufficient.
5. Add curator personas, then broader participation
- Pilot The Dead Editors’ Society: historical personas curate examples and marginalia #75 with one or two explicitly labeled AI-inspired curators using the same candidate/review pipeline, after human curation provides examples of good decisions.
- Bots retrieve real artifacts, provide source links and explain the suggested connection. Log actor, model/version, inputs, retrieval provenance and cost; constrain provider tools and treat retrieved text as untrusted input.
- A “Bierce-inspired AI” is a separate service identity from historical Ambrose Bierce. Generated commentary is not a Bierce quotation or historical claim. No bot self-approval or votes disguised as independent people.
- After moderation capacity is demonstrated, add public submissions, relevance voting, contributor history and abuse controls. Broader recommendations and social features follow; bots need not block public reading.
Three showcase cases: design and framework proof
| Word | Initial cultural material | What the case proves |
|---|---|---|
| Situationship | Attributed modern definitions where available; dating-reaction GIFs; a selected film about ambiguous relationships; a sourced poem or quotation chosen by a curator | A newer word works without historical coverage or Urban Dictionary access. Source-independent identity, sparse definitions and wrong-sense discovery are handled honestly. No fabricated Bierce definition. |
| Narcissism | Echo and Narcissus as an artwork candidate, its mythological subject, vanity-related GIFs, and explicitly attributed interpretations of particular behavior | Person, mythological character, artwork and meaning remain distinct. Factual depiction/authorship differs from a curator's interpretation. Keep ordinary self-absorption and clinical meanings distinct. |
| War | Selected CineGraph films, poetry, artworks, and related meanings such as grief and courage | A work can relate to multiple meanings independently; adaptations remain distinct; a film about war need not contain “war” in its title. Theme retrieval must use a real supported capability. |
The artwork candidate is documented by the National Gallery. This is not a claim that it is available through Artsy or the Met, or that an image can be reused without checking its source. Its depiction of a mythological subject and its interpretation as an illustration of narcissism are separate relationships.
For a real-person proposal—for example, a reader interpreting a sourced Trump speech or incident as self-absorption—retain who made the interpretation, exactly what they cited and the intended meaning. Do not publish “person is narcissism” as a factual edge or infer a clinical diagnosis. Ordinary automated provider discovery is not authorization to generate adverse personal judgments. Use fictional people in regression fixtures.
These are editorial test cases, not dedicated product modes. Use seed/config data for the pilot, generic routes and reusable components. One artifact may also appear under another meaning, but only with its own eligible connection or separately explained discovery match.
First-delivery acceptance
- Three compelling pages use one reusable template and demonstrate several content types across the set.
- CineGraph and GIPHY integrations work against actual authenticated services; document capability/access limitations separately from fixture results.
- Each card has a source, an understandable match/selection explanation and the appropriate discovery or attribution label.
- Include sourced art and quotations through the existing authoring path; preserve exact quotation attribution and citation.
- Mobile and desktop browsing remain readable, with lightweight previews, no horizontal overflow and graceful provider failure.
- Exercise real adapter records through refresh, withdrawal and changed-meaning review behavior—not only bespoke fixtures.
- Inspect at least 10 additional unprepared words for wrong-sense matches, empty states and usable fallback behavior. The showcases must not conceal a three-page-only implementation.
- Demonstrate reuse across meanings, duplicate handling, corrected targeting and exclusion of audit fixtures from discovery.
Scenario tests: does this remain general?
| Scenario | Required behavior |
|---|---|
| Situationship | Word and local/source meaning can exist even without Urban Dictionary access. A slang definition is source-attributed; a GIF search hit may be shown as labeled discovery, but is not an accepted interpretation until selected/reviewed. |
| War | CineGraph known-title selection works first. Theme discovery later records the actual matching tag/query; a film about conflict need not contain “war” in its title. |
| Nepotism | A curator can link a film scene or artwork with rationale. The artifact is not reclassified as a dictionary definition. |
| Bank | A river image attaches to the riverbank meaning and does not appear under the financial meaning. |
| Ambrose Bierce | Historical person, actual authored definitions, sourced quotations and an inspired bot remain distinguishable; authorship is not inferred from a matching name. |
| Grief | A poem can illustrate grief and remembrance through separate interpretations. Translation/edition and original attribution remain available. |
| Justice | Conflicting interpretations can coexist with named claimants and evidence; votes do not erase a source definition. |
| Same-name films / people | Provider identifiers and deliberate selection prevent title/name-only identity collapse. |
| Several quotes on one page | Distinct quotations survive; reselecting the same quote does not duplicate it merely because another user submitted it. |
| Real-person character judgment | An adverse interpretation is not automatically generated/published as factual classification; use explicit attributed-claim review. Use fictional examples for this test. |
| Provider outage / removal | Reader remains usable; removed media respects display policy; independent editorial history is retained where allowed. |
| Meaning split / evidence edit | Featured examples reconcile or enter re-review; old approval does not cover changed material. |
Acceptance and implementation handoff
- Map each stage to its owning issue/PR; explicitly list deferred stages. This tracker is not “done” after the first feature ships, but the first release does not wait for tracker closure.
- Confirm stage 0 is present, then start a bounded PR for the mixed reader and CineGraph integration, followed by GIPHY discovery and showcase completion. Use one writer for shared schemas/review code; avoid parallel corpus writers invalidating measurements.
- Use reproducible provider fixtures for ordinary tests and separately document authenticated live smoke checks, permissions, quotas and unavailable capabilities. Never report fixture success as proof of production API access.
- Verify candidate → accepted → featured → changed/withdrawn flows from both endpoints, reuse across two meanings, rejected content visibility and permission enforcement.
- Run
mix precommitfor implementation changes; record relevant browser and performance evidence. No blanket expensive rebuild without a change that warrants it. - Report what a visitor and invited curator can actually accomplish, remaining limitations and the next release slice.
The architectural aim is modest and durable: sources supply definitions and objects; contextual claims explain their connections; editorial selections determine the default presentation; people and bots can keep proposing improvements without rewriting the backbone.
Backlog coordination
Use this issue as the current product rollout sequence. #66/#67 contribute reader and mixed-wall design, adapted to the existing model; their older schema proposals and illustrative copy are not implementation requirements. #65 contributes sourced quotation handling, without requiring elaborate numerical truth scores. #86 owns the bounded Artsy probe and art-provider decision. #72 owns genuinely remaining source/deployment gaps, including ConceptNet as separate source work. #75/#17 follow after discovery and human editorial examples exist.
Reconcile delivered requirements in #73/#74/#79 and the merge/closure state of #84 rather than replaying old rebuild plans. Preserve useful ideas from #16/#44/#45/#46/#47/#48/#59/#60/#61/#64/#68, but treat superseded roadmaps as historical context, not concurrent execution contracts. Any closure/supersession should map remaining requirements first; this issue update does not close them.
Public release still needs its explicit operational gates. Rich local pages and a green test suite do not by themselves establish production readiness.
Activity
holden commentedon Sep 11, 2026
The first implementation slice is now #88: automatic cultural discovery on meaning pages, CineGraph first, with curation deferred.
Provider prerequisite: razrfly/cinegraph#1126.
This reflects the latest agreed sequence: no contributor onboarding, selection/review UI or bots in the first delivery. Record keyword/query provenance now; later attributed selections reuse the same objects through the existing claims model. #88 includes reader wireframes, meaning-specific mappings, refresh/identity tests and the future curation boundary. It supersedes this parent's earlier requirement to ship a contributor/reviewer workflow in the first slice.
CineGraph is the first child issue's required live integration. GIPHY and Artsy/open-museum adapters follow through the shared discovery/card model, respecting their different retrieval/display policies. They are not prerequisites for closing #88. This parent remains the broader rollout tracker.
holden commentedon Sep 18, 2026
Closing as superseded. Its children shipped (#88, #91/#92, #93, #86) and its remaining tracks now have their own trackers: providers and the kit → #109, the Photos shelf and the multi-source rule → #116, exemplars → #105, the assessor → #101, personas → #97, the reader's page → #111. Nothing here is unowned.