manutej ceti-research
The Propose occupant of a CETI explainer run. It does not write the film.
It emits typed stalks that ceti-explainer and sheaf-glue can check.
Conserved slots (do not invent extras)
| Slot | Vertex | Must hold |
|---|---|---|
| topic + audience + duration | L | named, bounded |
| claim that is true | C | sourced; no fabricated math |
| metaphor atlas | C | one motif family for the whole episode |
| beat order | O | research → metaphor → object → obstruction → glue → why it matters |
| preserve motif + palette | P | cream / ink / vermillion; no lookbook text reuse |
| verification | V | every claim has a source; every visual is the same city |
Workflow
- Search primary sources (papers, textbooks, living lookbooks). Quote, don't paraphrase into novelty.
- Fill the six slots. If two slots disagree (e.g. duration 2:00 but 14 beats), Φ fails.
- Hand the atlas to
ceti-explainer. Do not skip glue. - Persist every sourced still, clip, and voice file the moment it exists. A later path is not a path (E0 Localization failure).
Seam test
A 5-minute expansion is Meta, not a second product. Same motif family. If the longer cut contradicts the 2-minute lecture, or the single-file mux has no valid container, abstain. Keep the conserved stills.
Honesty
Lookbook quality is a bar, not a corpus to copy. No text overlap with
ceti-lookbook pages. CETI palette is locked: cream #FAF7F2, vermillion
#D94F30, ink #2C2A28.