Layered Decomposition Brief

This brief supplements ralph-methodology.md with specific guidance on decomposing Japanese→English seeds into overlapping, layere…

thomascassidyzm updated 1mo ago
Claude CodeGeneric
View source ↗
# Layered Decomposition Brief — eng_for_jpn

> **This brief supplements ralph-methodology.md with specific guidance on decomposing Japanese→English seeds into overlapping, layered LEGOs.**

---

## The #1 Anti-Pattern: Giant Single M-LEGOs

**THIS IS WHAT BAD DECOMPOSITION LOOKS LIKE:**

SEED: "But I don't worry about making mistakes." BAD: L1 [M] "間違えることを心配しない" → "I don't worry about making mistakes" → ONE giant LEGO wrapping the entire sentence → Learner gets ONE opaque blob, ZERO recombination


**The learner can't do ANYTHING with this.** They can't say "I don't worry about the answer" or "making mistakes is good" because they never learned the parts.

---

## When to Use Overlapping LEGOs

**Overlapping LEGOs exist to reveal things the learner can't figure out from simple tiling.** Use them when there's something non-obvious happening between the small piece and the big piece. Don't use them when the composition is transparent.

### Overlap WHEN:

**1. Grammatical transformation (conjugation, morphology)**
The target language changes the word form when combining pieces:

戻る (to come back) → 戻りたい (want to come back) Stem changes: 戻る → 戻り + たい Without overlap: learner knows 戻る but can't produce 戻りたい With overlap: sees 戻る alone, then 戻りたい — infers the たい conjugation


**2. Word order differences**
The target language puts pieces in a different order than the known language:

English: "speak later on" (verb + time) Japanese: 後で話す (time + verb — REVERSED) Without overlap: learner might guess 話す後で With overlap: sees 後で alone, sees 話す alone, then sees 後で話す — infers the order


**3. Particle/glue insertion**
Pieces need connective tissue that isn't obvious:

間違える (mistakes) + 心配する (worry) → 間違えることを心配する Particle を and nominaliser こと appear between them Without overlap: learner has no way to know about こと or を With overlap: sees the connected form, infers the glue


### DON'T overlap when:

**The composition is transparent.** If the learner could correctly guess the combined form from knowing the pieces, an overlapping LEGO adds nothing — it's just boring repetition.

English: "with you now" → あなたと今 If learner knows "with you" → あなたと and "now" → 今 Combined form is just あなたと + 今 = obvious No hidden grammar, no reordering, no particles → DON'T create an overlapping LEGO for this


**The art is knowing the difference.** Japanese has a LOT of hidden grammar and reordering compared to English, so overlapping LEGOs will be common. But not universal.

---

## The Decomposition Algorithm (Revised)

For each seed, work **bottom-up**:

### Step 1: Identify New Concepts
What does this seed introduce that the learner hasn't seen?

### Step 2: Ask "What's Hidden?"
For each pair of concepts that combine in the seed, ask:
- Does the word form change? (conjugation → overlap)
- Does the order flip? (word order → overlap)
- Do particles/connectors appear? (glue → overlap)
- Or is it just concatenation in the expected order? (→ no overlap needed)

### Step 3: Build the LEGO Stack
Create LEGOs bottom-up. For every "hidden" combination, create an overlapping M-LEGO that shows the learner how the pieces actually combine.

### Step 4: The Recombination Test
For each LEGO: **"Can this piece appear naturally in 3+ other sentences?"**
If not, it's too big or too specific. Redraw boundaries.

---

## Worked Example: Seed 16

SEED 16: 彼は後でみんなと一緒に戻りたいです。 Target: He wants to come back with everyone else later on.


### Analysis — What's Hidden?

| Combination | What happens? | Overlap? |
|---|---|---|
| 戻る → 戻りたい | Verb stem changes for desire form | YES — conjugation |
| 後で + 戻りたい | Time goes BEFORE verb (English: after) | YES — word order |
| 彼は + 戻りたい | Subject marked with は particle | YES — particle |
| みんなと一緒に + 戻りたい | Phrase order reversal from English | YES — word order |

### Resulting Decomposition:

L1 [A] "後で" → "later on" Standalone time word, recombines freely

L2 [M] "戻りたい" → "to come back" Verb in desire form. Recombines: "I want to come back", "come back tomorrow"

L3 [M] "後で戻りたい" → "to come back later on" ← OVERLAPPING: contains L1 + L2 ← REVEALS: time-before-verb order (English says "come back later", Japanese says "later come back")

L4 [M] "みんなと一緒に" → "with everyone else" Particle と absorbed. Recombines freely.

L5 [M] "彼は" → "he wants" Subject + は particle. Recombines: "he wants to speak", "he wants to learn"

L6 [M] "彼は戻りたい" → "he wants to come back" ← OVERLAPPING: contains L2 + L5 ← REVEALS: how subject attaches to verb (は particle placement)


**6 LEGOs, 3 overlapping pairs.** The learner sees 後で alone, 戻りたい alone, then 後で戻りたい — and infers that Japanese puts time BEFORE the verb. That's the teaching.

---

## Japanese-Specific Patterns

### Verb Conjugation = Always Overlap

Japanese verb conjugations are non-obvious to English speakers. Whenever a verb changes form, the learner needs to see the base form AND the conjugated form:

戻る → 戻りたい (desire: stem + たい) やめる → やめたくない (negative desire: stem + たくない) 話す → 話している (progressive: stem + ている) 覚える → 覚えられる (potential: stem + られる)


Each of these is an opportunity for an overlapping LEGO that reveals the conjugation pattern.

### Word Order = Usually Overlap

Japanese is SOV, English is SVO. Almost any multi-word combination will have a different order:
- "speak English" = 英語を話す (object before verb)
- "come back tomorrow" = 明日戻る (time before verb)
- "speak with everyone" = みんなと話す (companion before verb)

These order differences are worth revealing through overlapping LEGOs, especially in early seeds when the learner hasn't yet internalised Japanese word order.

### Particles = Always Bundle

は, が, を, に, で, と are never standalone LEGOs. Always absorbed into M-LEGOs that show them in context.

---

## Don't Overdo It

**Not every seed needs maximum overlap.** The goal is to reveal patterns, not to bore the learner with redundant drilling when the co

Maintain Layered Decomposition Brief?

Let people know it's listed here — add the badge (live metrics, light/dark aware) or a plain link to your README or docs.

[Layered Decomposition Brief on getagentictools](https://getagentictools.com/loops/thomascassidyzm-layered-decomposition-brief-eng-for-jpn?ref=badge)
npx agentictools info loops/thomascassidyzm-layered-decomposition-brief-eng-for-jpn

The second line is the CLI lookup for this page — handy in READMEs and docs.