Roadmap (text only — no scaffolding in code)

All items below are FUTURE and intentionally absent from the current release.

Shipped in v0.2: French (fr) and Tamil (ta), completing the T.H.E.F.T. language set (DECISIONS 13).

Neologism engine (compose affixes onto stems with constraint checking).
Analogy playground (A : B :: C : ?) over the affix graph.
Larger lexicon (beyond the 850-entry cap) with external reviewer workflow; promote draft → reviewed → published.
IPA coverage for all lemmas; audio.
Sandhi-aware Telugu/Hindi/Tamil segmentation display.
Export of the graph as RDF/Linked Data (kept out to avoid ontology creep).
Tamil and French review. The v0.2 French and Tamil records are first drafts; six meanings are flagged partial_coverage.

Known limitations

Nearly all content ships review_status: "draft"; counts are displayed honestly on /.
Atom structures for complex emotions/abstractions are coarse approximations (HYPOTHESIS).
Segmentations follow surface morphology; sandhi alternations recorded only as allomorphs.

Expert review workflow. A reviewer-facing view that lists draft records with their sources, so draft to reviewed can happen item by item.
Reference entry numbers. Verify and add CDIAL and DEDR entry numbers against the printed volumes, lifting the 0.7 confidence cap where justified.
Real-user performance data for Hindi-, Telugu- and Tamil-heavy pages, then reconsider the font strategy (decision 12).
Per-page social cards (one per word or meaning) instead of a single site-wide image.
Print or PDF edition of the technical report.