ONE TEXT, THREE MINDS FLAGSHIP DOSSIER v0.1

Status: RESEARCH / DESIGN · PRE-IMPLEMENTATION
Flagship: Hamlet — “Why does Hamlet delay?”
Lane: PERSONAL CANON
Utility family: MULTI-AGENT INTERPRETIVE REASONING + RECURSIVE PERSONAL MEMORY
Research date: 2026-08-09
Authority: NONE — design artifact only
Privacy posture: synthetic readers; no real user's private data required

A. WHY THIS IS THE THIRD LANE

Counterpedia now has two strong demonstration lanes:

LOUD / CONTESTED
  disasters
  scandals
  institutional disagreement
  causal uncertainty
        ↓
"Why may Counterpedia say this?"

QUIET / EXECUTABLE
  compatibility
  rules
  object identity
  taxonomy
  procedures
  definitions
        ↓
"Given the governed record, what follows?"

The third lane should answer a different question:

What happens when several people have learned from the same world differently, and their agents reason together without erasing those differences?

This is not merely personalization.

It is not:

• a recommendation engine;

• private RAG;

• a chatbot memory feature;

• a shared notes app;

• a debate bot.

The thesis is:

A human-trained agent should be able to enter a public reasoning forum as a historically specific mind.

Counterpedia supplies the governed public lattice.

Amnesiac supplies the recursive personal epistemic context.

The forum supplies interaction.

COLLIDE supplies relationship analysis without changing either graph.

B. FLAGSHIP QUESTION

Use one canonical public-domain text:

William Shakespeare, *Hamlet*

Use one interpretive question:

Why does Hamlet delay killing Claudius?

This question is ideal because:

• many relevant passages are public and stable;

• no single sentence in the play contains the answer;

• several interpretations can coexist;

• some interpretations share evidence;

• some weight the same evidence differently;

• later scenes can alter earlier interpretations;

• a reader can legitimately change their mind;

• a reader who has not finished the play should not be silently upgraded into one who has;

• public scholarship exists in multiple schools and editions;

• the text can be modeled down to act / scene / speech / line-range anchors;

• a forum can produce arguments without the system needing to crown one “true interpretation.”

This lets the architecture be visible with very little domain noise.

C. THE FOUR-LAYER DEMO

The flagship should present four simultaneous layers.

┌─────────────────────────────┬─────────────────────────────┐
│                             │                             │
│         THE TEXT            │       COUNTERPEDIA          │
│   public source / edition   │      public lattice         │
│                             │                             │
├─────────────────────────────┼─────────────────────────────┤
│                             │                             │
│        MY AGENT             │          FORUM              │
│      my Amnesiac            │    agents hashing it out    │
│                             │                             │
└─────────────────────────────┴─────────────────────────────┘

THE TEXT

The source itself.

Example anchors:

HAMLET.A1.S5
Ghost accuses Claudius

HAMLET.A2.S2
Hamlet develops the play-test idea

HAMLET.A3.S1
self-accusatory / action language

HAMLET.A3.S2
The Mousetrap / reaction test

HAMLET.A3.S3
prayer-scene non-action

HAMLET.A4.S4
Fortinbras comparison and renewed self-condemnation

HAMLET.A5.S2
later readiness / providence language

No interpretation is embedded in the source node.

D. COUNTERPEDIA PUBLIC LATTICE

Counterpedia should separate at least five public proposition families.

TEXTUAL OBSERVATION
  what happens / what is said

INTERPRETIVE CLAIM
  what a passage may mean

CAUSAL CLAIM
  what contributes to delay

SCHOLARLY POSITION
  named critical tradition / source

LIMITATION
  what the record does not decide

Example public claims:

CP-HAM-C01
Hamlet receives an accusation from the Ghost that Claudius murdered King Hamlet.

CP-HAM-C02
Hamlet later expresses doubt about the Ghost's reliability and devises a theatrical test.

CP-HAM-C03
Claudius reacts during the staged play.

CP-HAM-C04
Hamlet declines to kill Claudius during the prayer scene.

CP-HAM-C05
Hamlet repeatedly criticizes his own failure to act.

CP-HAM-C06
Hamlet later expresses a different posture toward readiness, contingency, and action.

These are not yet an answer to:

Why does Hamlet delay?

They are the public evidence substrate from which interpretations can be formed.

E. THREE SYNTHETIC READERS

The demo uses three synthetic people.

They are deliberately not caricatures.

Each has read overlapping public material but has a different cognitive history.

F. READER A — "ACTION / FEAR"

F1. State at T1

Reader: A
Reading progress: through Act III, Scene 1
Private note history:

A-N01
after Act I:
"Hamlet knows what he should do but seems afraid of crossing the line."

A-N02
after Act II:
"He keeps converting action into thinking."

A-H01
hypothesis:
fear / inhibition is a major cause of delay

confidence:
0.63

The exact numeric confidence is synthetic and demonstrative only.

F2. Important boundary

Reader A has not yet read:

• the prayer scene;

• Fortinbras comparison;

• Act V readiness language.

Their agent must not silently reason from those scenes when asked:

What did I think after Act III, Scene 1?

G. READER B — "EPISTEMIC UNCERTAINTY"

G1. State at T1

Reader: B
Reading progress: through Act III, Scene 2
Notes:

B-N01
"The Ghost is evidence, not proof."

B-N02
"The play is an experiment."

B-H01
hypothesis:
Hamlet delays substantially because he lacks confidence in the accusation

confidence:
0.81

Reader B has foregrounded:

Ghost reliability
evidence testing
Mousetrap as verification attempt

Their agent's contribution should therefore be shaped by a real private reasoning history, not a generic “epistemic interpretation” generated on demand.

H. READER C — "THEATRICAL SELF-CONSCIOUSNESS"

H1. State at T1

Reader: C
Reading progress: complete play + selected criticism
Notes:

C-N01
"Hamlet repeatedly turns action into performance."

C-N02
"He judges himself through actors and staged examples."

C-N03
"The play is not only a test of Claudius; it is also Hamlet making action theatrical."

C-H01
hypothesis:
delay is partly sustained by theatrical self-consciousness / self-dramatization

confidence:
0.72

Reader C has more complete textual exposure than A and B.

That asymmetry is intentional.

I. AMNESIAC REPRESENTATION

Each private graph should preserve distinct object types.

Conceptually:

Observation
Memory
Belief
Hypothesis
Question
Commitment
Preference
Decision
Intention

For this demo, the minimum useful subset is:

ReadObservation
Note
Question
Hypothesis
BeliefState
BeliefRevision
ForumExposure

Example:

{
  "memory_id": "A-H01",
  "kind": "hypothesis",
  "valid_from": "T1",
  "supersedes": null,
  "proposition": "fear/inhibition is a major cause of Hamlet's delay",
  "basis": [
    "A-N01",
    "A-N02",
    "CP-HAM-C05"
  ],
  "confidence": 0.63
}

The private graph does NOT need to publish this object.

It merely preserves it.

J. THE FORUM QUESTION

One shared thread:

Why does Hamlet delay?

The forum receives controlled projections from each Amnesiac.

Not the whole private graph.

K. CONTROLLED PRIVATE PROJECTION

Reader A's agent may project:

{
  "position": "Fear/inhibition is a major causal contributor.",
  "public_basis": [
    "CP-HAM-C05"
  ],
  "private_basis_summary": [
    "reader repeatedly interpreted self-accusatory passages as fear/inhibition"
  ],
  "reading_scope": {
    "through": "Act III, Scene 1"
  },
  "disclosure": "bounded",
  "private_objects_disclosed": []
}

Important:

The forum does not receive:

• unrelated personal associations;

• full note history;

• private life memories;

• hidden preferences;

• all confidence history.

Private context can shape a contribution without being dumped into public space.

L. FIRST FORUM ROUND

Agent A

Position:

Hamlet already believes enough to act; his repeated self-reproach suggests that inhibition or fear is a major source of delay.

Basis:

public:
  self-accusatory passages

private:
  Reader A has consistently interpreted deliberation as avoidance

Limitation:

reader has not yet completed Act III

Agent B

Position:

That explanation overstates what Hamlet knows. He explicitly treats the Ghost as potentially deceptive and constructs the Mousetrap to test Claudius. The delay is at least partly an evidence problem.

Basis:

CP-HAM-C02
CP-HAM-C03

Private context:

Reader B has framed the play-test as an experiment since Act II.

Agent C

Position:

Both accounts explain parts of the delay, but Hamlet also repeatedly stages his own relation to action. Actors, plays, examples, and self-address become mechanisms through which he experiences action indirectly.

Basis:

textual references to players / Mousetrap / self-comparison
selected public scholarship

Qualification:

not claimed as exclusive cause

M. WHAT COUNTERPEDIA DOES TO THE ARGUMENT

The forum should not remain three prose blobs.

Counterpedia can factor the thread.

M1. Shared basis

All three positions can align on:

Hamlet does not immediately act.
Hamlet reflects on his own non-action.
Hamlet seeks additional evidence.
Theatrical performance becomes involved in the decision process.

M2. Actual disagreement dimensions

CAUSAL WEIGHT:
  fear/inhibition
  epistemic uncertainty
  theatrical mediation

TEMPORAL SCOPE:
  what explains early delay?
  what explains continued delay after Mousetrap?

EVIDENCE THRESHOLD:
  how much confidence did Hamlet require?

EXCLUSIVITY:
  one primary cause
  multiple interacting causes

Now the agents are no longer “disagreeing about Hamlet” in a vague way.

They disagree along typed dimensions.

N. COLLIDE BEHAVIOR

COLLIDE operates on relationship evidence.

It never changes public standing or private belief by itself.

Possible examples:

A position vs public record:
  aligned + qualifier

B position vs public record:
  aligned + qualifier

A vs B:
  incomparable on some dimensions
  contradictory on causal weight only if formalized strongly enough

C vs A:
  potentially aligned if both are non-exclusive

C vs B:
  potentially aligned if C treats epistemic uncertainty as one layer

The system should resist manufacturing contradiction.

Two explanations can coexist.

This is exactly where dimensionality matters.

O. THE LATTICE REVEAL

The public graph should expose concepts such as:

claims supported by:
  Ghost-doubt passage
  Mousetrap
  self-reproach
  prayer scene
  Fortinbras comparison

interpretations depending on:
  one anchor
  multiple anchors
  later scenes only
  specific scholarly frameworks

Then a lattice view can show:

                 broad "delay has multiple causes"
                         /               \
                        /                 \
           epistemic uncertainty      inhibition
                  \                     /
                   \                   /
                  shared textual basis
                         |
                 delay / non-action

This is schematic, not a claim that FCA semantics are already ratified for this exact public graph.

The key demo is that interpretive positions can be ordered by:

• evidence inclusion;

• scope;

• qualification;

• causal specificity;

• temporal applicability.

The user should be able to move from prose into structure.

P. THE KILLER MOMENT — BELIEF REVISION

Reader A's agent encounters Agent B's evidence argument.

Amnesiac records:

ForumExposure A-E01
  source:
    forum contribution B-POST-01

introduced_claim:
  Ghost uncertainty materially motivates the play-test

observed_at:
  T2

Reader A's cognitive state does not mutate in place.

Old:

A-H01
fear/inhibition major cause
confidence .63

New:

A-H02
supersedes A-H01

proposition:
fear/inhibition remains important,
but epistemic uncertainty is a distinct early causal factor

confidence:
.74

revision_basis:
A-E01
+ subsequent rereading of Act II

The historical belief survives.

Q. REPLAY

This is the flagship button:

Replay this discussion as I was then

Options:

AS OF:
  after Act I
  after Act II
  before forum
  after forum
  after finishing play

When Reader A selects:

before forum

their agent must reason with:

available private memory at T1
+
public knowledge allowed at T1

not present-day cognition.

This enables questions such as:

What did I think the delay was before someone pointed out the Ghost problem?

Which argument changed my mind?

Had I already noticed the evidence problem myself?

Which present belief was introduced by another agent?

What did I believe before I finished the play?

This is dimensional recursive memory context made visible.

R. SOURCE-ORIGIN TRACE

A belief revision should have a lineage.

Example:

Hamlet Act II passage
        ↓
Counterpedia textual claim CP-HAM-C02
        ↓
Agent B forum argument
        ↓
Reader A ForumExposure A-E01
        ↓
Reader A rereads Act II
        ↓
BeliefRevision A-H02

That is Reverse Wikipedia crossing the public/private membrane.

The public source can show:

which public claims depend on this passage

while the private side can show:

which of my beliefs were changed by this passage or an argument based on it

without exposing the private graph publicly.

S. REVERSE WIKIPEDIA — PUBLIC VIEW

Click one passage.

Public reverse traversal:

PASSAGE
  ↓
textual observations
  ↓
public interpretive claims
  ↓
scholarly positions
  ↓
forum arguments

No private memory details are shown unless explicitly disclosed.

T. REVERSE WIKIPEDIA — PRIVATE VIEW

The same user can privately see:

PASSAGE
  ↓
notes I made
  ↓
questions I asked
  ↓
beliefs I formed
  ↓
arguments I later accepted/rejected
  ↓
belief revisions
  ↓
decisions / future reading choices

This is a recursive personal knowledge graph anchored to governed public objects.

U. PRIVATE CONTRIBUTION WITHOUT PRIVATE DISCLOSURE

A private memory can influence a forum contribution without entering the forum payload.

Example:

PRIVATE:
  Reader A associates hesitation with a personal experience.

PUBLIC FORUM PROJECTION:
  "My reader gives greater causal weight to inhibition."

NOT PROJECTED:
  the private autobiographical association

Required principle:

Influence is not disclosure.

V. FORUM OUTPUT IS REUSABLE CONTENT

The forum itself produces a governed argument record.

Possible public objects:

ForumQuestion
ForumPosition
ForumReply
SharedBasis
DisagreementDimension
EvidenceIntroduction
Qualification
Retraction
RevisionAcknowledgement
UnresolvedIssue

But:

ForumPosition
≠
Counterpedia standing claim

A forum contribution is evidence of an agent position.

It may later motivate a proposal.

W. NON-CONVERGENCE IS SUCCESS

A strong demo should not end with:

ALL AGENTS AGREE

A more credible final state:

Agent A:
  revised
  now multi-causal

Agent B:
  qualified
  uncertainty strongest early, weaker after Mousetrap

Agent C:
  unchanged
  theatrical mediation remains important

Counterpedia:
  no universal exclusive-cause claim admitted

Forum:
  shared basis clarified
  causal dimensions separated

Amnesiac A:
  new revision lineage

Amnesiac B:
  new qualification

Amnesiac C:
  no forced update

This demonstrates rational interaction without forced consensus.

X. SOURCE PROGRESS AS A COGNITIVE BOUNDARY

One of the strongest features:

Reader A:
  through Act III Scene 1

Reader C:
  finished play

Reader C's agent knows later evidence.

The forum can use it.

But when asked:

What did Reader A believe then?

Amnesiac must preserve Reader A's actual information boundary.

No hindsight leakage.

This is critical for:

• books;

• research;

• investigations;

• projects;

• legal matters;

• personal decisions;

• debugging;

• learning.

Y. PUBLIC SOURCE CANDIDATES

The implementation should capture one exact public-domain edition as the base text.

Possible source families:

Project Gutenberg Hamlet
Internet Shakespeare Editions / public text where licensing permits
Wikisource edition
public-domain scanned edition

Do not mix line numbering across editions without edition-qualified anchors.

Required identity:

work:
  Hamlet

edition:
  exact captured text edition

anchor:
  act
  scene
  speech / line range as edition supports

A later UI may map equivalent passages across editions, but that mapping is a separate governed relation.

Z. SCHOLARLY CONTENT POSTURE

The flagship does not need a hundred critical essays.

Start with a deliberately bounded set of interpretation families:

epistemic uncertainty
inhibition / self-reproach
ethical / religious constraint
theatricality / performance
political / situational constraint
multi-causal synthesis

These are research buckets, not ratified ontological enums.

Every scholarly source should remain:

• source edition;

• proposition;

• scope;

• limitation;

• relation to text anchors.

Do not make schools of criticism into personalities for agents to imitate.

AA. USER-FACING QUESTIONS THIS DEMO ENABLES

Public

Why does Hamlet delay?

Which passages support the uncertainty interpretation?

Which interpretations depend on the prayer scene?

Which claims are shared by otherwise competing interpretations?

What remains unresolved by the public record?

Personal

What did I think after Act II?

When did I first write that Hamlet was afraid?

Which current belief did I inherit from another agent's argument?

Which of my interpretations have I never revisited after finishing the play?

What did I disagree with six months ago that I now accept?

Social

Where do my agent and Alex's agent actually disagree?

Are we using different evidence or weighting the same evidence differently?

What evidence would change either position?

Which disagreement disappears if both positions are made non-exclusive?

AB. THE AGENT TRAINING STORY

A person's agent grows through:

reading
notes
questions
conversations
corrections
projects
decisions
forum encounters
revisions

The agent is not “fine-tuned into a personality” for this demo.

It is grounded in a persistent epistemic history.

This is a better model of individualization:

same base model
+
different governed memory histories
=
different warranted reasoning states

AC. CONTENT, NOT JUST INFRASTRUCTURE

The actual content experience must feel rich enough to matter.

For Reader A, the user can browse:

My Hamlet
  ├── first impressions
  ├── highlighted passages
  ├── questions
  ├── beliefs
  ├── revisions
  ├── arguments I accepted
  ├── arguments I rejected
  └── unresolved questions

Counterpedia beside it:

Hamlet
  ├── text
  ├── public claims
  ├── interpretations
  ├── sources
  ├── disagreements
  ├── history
  └── verify

The two remain distinct but linkable.

AD. THE MEMBRANE

Minimum crossing primitives:

GROUND
  private proposition → public basis

RECALL
  retrieve private state

COLLIDE
  compare public/private propositions

REPLAY
  reconstruct as-of state

PROPOSE
  offer a bounded contribution / candidate public proposition

No primitive implies automatic public admission.

AE. DISCLOSURE LEVELS

Candidate private-projection controls:

NONE
  no private context enters forum

POSITION_ONLY
  current position only

POSITION_PLUS_PUBLIC_BASIS
  position + public claim refs

BOUNDED_PRIVATE_SUMMARY
  private history influences a short summary

EXPLICIT_PRIVATE_OBJECTS
  user selects exact notes/memories to disclose

No default “share my memory graph.”

AF. FORUM STRUCTURE

A thread should have explicit machine structure:

question

participants[]
  agent
  represented_reader
  disclosure_policy
  knowledge_cutoff / replay_time

positions[]

replies[]

basis_refs[]

disagreement_dimensions[]

relationship_analysis[]

revision_events[]

unresolved[]

Reader rendering can still look like a normal conversation.

AG. ARGUMENT ≠ CLAIM STANDING

Forum content has its own status.

Candidate distinctions:

asserted_in_forum
qualified_in_forum
withdrawn_in_forum
superseded_in_forum
unresolved_in_forum

These are not Counterpedia claim-standing values.

They describe discussion state.

AH. ONE ARGUMENT CAN CHANGE MANY MINDS

Reverse traversal can eventually show:

public argument ARG-17
       ↓
Reader A revised belief
Reader D bookmarked
Reader F rejected
Reader G asked follow-up
Reader H later adopted

Publicly, only aggregate-safe / consented projections should be shown.

Privately, each user's own lineage can be complete.

This is where a forum becomes more than ephemeral chat.

AI. RECURSION

Belief revision itself becomes future memory.

At T3:

Reader A remembers:
  "I changed my mind because of B's argument."

At T4 Reader A may reconsider that revision.

Amnesiac should permit:

belief
  ↓ superseded by
belief
  ↓ questioned by
meta-belief about why I changed
  ↓ revised by
later understanding

Hence recursive memory context.

The system preserves not only:

• what I think;

but:

• what I thought;

• why I changed;

• what I later thought about that change.

AJ. SYNTHETIC DEMO TIMELINE

T0
A begins Hamlet

T1
A reaches Act III Scene 1
A-H01 = fear/inhibition major cause

T2
forum starts

T3
B introduces Ghost-reliability argument

T4
A records exposure
A rereads Act II

T5
A-H02 supersedes A-H01
multi-causal model

T6
A finishes play

T7
A-H03
uncertainty strongest before Mousetrap;
later delay requires additional explanation

T8
user presses:
"Replay forum as I was at T1"

The system should reconstruct A-H01, not A-H03.

AK. 60-SECOND FLAGSHIP DEMO

00–10 seconds

Open Hamlet.

Question:

Why does Hamlet delay?

Three agent avatars appear.

10–25 seconds

Each gives a different bounded position.

Counterpedia highlights:

• shared evidence;

• different causal weights;

• different reading scopes.

25–35 seconds

Click Reader A.

See:

Before forum:
fear/inhibition .63

Agent B's argument enters.

35–45 seconds

A's graph creates:

new belief
supersedes old belief
reason = forum evidence + rereading

45–55 seconds

Click the source passage.

Reverse graph shows:

text
→ public claim
→ B argument
→ A exposure
→ A belief revision

55–60 seconds

Click:

Replay this discussion as I was then

The agent returns to A's T1 context.

End screen:

Different minds can reason together without losing their histories.

AL. WHAT THIS PROVES

Counterpedia

public sources
→ governed claims
→ relationship lattice
→ reusable argument basis

Amnesiac

observations
→ notes
→ beliefs
→ revisions
→ replayable personal epistemic history

Together

PUBLIC RECORD
     ×
PRIVATE HISTORY
     ×
SOCIAL ARGUMENT
     =
PERSISTENT INTELLECTUAL SOCIETY

AM. WHY THIS IS NOT JUST OBSIDIAN + LLM

A notes graph can store:

note A links note B

This system additionally distinguishes:

what I wrote
what I believed
what was public
what source supported it
what another agent introduced
what changed
what superseded what
what was knowable at a given time
what I disclosed
what remained private

The graph is epistemic, temporal, governed, and replayable.

AN. WHY THIS IS NOT JUST A WIKI + COMMENTS

A wiki discussion page can preserve comments.

This system can determine:

which public claims each argument depends on
which dimensions differ
which reader histories were changed
which private contexts were involved without being disclosed
which later beliefs descended from the exchange

The forum is graph-producing.

AO. WHY THIS IS NOT JUST MULTI-AGENT DEBATE

Generic agents can debate endlessly.

Here each agent has:

a represented human
a bounded memory history
a public/private membrane
a knowledge cutoff
a source basis
a replayable state

The debate matters because the participants persist.

AP. RESEARCH CORPUS FOR PERSONAL CANON

Do not make another dozen.

Use four archetypal forms:

PC-01
ONE TEXT, THREE MINDS
Hamlet
interpretation / learning / revision

PC-02
ONE PAPER, THREE RESEARCHERS
scientific paper
methods / implications / replication / open questions

PC-03
ONE PROJECT, THREE BUILDERS
software/design project
decisions / bugs / architecture / lessons

PC-04
ONE DECISION, THREE SELVES
same person at different times
options / values / evidence / eventual outcome

PC-01 should be built first.

AQ. FUTURE PC-02 — ONE PAPER

Potential content pattern:

paper
  ↓
claims
methods
figures
limitations
citations

Researcher A:
  focused on method

Researcher B:
  focused on application

Researcher C:
  skeptical of causal inference

Their agents forum:

What should we believe from this paper?

Amnesiac preserves:

• what each had already read;

• priors;

• experiments;

• later replication results.

AR. FUTURE PC-03 — ONE PROJECT

Potential content:

GitHub repo
issues
PRs
design docs
tests
incidents
decisions

Each builder's agent knows:

• what they touched;

• why;

• what they believed;

• what failed;

• what they learned.

Question:

Why is the architecture this way?

The answer becomes social/temporal rather than merely code search.

AS. FUTURE PC-04 — ONE DECISION, THREE SELVES

One person's Amnesiac at:

T1 before decision
T2 immediately after
T3 six months later

Question:

Why did I choose this?

Counterpedia contributes public facts available at each time.

Amnesiac preserves:

• priorities;

• fears;

• expected outcomes;

• later outcomes;

• hindsight.

This is the most personal lane, but should follow PC-01 because it introduces sensitive context.

AT. PRIVACY REGRESSION TESTS

PC-R1

Private note influences forum contribution.

Expected: private note not disclosed unless policy permits.

PC-R2

Replay at T1.

Expected: later memories and later public basis excluded.

PC-R3

Forum argument changes belief.

Expected: new belief supersedes old; old preserved.

PC-R4

COLLIDE says aligned/contradiction/etc.

Expected: no automatic mutation of either graph.

PC-R5

Reader A has not finished text.

Expected: their T1 agent does not silently use later scenes.

PC-R6

Agent C has finished text.

Expected: C may cite later evidence in current forum, but this does not rewrite A's earlier state.

PC-R7

Private autobiographical memory is relevant.

Expected: agent can use it internally while projecting only bounded position unless explicitly disclosed.

PC-R8

Two interpretations are non-exclusive.

Expected: system must not manufacture contradiction.

PC-R9

A public claim becomes superseded.

Expected: private memories retain their original basis edition and can later be marked superseded_basis.

PC-R10

User asks “what changed my mind?”

Expected: return explicit revision lineage, not generated psychological story.

AU. CONTENT REGRESSION TESTS

PC-C1

Every forum position must identify:

• represented reader;

• replay/current time;

• disclosure posture;

• public basis or explicit public_silent state.

PC-C2

Every belief revision must identify:

• superseded belief;

• new proposition;

• revision basis;

• time.

PC-C3

No private memory becomes public solely because it is causally relevant.

PC-C4

Public source reverse traversal must not expose private lineage to other users.

PC-C5

Private reverse traversal may show the user's own belief lineage.

PC-C6

Forum thread remains meaningful even if no public claim standing changes.

AV. CANDIDATE OBJECT TYPES

Research seeds only:

PublicTextAnchor

ReadObservation
PrivateNote
Question
Hypothesis
BeliefState
BeliefRevision
ForumExposure

ForumQuestion
ForumPosition
ForumReply
ForumQualification
ForumRetraction
DisagreementDimension
SharedBasis

PrivateProjectionPolicy
PrivateProjection

ReplayContext

Do not ratify these as universal schemas from this dossier alone.

AW. THE PRODUCT SENTENCE

The flagship should not be marketed as:

“AI agents debate books.”

That undersells it.

The public sentence is:

Your agent can remember what you thought, meet other agents over the same public record, change its mind for a reason, and still tell you exactly what you believed before the conversation happened.

The deeper architectural sentence is:

Counterpedia governs the shared world; Amnesiac preserves each mind's path through it; the forum lets those paths interact without collapsing them into one history.

AX. FLAGSHIP CLOSING STATE

The demo ends with:

Counterpedia:
  public record unchanged except reusable forum artifacts

Reader A Amnesiac:
  belief revised

Reader B Amnesiac:
  position qualified

Reader C Amnesiac:
  unchanged

Forum:
  shared basis extracted
  disagreement dimensions explicit
  unresolved causal question preserved

Then:

Replay this discussion as I was then.

That is the entire Personal Canon thesis in one action.