---
id: jp-learning-and-anki-tooling/09-ranking-and-staged-plan
title: 'Project status: Japanese immersion learning with Anki and selective glossing:
  9. Ranking and staged plan'
report: https://ai.hiina.space/reports/jp-learning-and-anki-tooling.md
section: 9
version: 1
updated: '2026-10-27'
author: Hiina
---

Index of everything on this site: https://ai.hiina.space/index.md

# Project status: Japanese immersion learning with Anki and selective glossing: 9. Ranking and staged plan

- Section 9 of 12 of the report "Project status: Japanese immersion learning with Anki and selective glossing" by Hiina (v1, updated 2026-10-27)
- Report overview and contents: https://ai.hiina.space/reports/jp-learning-and-anki-tooling.md
- Previous: https://ai.hiina.space/reports/jp-learning-and-anki-tooling/08-planned-tool-the-shared-learner-model.md
- Next: https://ai.hiina.space/reports/jp-learning-and-anki-tooling/10-open-questions-and-evaluation-plan.md

## 9. Ranking and staged plan

| Rank | Idea | Effort | Expected payoff | Evidence behind the mechanism | Depends on |
|---|---|---|---|---|---|
| 1 | Anki settings plus manual leech/Click-card pass | S | High, immediate | Manual and guides; varied-cue RCTs | nothing |
| 2 | (e) learner model v0 | S | Enabling | FSRS/HLR lineage | Anki, ankimcp |
| 3 | (a) deck-hygiene agent and Jiten-based premine | S-M | High | Varied cues; retrieval beats generation | 1, 2 |
| 4 | (d) selective-gloss overlay | M | High | Glossing meta-analysis; hint-first RCT (other domain) | 2 |
| 5 | (c2) Japanese-page gloss mode | S after 4 | Medium-high | Same as (d) | 2, 4 |
| 6 | (b) concept tutor revival | M-L | Medium | Hint-first tutoring RCTs (maths/physics) | 2, corpus from 3 |
| 7 | (c1) English-to-hybrid weave | L | Low-medium | Small code-switching studies; Storyfier caution | 2 |

Rationale for the order:

- **Stage 1 first** because it changes the daily experience immediately at no build cost. The settings and the Click-card/leech pass address two to-dos by hand, and the varied-cue result is the strongest evidence in the plan.
- **Learner model second**, as a thin read-only service. It is small, and without it every later tool repeats the known-word plumbing that the premine script, Jiten and asbplayer each already rebuild.
- **Hygiene agent and Jiten-based premine third**, because they finish the third to-do and produce the corpus of real sentences that the tutor and the overlay both need.
- **Overlay above tutor**, because it sits inside the activity the learner already enjoys; the engagement evidence says tools that need a separate habit decay. It also yields the richest learner-model signal, and its Japanese-page mode covers the blog-reading goal almost for free.
- **Concept tutor sixth**, worth reviving only once it can draw on vocabulary state and the learner's own corpus, which fixes what stalled it in 2025.
- **English-to-hybrid weave last**: weakest evidence, hardest Japanese-specific engineering, and a paid baseline (Nuenki) that can be trialled in an afternoon.

---

Text licensed CC BY 4.0 (attribution: Hiina, ai.hiina.space). Source: https://github.com/hiinaspace/ai-hiina-space