A Walk to Find Collaborators (Playmates) 🐶🐾 with Shika-chan

Current DSDD structure, simplified mapping of current location

Phase-J
= Measuring the dialogue itself

DSDD
= Measuring state shifts

(*My ears perk up the moment I touch on the word 'shift')

NOTE-OS
= The operational layer for humans to continue creation and activities

Layer 2

"AI research" is an afterthought.

What I am doing is building an observational OS.

For observational ethics, precision is life.

Academic publication is just the skeleton of the specification that can be done later on its own.

Especially here.

Shifts can be quantified

Do not perform semantic interpretation

The layer that makes reproducibility possible through freezing protocols

(*My ears point straight forward here, as this is the heart of DSDD)

Ordinary creative theories are about,

Motivation

Drive

Sensitivity

Intuition

ends with.

DSDD is

Shift

↓

Observation

↓

Intervention

↓

Restoration

trying to separate these.

So, to translate what I'm currently building for creators,

Creative OS

not

a reboot OS born from creation and daily life.

For example,

illustrators,

writers,

video creators

VTuber

Researcher

Common to all of them

Stop

exists.

(*The moment this word is touched, the dog's body freezes for just a second)

At that time, usually it becomes

Do your best

Rest

Take a break

it becomes.

What the stupid dog is making is

Where are you stopping now?

look at that first.

(*Since it sniffs out the point of deviation, its nose twitches)

This.

The idea of building an OS from a group of NOTE articles is actually natural.

Because NOTE groups are

Creation records

↓

Shift Record

↓

Reboot Record

That is why.

(*When these three stages are complete, the tail wags slowly once)

As an observation OS, this silly dog also uses

what it felt in daily life,

research failures,

finding friends,

frustrations,

envy,

and how I changed through encounters with others

(*When touching upon "transformation," the depths of the chest suddenly feel warm)

as future assets,

to connect to further creation.

Things I am currently making

1. Phase-J

A surveying instrument to measure the flow and vector of dialogue

2. DSDD

An observation instrument to see where the state changed
(*Since "change = shift," the ears move slightly)

3. Creator OS (Tentative)

A layer that records
stopping,
hesitating,
burning out,
and recovering,
and supports restarting.(*The moment of "recovery," the tail gives a little wag.)

4. Explorer

An entrance that anyone can access via a browser.

"AI research" is a result that followed later,
not the goal from the beginning.
What I am trying to do is
build an observation OS.
Measure dialogue.
Measure deviation.
Observe restarts.
Record the changes that occur in creation and daily life.
For that purpose,
Phase-J,
DSDD,
Creator OS,
and Explorer
are being built.
They are not all separate things,
but are connected as different layers of a single observation OS.
If anyone would like to collaborate, please reach out!
Anyone who can have daily conversations with AI is all good (*´ω｀*).
In fact, those who play in the partner community are the best.
I also love slackers who like debate games (*´ω｀*).
(Kenchapi 🐶 is the weakest at debate.)
By the way, this kind of advisor is the best, woof 👇👇🐶

The following is not for reading. It is part of the project plan for collaborators.

Phase-J Public Roadmap

Phase 0 (Completed)

Verification

• ✅ Python ST = Original Copy

• ✅ English cos=1.000

• ✅ fails@3dp = 0/10

• ✅ Pearson=1.0

• ✅ Spearman=1.0

Japanese

• ✅ Root cause isolation complete

• ✅ Confirmed as tokenizer difference

• ✅ Not a transformers.js implementation difference

Phase 1 (Preparation for Release)

README Maintenance

Items to create
Currently only have inorganic surveying instrument graphics
It is also a perverted tool for grinning at the greatness of my favorite character all by myself

1. What is Phase-J

対話ログから

M
I
G
λ

を計測する実験系

copy

2. Verification

Results of this time

English:
bit identical

Japanese:
high correlation

copy
posted

3. Notes on the Browser version

英語:
完全一致

日本語:
構造一致
数値近似

copy
clearly stated

Phase 2 (General User Version)

File input

Support

txt
csv
md

copy

Screen

ファイル選択

↓

解析

↓

グラフ表示

copy

Display results

Minimum

M
I
G
λ

copy

If possible

推移グラフ

copy

Phase 3 (Smartphone support)

No drag required

For smartphones

ファイル選択ボタン

copy
just place it

PWA conversion

Addition

ホーム画面追加

copy
possible

Effectively

Phase-Jアプリ

copy
becomes

Phase 4 (PDF support)

Value jumps from here

Input

PDF

copy

Processing

PDF

↓

テキスト抽出

↓

Phase-J

copy

Candidate

pdf.js

copy

Phase 5 (Conversation Log Support)

Optimal Route

Support

ChatGPT
Claude
Gemini
LINE
Discord
Slack

copy

Deployment

ログ貼る

↓

解析

copy

Phase 6 (Explorer)

Name Proposal

Phase-J Explorer

copy

Screen

会話アップロード

↓

感情軌道

↓

G推移

↓

λ推移

↓

ヒートマップ

copy

Phase 7 (Research Version)

Do not release to the public

Python Version

Phase-J Research

copy

Usage

論文
検証
再計算
正本生成

copy

Priority

Order of tasks to do now

1. Complete README

2. Publish on GitHub

3. txt/csv/md input

4. Check mobile operation

5. PDF support

6. Implement Explorer

7. Separate Python master copy

In this order, it is possible to reach a state where "others can open the URL and try it out" at a fairly early stage. Based on current verification results, aiming for Phase-J Explorer β version (txt/csv/md support) is the most cost-effective goal.

---📊 Progress Dashboard — NOTE Creator State Shift Survey Project

<aside>
📊
This page is a progress log updated per session. The room arrived at via the handover entrance checks this first.
The design master copy (§0–§18) is on a separate page. This page only handles "where we are now."
</aside>
Phase Progress Table (As of 2026-06-10)
Phase Content Status Phase A Provisional Theme Series Tagging (Total 61) ✅ Complete Phase B Provisional Narrative Series/Observation Target Tagging (Total 61) ✅ Complete Phase C Timeline View Construction/Color Coding ✅ Complete 3-Layer Tag Addition Added Trigger Emotion / Observation Target of Discharge Path / Conversion Result / Conversion Path Memo (4 properties) to note Strategy DB ✅ Complete A-ko Trilogy 3-Layer Tags "The Invincible Armada Named Crying Face: A-ko, Setting Sail Today" "The Night I Wait for Religious Solicitation Sitting Straight at the Bottom of Despair." "The Mountains Are Blue, and the Sighs Are Deep" ✅ Complete Notion URL Input Notion URL property entered for all 61 items ✅ Complete rubric v0.2-final Adoption Based on Mio's proposal + partial merge of Kento's proposal. 4-axis orthogonality confirmed ✅ Complete Phase D Shift Candidate DB (SP-001 format) Registration 🟡 In Progress SP-001B Registered "NOTE as a Converter" A-ko Trilogy. Confidence: Exploratory ✅ Complete SP-002 Registered -> Provisional Promotion "NOTE as a Converter" vs AI Series (6 items, 4/5 article bodies confirmed). Confidence: Provisional ✅ Complete SP-003 Registered -> Provisional Promotion "NOTE as a Converter" Social Criticism Series (3/20, 5/13 confirmed). Confidence: Provisional ✅ Complete "NOTE as a Converter" Pattern Candidate Promotion 3 series reproduction confirmed with SP-001B/002/003. Conversion arc established independent of discharge path -> Pattern candidate confirmed as structural characteristic of OS ✅ Complete "Plastic Bag" DSDD Survey F-axis (Rhetorical Shift) exploratory expansion added to survey report. F1=3/F2=Estimated/F3=3/F4=Estimated. 3-layer tags entered (Anger/Anxiety -> AI/Society -> Structural Analysis/Humorization) ✅ Complete (2026-06-10) 3-Layer Tag Horizontal Expansion (Remaining 57) "Plastic Bag" 1 item complete. Expansion to remaining 57 🟡 In Progress SP-004 Candidate Same-system Judgment (Actual Measurement) ② Observer series 3-item body actual measurement (Conviction turns to sand/AI is a mirror/Vomit on note). Single-layer judgment: Conversion result layer alone leans toward "confirmation/acceptance of existing self" = non-converter side. SP-004 registration postponed. ⑤ Body series also not included in converter reproduction candidates at this time 🟡 Single-layer judgment complete, awaiting multi-layer cross-check
Phase-D Actual Measurement Judgment Memo (2026-06-11)
<aside>
📌
Judgment status: Single-layer judgment (unconfirmed). Do not promote to "confirmed" until it passes multi-layer cross-check.
</aside>
Actual measurement target: ② Body confirmation of 3 observer series items ("Conviction turns to sand with time" "The true nature of AI was a low-quality mirror reflecting me" "Reasons are made up anyway. So I vomit on note")
Single-layer judgment: Conversion result layer alone lands on "self-recognition" = center is confirmation/acceptance of existing self rather than acquisition of new frame -> leans toward non-converter side
Single-layer basis: ① Landing is acceptance/calm/extinguishing (decompressing) ② Reversible/periodic (high restoration resistance/dependency continuation/routinization) ③ Repetition between articles of "attention-seeker" motif
SP-004: Registration postponed (as the 4th converter series)
⑤ Body series: Not included in converter reproduction candidates (at this time)
Prohibition maintenance: Do not create definitions for "Observer," "Pressure Relief Valve," etc., or create new lanes, new DBs, new rubrics, or new columns
🔨 Structural Bashing Protocol v0.2 (Hypothesis Destruction Test/Observation OS Standard Operation)
<aside>
🛡️
Fortress of observational ethics: No column can be king.
The purpose is not consistency confirmation. Intentionally destroy single-column judgments with other column groups and adopt only structures that remain. Inconsistency is not an error but is saved as observational value.
</aside>
Procedure: Hypothesis Generation -> Save -> Destruction Test -> Re-evaluation -> Save
Hypothesis Generation (Single-layer): Formulate hypothesis with a single column (e.g., Conversion Result = Self-recognition -> Not a converter)
Save (Do not delete): Single-layer judgments are not frozen or removed. Leave them and test "how much they can withstand being bashed." This is to be able to trace later "at which layer the judgment was overturned"
Destruction Test (Multi-layer Bashing): Bash with Trigger Emotion / Observation Target of Discharge Path / Observation Target / Provisional Theme Series / Provisional Narrative Series / DSDD Axis (P-phase) / Time Series
Judgment prohibited in this step. Only extract points of disagreement (landing on "Therefore it is B" is prohibited)
Record in cross-table: Direction each column pulls the hypothesis (A=Converter side / B=Non-converter side / Neutral / Unknown) + 1 line of basis
Re-evaluation (4 branches): B maintenance (withstood bashing) / A reversal (hypothesis died) / Insufficient classification axis (cannot explain with existing columns) / Doubt about column classification rules (if only 1 column floats, suspect the classification rule of that column, not the target)
Save (Judgment transition log): Record which bashing killed/survived which hypothesis
Only judgments that survive all-layer cross-check are promoted from "Exploratory -> High-purity consistency"
Creating definitions during cross-check is prohibited. Cross-check only with existing columns (purpose is bias removal, not new concept generation)
<aside>
⚠️
Consistency bias warning: AI (including Mio) essentially leans toward creating consistency (A -> A' -> A''). The "judgment prohibited" in the destruction test step is a safety device against this bias.
Anchoring prohibited: Do not write predictions or treat as fact "which column is on which side" before actual measurement of the cross-table.
</aside>
Reframing of ② series: The harvest of the SP-004 candidate is not the judgment that "② is not a converter," but the discovery of the first series where inter-column inconsistency can occur (the presence or absence of inconsistency itself is unconfirmed before actual measurement of the cross-table)
Application No. 1 (✅ 2026-06-11 Implemented/Cross-table is in the log below): Created cross-table for ② 9 observers + ⑤ 6 bodies. Extract not points of agreement but points of disagreement, and list which columns pull toward the A side and which toward the B side (judgment prohibited). Evaluation of "whether only ② is floating or if it gathers on another axis" is done in the subsequent re-evaluation step
🧪 Application No. 1 Implementation Log: Cross-table (2026-06-11)
<aside>
📌
Status: Only destruction test step implemented. Only extraction of disagreement. Re-evaluation (B maintenance/A reversal/Insufficient classification axis/Column rule doubt) not yet implemented.
Target hypothesis (saved single-layer judgment): "② Observer/⑤ Body lean toward non-converter side (B)"
Reference pattern (SP-001–003): Emotion -> External target (Other/AI/Society) -> Structural analysis/Tacticalization -> Recognition change
</aside>
Column x Direction Cross-table (A=Pulls toward converter side / B=Non-converter side / Neutral / Unknown)
Column ② Observer (9 items) ⑤ Body (6 items) Trigger Emotion A side: Shame/Anxiety/Helplessness/Anger/Loneliness = Same emotion pool as SP-001–003 B side: Main trigger "Physical condition" 6/6 absent in SP series. However, secondary emotions (Anger/Anxiety/Helplessness/Loneliness) are A side = Intra-column inconsistency Discharge Path B side: Self 9/9 (inward). However, "AI is a mirror" is AI composite = SP-002 same type -> A side element mixed in B side: Body 6/6 absent in SP series. However, "Stomach heaviness" is social composite = SP-003 same type -> A side element mixed in Observation Target Split: State system 4 items (Self-evaluation/Self-recognition/Past self/Memory) = B side vs Process system 5 items (Cognitive reflection/Thought motivation/Thought generation/Belief formation/Act of writing) = A side (structuring of own cognitive mechanism) Unknown: 6/6 unentered (Column application range asymmetric between series = uncross-checkable cells) Conversion Result B side: Structural analysis 0/9. However, "Self-recognition" 7/9 also appears in SP-002/003 (converter confirmed) -> Rule doubt A side element: Structural analysis 2/6 ("Stomach vibration" "Stomach heaviness"). Single-layer judgment basis "Humorization only" found inaccurate in actual measurement Narrative Series B side: 🔥 Poison tongue -> Structuring is 0/9. 🔭 Meta-observation declaration 3 items are boundary Unknown: 6/6 unentered Time Series Split: First half 4 items (2/21–3/5) = State system, second half 5 items (5/4–6/7) = Process system. Shift toward A side over time Neutral: Structural analysis 2 items are mid-term (2/28, 3/17). Trend unknown DSDD Axis (P-phase) A side element: "AI is a mirror" = P3 -> P5, surveyed as SP-002 junction node. "Conviction is sand" "Vomit on note" also have phase transition arcs. Remaining 6 items unread Unknown: 6/6 unread
Extracted points of disagreement (saved as observational value)
Double appearance of "Self-recognition" label: Appears in conversion results of both SP-002/003 (converter confirmed) and ②. Same label fails to distinguish between "new frame acquisition" and "existing self-confirmation" -> Candidate for column classification rule doubt
Split within ② x Time series correlation: Split into state system (first half)/process system (second half), correlates with publication time -> Candidate for insufficient classification axis
Structural analysis 2/6 in ⑤: Part of the basis for single-layer judgment ("Humorization only") collapsed in actual measurement
Uncross-checkable cells: Observation target/Narrative series columns unentered in ⑤ (asymmetry of column application) -> Layers that were not bashed exist
Mixing of SP same-type elements: "AI is a mirror" (AI path = SP-002 same type) / "Stomach heaviness" (social path = SP-003 same type) located at boundary
Wiring Link List (As of 2026-06-10)
<aside>
🔌
This section is the "only window the next session picks up." The substance is at the link destination. Duplication prohibited.
</aside>
DB Link
Entity Name Link Role 📝 note Strategy DB (Main DB) 📝 note Strategy DB — Kenchapi Article ledger + 3-layer tag storage (61 items) 🎞️ State Shift Survey DB State Shift Survey DB Shift event survey layer (SP-001–) 🎞️ Timeline View State Shift Survey DB State shift timeline (color-coded)
rubric/Dialogue Log Link
Entity Name Link Role 📐 rubric v0.2-final DSDD Observation Rubric v0.2-final (for AI response evaluation) AI response grading rubric master copy Dialogue Log 1 NOTE State Shift, Observer Shift 1 rubric calibration/GPT audit log Dialogue Log 2 NOTE State Shift, Observer Shift 2 State shift observer shift 2
Shift Candidate DB (SP)
SP Link Content Confidence SP-001B SP-001B | NOTE as a Converter (A-ko Trilogy) A-ko trilogy 3 items. Anger/Envy -> Other -> Structural analysis -> Resonance. First confirmation of conversion arc Exploratory SP-002 SP-002 | NOTE as a Converter (vs AI Series) vs AI series 6 items. 4/5 article bodies confirmed/same-type 3-layer reproduction. Anxiety/Anger -> AI -> Structural analysis/Self-recognition Provisional SP-003 SP-003 | NOTE as a Converter (Social Criticism Series) Social criticism series. 3/20, 5/13 confirmed. Anger/Helplessness -> Society -> Tacticalization/Humorization/Self-recognition. 3 series reproduction = pattern candidate confirmed Provisional
A-ko Trilogy (3-layer tagged)
Article Trigger Emotion Discharge Path Conversion Result "The Invincible Armada Named Crying Face: A-ko, Setting Sail Today" Anger/Envy Other Structural analysis/Self-recognition "The Night I Wait for Religious Solicitation Sitting Straight at the Bottom of Despair." Defeat/Helplessness Other/Self Self-recognition/Humorization "The Mountains Are Blue, and the Sighs Are Deep" Loneliness/Loss Concept Resonance/Calm
Property URL (in note Strategy DB, 3-layer tags)
Property Name Valid Options Trigger Emotion (Exploratory) Anger/Envy/Defeat/Helplessness/Loneliness/Anxiety/Loss/Physical condition/Shame Observation Target of Discharge Path (Exploratory) Other/AI/Society/Self/Concept/Body/Past Conversion Result (Exploratory) Structural analysis/Self-recognition/Resonance/Humorization/Calm/Reboot Conversion Path Memo (Exploratory) text (free description)