Elamur Groupantsmova-turtle-make
The groupant beside tools/turtle/src/make.rs in the Mova repository: what the code computes, what it never does, how to go back from its outputs, its invariants, cost and checks.
Facts
Implementation material. A Mova code model describes this code at one source revision: partial models, finite checks and their stated scope. It is not a reviewed general exposition, and its labels are shown as written, never raised.
Components
| component | label | where |
|---|---|---|
| approximant | partial | Forward approximant |
| ast | partial | syntax.json |
| composition | partial | Composition |
| cost | partial | Cost |
| descriptant | present | Descriptant |
| invariant | partial | Invariant |
| tests | finite evidence | tests |
Obernants
The typed relations around the forward model: the behaviour complement (what the code never does), the reverse reconstruction (every input compatible with an output) and specialised views.
| id | type | label | where |
|---|---|---|---|
behavior_complement | behavior_complement | partial source derived model | behavior_complement |
reverse_reconstruction | reverse_reconstruction | partial source derived model | reverse_reconstruction |
Model text
Descriptant
What Sea Turtle makes: gap questions from a Nautilus layer as draft lessons (wrong options drawn by FNV-1a from the layer, its class first, each held to the fit check; of the first twelve accepted the three shown least; the right one starred at a drawn place; a gap asked once per making; lessons dealt three exercises to two tests and closed when they keep the school's rule, a tail joining the lesson before), the made journal and its counts, the lessons_task teacher question, and LessonJob for the standing loop (held-out names never read; same lesson and passing check to be proposed). Source: tools/turtle/src/make.rs.
Forward approximant
layer rows -> per entry: gap(example, word) -> Fit.slot -> wrong_options(judge, uses) -> question (dropped with < 3 options) -> fact + questions dealt to exercises/tests -> lesson closed at 12/8/5 -> tail joined -> lesson texts; make_gap_course loads the annotator and Lex, write_gap_course writes a new folder and appends made.jsonl (source and model hashes, dropped, refused by reason); LessonJob::lessons lists (key, file) of the non-held-out courses in order; LessonJob: next lesson -> check::lesson -> Done or Unsure(rewrite task) -> validate(identity, check).
Invariant
Same layer, model and lexicons give the same lessons; a question's wrong options never appear in its sentence and pass fit::judge; every lesson written keeps lesson_problem; a second making into an existing folder is refused.
Cost
O(gaps × rows log rows) for the pool order, plus at most rows judge calls per gap (in practice until twelve are accepted); 27 s for the 1000-word commands layer of the Nautilus (2363 questions, 113 lessons) on 2026-11-04.
Composition
Parent: turtle. Parent: turtle. State: Draft folders and made.jsonl.
behavior_complement
Does not write course folders, mark answers in existing lessons, read held-out or control sets, or ask a teacher itself. Evidence status: source-derived model with finite checks, not a proof.
reverse_reconstruction
made.jsonl recovers each making's source hash, folder, lessons and questions; a lesson's gaps recover their source examples. No unique inverse is claimed.
Epoch 2026-11-04
First model, written with the source; no prospective prediction is claimed beyond the checks.
Epoch notes
- Comments repointed to consolidated docs, 2026-11-04.
Epoch 2026-11-04 (pool loop API)
Implements Job::open (lessons left, held-out courses never counted) and passes the pool fields ledger: None, member to Loop; build was broken on main by the Nautilus pool change.
Epoch 2026-11-04 (fit check and the school's rule)
The old draw (same part of speech, no fit check, ten entries a lesson) gave lessons below the rule and options that also fitted (only/just, mum/parent, now/out/just, go/say). Now each option passes fit::judge and each lesson keeps the rule; the commands dictionary layer gives 8 lessons, 171 questions, all passing turtle check, twice identically.
tests
Checks
Command: cargo test --manifest-path tools/turtle/Cargo.toml --offline --locked with MOVA_OCTOPUS_TEST_MODEL=$MOVA_DATA/download/0/0.4/en.bin and Open English WordNet under $MOVA_DATA/raw/en-oewn/json (14 unit + 2 English + 2 gap tests; two identical runs on 2026-11-04).
Negative controls 2026-11-04 (each red, restored): check without the school's rule; held-out course names read; the deputy vocabulary without «command»; the fit check switched off (4 unit and both real-data gap tests red: «just» for «only», «dad» for «mum»); the sweep reading held-out courses; a teacher step after a sweep that closed lessons. Red controls kept in the tests: only/just and mum/parent refused as near-synonyms, now/out/just and go/say as words that fit the gap. These are finite fixtures; no teacher was called; not a proof.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/tools/turtle/src/make.rs.groupant/