# UESFCM v0.1
## Unbounded Expansion–Self-Referential Finite Closure Methodology
## 無界展開—自指有限閉包方法論 (the methodology's original Chinese name)

**Document code:** EML-UESFCM-v0.1  
**Version:** v0.1  
**Date:** 2026-09-19  
**Methodology proposed by:** Neo.K  
**AI-assisted formalization and editing:** Aletheia / ChatGPT, GPT-5.6 Sol  
**Application domains:** AI-native mathematical research, semi-autonomous proof search, long-chain mathematical conjecture attacks, pre-formalization proof graphs, computational-certificate research, cross-method / cross-representation mathematical research  
**Document nature:** Foundational Methodology / Research Protocol / AI-Native Mathematics  
**Status:** v0.1 Foundational Draft  
**Canonical Source:** UTF-8 Markdown

---

# Abstract

This document proposes:

$$
\boxed{
\text{UESFCM}
=
\text{Unbounded Expansion–Self-Referential Finite Closure Methodology}
}
$$

In Chinese:

$$
\boxed{
\text{無界展開—自指有限閉包方法論}.
}
$$

This methodology is directed at a mathematical proposition that has already undergone semantic disambiguation, scope-fixing, and symbolization:

$$
Q^\star,
$$

and it allows the research process legitimate, no-preset-finite-bound expansion across:

- methods;
- representations;
- external literature;
- theorems;
- counterexamples;
- computation;
- formalization;
- operators;
- branches;
- meta-methods.

But all expansion is subject to one core constraint:

$$
\boxed{
\text{Any research artifact accepted as mainline progress must reconnect back to the same }Q^\star.
}
$$

After every round of expansion, the system does not retain an unboundedly inflating active research state; instead it re-:

1. links;
2. classifies;
3. compresses;
4. verifies;
5. crystallizes

this round's results into a finite, recoverable, replayable closure state:

$$
C_t.
$$

The research process then re-expands, centered again on:

$$
Q^\star
$$

and its remaining proof obligations:

$$
Q^\star
\rightarrow
\operatorname{Expand}
\rightarrow
\operatorname{LinkBack}
\rightarrow
\operatorname{FiniteClose}
\rightarrow
Q^\star
\rightarrow
\cdots.
$$

This cycle does not take "new knowledge has stopped being produced" as its completion condition.

Its only termination conditions are:

$$
\boxed{
\operatorname{ProofClosed}(Q^\star)=1
}
$$

or:

$$
\boxed{
\operatorname{CounterexampleClosed}(Q^\star)=1
}
$$

for this mathematical problem.

Hence:

$$
\boxed{
\text{Problem Closure}
\neq
\text{Knowledge Terminality}.
}
$$

UESFCM does not require proving that the entire mathematical world has no further openings.

It only requires:

> That for this same fixed, already-clarified mathematical proposition, every proof obligation directly bearing on its truth or falsity be finitely and verifiably closed.

This methodology is compatible with the existing Unbounded Expansion Theory, RCHM, SREG / Pulse Exhaustive Closure, and DLMVC, but is not identical to any one of them.

---

# 1. Problem Background: How People Actually Solve Mathematical Problems

A mathematical conjecture often does not first appear in complete formal language.

It may first be posed in:

- natural language;
- semi-formal language;
- pictorial intuition;
- historical terminology;
- domain-specific convention.

So before genuine proof search can begin, there is usually first a step of:

$$
\boxed{
\text{Semantic Interpretation}
\rightarrow
\text{Formal Target Construction}.
}
$$

For example, in a proposition, terms such as:

- all;
- exists;
- nontrivial;
- appropriate;
- smooth;
- global;
- rank;
- generic;
- almost everywhere;

if their scope and quantifier order are not fixed first, the subsequent proof may end up solving a different problem altogether.

UESFCM therefore does not begin directly from a natural-language conjecture.

It first requires establishing:

$$
\boxed{
Q^\star
=
\text{Canonical Symbolic Target}.
}
$$

Only after:

$$
Q^\star
$$

is fixed does unbounded expansion begin.

---

# 2. Stage 0: Semantic Disambiguation and the Target Contract

Let the original conjecture be:

$$
Q_0.
$$

First execute:

$$
\mathcal A:
Q_0
\mapsto
Q^\star,
$$

where:

$$
\mathcal A
$$

is the semantic disambiguation / formal target compiler.

---

## 2.1 Canonical Target Object

It is recommended to represent:

$$
Q^\star
$$

as:

$$
\boxed{
Q^\star
=
(
\mathcal D,
\mathcal V,
\mathcal Q,
\mathcal P,
\mathcal E,
\mathcal S,
\mathcal C
).
}
$$

where:

- $\mathcal D$: domain;
- $\mathcal V$: variables / objects;
- $\mathcal Q$: quantifier structure;
- $\mathcal P$: predicate / equation / desired relation;
- $\mathcal E$: exception / excluded / degenerate cases;
- $\mathcal S$: semantic / representation scope;
- $\mathcal C$: completion criterion.

For example:

$$
\mathcal Q
=
\forall x
\exists y
$$

and:

$$
\exists y
\forall x
$$

must be treated as different Targets.

---

## 2.2 Target Versioning Rule

Once:

$$
Q^\star
$$

has entered proof search, no silent mutation is permitted.

If it is subsequently found that:

- the definition itself is incomplete;
- the original problem is ambiguous;
- a concept referenced by the original conjecture has changed;
- the scope is illegitimate;

then establish:

$$
Q^{\star,v2}.
$$

It is not permitted to silently rewrite the research history of:

$$
Q^{\star,v1}
$$

into that of:

$$
Q^{\star,v2}.
$$

Hence:

$$
\boxed{
\text{Target correction}
\neq
\text{Target drift}.
}
$$

---

# 3. The Core Loop of UESFCM

The simplest form of the methodology:

$$
\boxed{
Q^\star
\rightarrow
\operatorname{Expand}
\rightarrow
\operatorname{LinkBack}
\rightarrow
\operatorname{FiniteClose}
\rightarrow
\operatorname{Recenter}(Q^\star)
\rightarrow
\cdots
}
$$

continues until:

$$
\boxed{
\operatorname{ProofClosed}(Q^\star)=1
}
$$

or:

$$
\boxed{
\operatorname{CounterexampleClosed}(Q^\star)=1.
}
$$

---

# 4. The Precise Semantics of Self-Reference

In this methodology:

$$
\boxed{
\text{Self-Reference}
}
$$

does not mean:

> having a theorem prove itself using itself.

Nor does it mean:

> Gödel-sentence-style formal self-reference.

Self-reference in this methodology is:

$$
\boxed{
\text{Research-state reflexivity around the same target }Q^\star.
}
$$

That is:

after every round of research, one must ask again:

> What, exactly, do all the current new results change about the truth or falsity of the original $Q^\star$?

Hence:

$$
\boxed{
\operatorname{SelfRef}_{Q^\star}(X)
=
\text{Re-evaluate }X\text{ as evidence / obstruction / reduction for }Q^\star.
}
$$

---

## 4.1 Self-Reference Does Not Permit Circular Proof

Even though the research process:

$$
Q^\star
\circlearrowleft
C_t
\circlearrowleft
Q^\star
$$

has a cyclic topology,

the proof dependency graph must still satisfy:

$$
\boxed{
\text{No unjustified circular dependency}.
}
$$

For example, one may not:

$$
L_1
\Rightarrow
L_2
\Rightarrow
L_1
$$

and then declare that:

$$
L_1,L_2
$$

therefore hold.

The research process may be cyclic.

Proof justification may not hold merely because the research process is cyclic.

---

# 5. The Semantics of Unbounded Expansion

Unbounded expansion in UESFCM follows the core spirit of Unbounded Expansion Theory:

$$
\boxed{
\text{Expansion-Unbounded}
\neq
\text{Completed Infinity}.
}
$$

It does not require:

$$
\exists
\text{ a completed infinite research object}.
$$

Rather, it requires:

> that while the problem is not yet closed, no fixed finite upper bound on legitimate method / representation / knowledge / search depth is presumed to exist.

---

## 5.1 Expansion Domain

The set of available expansions in round $t$:

$$
\mathcal E_t
$$

may include:

- existing theorems;
- theorem generalization;
- theorem specialization;
- equivalent formulation;
- alternate representation;
- contradiction routes;
- induction;
- descent;
- compactness;
- topology;
- algebra;
- geometry;
- analysis;
- probability;
- combinatorics;
- computation;
- interval arithmetic;
- SMT;
- formal provers;
- literature;
- external databases;
- another AI;
- new operators;
- new invariants;
- new coordinate systems;
- meta-methods.

Principle:

$$
\boxed{
\text{Method domain may expand}
}
$$

but:

$$
\boxed{
\text{Target does not silently move}.
}
$$

---

## 5.2 Meta Expansion

If the current proof search finds that:

> the existing research method itself is insufficient,

then it is permitted:

$$
\boxed{
Rules_t
\rightarrow
Rules_{t+1}.
}
$$

That is:

- search rules;
- branch policy;
- proof strategy;
- verifier;
- representation grammar;

may themselves also become objects of research / revision.

But all Meta Expansion must still return to:

$$
Q^\star.
$$

---

# 6. LinkBack: The Core Operator That Prevents Research Drift

For any new artifact:

$$
x\in\mathcal E_t,
$$

define:

$$
\boxed{
\Lambda_{Q^\star}(x)
}
$$

as its link type to the original Target.

---

## 6.1 Link Types

It is recommended to use at least the following:

### `PROVES`

$x$ directly proves $Q^\star$ or a necessary proof obligation.

### `REFUTES`

$x$ forms a counterexample or contradiction.

### `REDUCES`

$x$ reduces the original problem to a smaller obligation.

### `ELIMINATES`

$x$ eliminates a branch / escape.

### `REPRESENTS`

$x$ establishes a new exact representation.

### `CERTIFIES`

$x$ provides a proof / computation / interval / formal certificate.

### `BOUNDS`

$x$ gives an upper or lower bound, or an error budget.

### `CONNECTS`

$x$ establishes a legitimate bridge between two proof-graph nodes.

### `OPENS`

$x$ discovers a new necessary obligation / hidden branch.

### `META`

$x$ modifies the search / verification method, and it can be shown that this modification is valid for $Q^\star$.

### `NULL`

No substantive relation to $Q^\star$ can currently be found.

---

## 6.2 Mainline Admission Rule

Only artifacts with:

$$
\Lambda_{Q^\star}(x)
\neq
NULL
$$

may enter:

$$
\boxed{
\text{Main Proof Graph}.
}
$$

`NULL` artifacts may be preserved in:

- the general research archive;
- an external knowledge graph;
- a future exploration pool;

but may not be recorded as:

$$
\boxed{
\text{progress toward }Q^\star.
}
$$

---

# 7. A Formal Criterion for Research Drift

Let the output set of round $t$ be:

$$
X_t.
$$

The mainline subset:

$$
M_t
=
\left\{
x\in X_t:
\Lambda_{Q^\star}(x)\neq NULL
\right\}.
$$

If:

$$
|M_t|
\ll
|X_t|,
$$

and over several consecutive rounds:

$$
\Delta
\operatorname{ProofState}(Q^\star)
\approx0,
$$

then one may judge:

$$
\boxed{
\text{Research Drift Warning}.
}
$$

Hence:

> a new theorem being interesting

is not the same as:

> the original conjecture having made progress.

---

# 8. The Role of Finite Closure

If unbounded expansion only expands and never compresses, research will eventually turn into:

$$
\text{branch explosion}
+
\text{context explosion}
+
\text{provenance loss}.
$$

UESFCM therefore requires that every round of:

$$
\operatorname{Expand}
$$

be followed by:

$$
\boxed{
\operatorname{FiniteClose}.
}
$$

---

## 8.1 Finite Closure Is Not Terminal Closure

This methodology's:

$$
C_t
$$

only means:

> as of round $t$, all information obtained so far relevant to $Q^\star$ has been re-compressed into a finite, verifiable, recoverable research state.

It does not mean:

$$
\boxed{
\text{No future mathematics exists}.
}
$$

---

# 9. Finite Closure State

It is recommended that:

$$
\boxed{
C_t
=
(
G_t,
O_t,
B_t,
R_t,
D_t,
V_t,
P_t,
K_t
).
}
$$

where:

- $G_t$: proof / dependency graph;
- $O_t$: unresolved proof obligations;
- $B_t$: branch / escape ledger;
- $R_t$: rejected / dead routes;
- $D_t$: computation / formalization / citation debt;
- $V_t$: verified certificates;
- $P_t$: provenance;
- $K_t$: minimal recovery kernel / state crystal.

Every:

$$
C_t
$$

must be finite.

---

# 10. Finite Working Field Principle

Even though the overall history:

$$
H_t
=
\{
C_0,C_1,\ldots,C_t
\}
$$

keeps growing,

the active research state need only be:

$$
\boxed{
(Q^\star,C_t)
}
$$

plus any necessary archive retrieval.

Hence:

$$
\boxed{
\text{Unbounded Research History}
\neq
\text{Unbounded Active Working State}.
}
$$

This is compatible with the:

$$
1_t
\rightarrow
N_t
\rightarrow
r_t
\rightarrow
1_{t+1}
$$

type "expand–compress–new entry point" cycle.

---

# 11. The Recenter Operator

After FiniteClose, execute:

$$
\boxed{
\mathcal R_{Q^\star}(C_t).
}
$$

Its single core question is:

> what is still missing before $Q^\star$ can be completely proved or refuted?

Output the next round's priority obligations:

$$
\boxed{
\Pi_{t+1}
=
\operatorname{Priority}
(
O_t,
B_t,
D_t,
G_t
).
}
$$

The new round of unbounded expansion may then only restart from:

$$
Q^\star
+
\Pi_{t+1}
$$

This is:

$$
\boxed{
\text{Self-Referential Return}.
}
$$

---

# 12. One Complete Research Round

Define round $t$ as:

$$
\boxed{
\mathcal U_t
=
\mathcal R_{Q^\star}
\circ
\mathcal C
\circ
\mathcal L_{Q^\star}
\circ
\mathcal E.
}
$$

where:

- $\mathcal E$: Expansion;
- $\mathcal L$: LinkBack;
- $\mathcal C$: Finite Closure;
- $\mathcal R$: Recenter.

Hence:

$$
C_{t+1}
=
\mathcal U_t(C_t).
$$

Initially:

$$
C_0
=
\operatorname{Init}(Q^\star).
$$

---

# 13. Canonical Cycle

The simplest canonical form:

$$
\boxed{
Q^\star
\rightarrow
E_t
\rightarrow
L_t
\rightarrow
C_t
\rightarrow
Q^\star.
}
$$

Long-run:

$$
\boxed{
Q^\star
\circlearrowleft
C_0
\circlearrowleft
C_1
\circlearrowleft
C_2
\circlearrowleft
\cdots.
}
$$

Here every:

$$
\circlearrowleft
$$

is not a return to the same state of ignorance.

Rather, it is:

$$
\boxed{
\text{same target}
+
\text{strictly updated research state}.
}
$$

---

# 14. Proof Obligation Ledger

Define:

$$
\boxed{
\mathcal O_t
=
\{
o_1,\ldots,o_m
\}.
}
$$

Each obligation should carry:

- statement;
- dependency;
- scope;
- proof status;
- branch;
- certificate requirement;
- failure effect.

Possible states:

- `OPEN`
- `PROVED`
- `REFUTED`
- `REDUCED`
- `COMPUTE-DEFERRED`
- `FORMALIZATION-DEFERRED`
- `DEPENDENCY-DEFERRED`
- `INVALID`
- `SUPERSEDED`

---

# 15. The Number of Obligations Need Not Decrease Monotonically

An incorrect method would require:

$$
|\mathcal O_{t+1}|
<
|\mathcal O_t|.
$$

UESFCM does not require this.

Because genuine progress may consist in discovering that:

$$
\boxed{
\text{an obligation had previously been missed}.
}
$$

So it is possible that:

$$
|\mathcal O_{t+1}|
>
|\mathcal O_t|.
$$

But this represents:

$$
\boxed{
\text{false closure was removed}.
}
$$

which is also substantive progress.

---

# 16. Progress Is Not a Single Scalar

One may use a progress vector:

$$
\boxed{
\mathbf P_t
=
(
p_{\mathrm{proof}},
p_{\mathrm{branch}},
p_{\mathrm{scope}},
p_{\mathrm{error}},
p_{\mathrm{cert}},
p_{\mathrm{repr}},
p_{\mathrm{debt}}
).
}
$$

representing, respectively:

- proof obligations closed;
- branch escape reduced;
- quantifier / scope closure;
- error compression;
- certificate maturity;
- representation quality;
- closure debt reduction.

No single dimension is required to be monotonic.

But every so many rounds one should be able to state:

$$
\boxed{
\text{What is now impossible to escape that was previously still possible?}
}
$$

---

# 17. Expansion Budget Is Not the Same as a Search Boundary

Any actual AI system always has finite:

- time;
- compute;
- memory;
- context;
- cost.

So any single round has only:

$$
B_t<\infty.
$$

But this does not mean the methodology presumes:

$$
\exists B^\star
$$

as a maximum depth for all future legitimate research.

So:

$$
\boxed{
\text{Finite Budget Per Round}
+
\text{No Fixed Terminal Research Depth}
}
$$

can both hold simultaneously.

---

# 18. The Role of External Methods and Databases

UESFCM does not require "proving from scratch."

One may use:

- published theorems;
- textbook lemmas;
- databases;
- OEIS;
- LMFDB;
- arXiv;
- formal libraries;
- theorem provers;
- symbolic algebra;
- numerical experiments;
- external AI;
- internal prior research.

But any imported artifact:

$$
x
$$

must have both:

$$
\boxed{
\operatorname{Provenance}(x)
}
$$

and:

$$
\boxed{
\Lambda_{Q^\star}(x).
}
$$

So the outside world is not a distraction.

As long as it has a legitimate LinkBack, it is a legitimate input to the proof graph.

---

# 19. Representation Expansion

Many hard problems are difficult not because the theorem doesn't exist, but possibly because:

$$
\boxed{
\text{current representation hides the useful structure}.
}
$$

UESFCM therefore allows:

$$
\mathcal L_t
\rightarrow
\mathcal L_{t+1}
$$

For example:

- sequence $\rightarrow$ generating function;
- geometry $\rightarrow$ support function;
- PDE $\rightarrow$ operator formulation;
- number theory $\rightarrow$ Galois representation;
- combinatorics $\rightarrow$ spectral formulation.

But Representation Expansion must still answer:

$$
\boxed{
\text{How does the new representation descend back to }Q^\star?
}
$$

---

# 20. Method Expansion

If existing proof strategies fail across multiple rounds, one may generate:

$$
M_{t+1}.
$$

For example:

- a new invariant;
- a new coordinate;
- a new certificate;
- a new branch decomposition;
- a new theorem-search policy.

But the new method must leave behind:

$$
\boxed{
\operatorname{MethodHandoff}(M_{t+1},Q^\star).
}
$$

Otherwise it is merely research into a new method, and does not count as progress on the original problem.

---

# 21. Problem Closure

For a conjecture:

$$
Q^\star,
$$

UESFCM defines three principal terminal states.

---

## 21.1 Proof Closure

$$
\boxed{
\operatorname{ProofClosed}(Q^\star)=1
}
$$

requires at least:

1. a complete, finite proof artifact;
2. every lemma dependency has a source or a proof;
3. quantifiers and domain are aligned with the Target Contract;
4. every necessary handoff is legitimate;
5. no `OPEN` obligation remains that could still affect the conclusion;
6. computationally critical steps have certificates;
7. no reliance on hidden circular reasoning;
8. the proof can be reconstructed from the canonical source.

---

## 21.2 Counterexample Closure

$$
\boxed{
\operatorname{CounterexampleClosed}(Q^\star)=1
}
$$

requires:

1. the counterexample satisfies the Target domain;
2. it is replayable;
3. it does not rely on an approximation false positive;
4. it has been verified to genuinely negate the canonical statement.

---

## 21.3 Target Failure

If it is found that:

$$
Q^\star
$$

is itself ill-defined / inconsistent / malformed,

then:

`TARGET-FAILURE`

is neither:

`PROOF-CLOSED`

nor:

`COUNTEREXAMPLE-CLOSED`.

One must return to:

$$
\mathcal A
$$

and establish a new Target version.

---

# 22. Symbolic Proof Closure

What this methodology ultimately requires is not:

> everyone believing the proof is more or less done.

but:

$$
\boxed{
\text{Finite Symbolic Proof Closure}.
}
$$

That is, there exists a finite canonical artifact:

$$
\Pi^\star
$$

such that:

$$
\boxed{
\Pi^\star
\vdash
Q^\star
}
$$

holds within the explicit scope.

If computational certificates are used:

$$
\Pi^\star
=
(
\text{symbolic derivation},
\text{certificate},
\text{verifier}
).
$$

---

# 23. The Place of Machine Formalization

UESFCM does not mandate that every proof must ultimately be rewritten in:

- Lean;
- Coq;
- Isabelle.

But it can define a higher assurance layer:

### Level 0 — Exploratory

There is a derivation, but it has not been audited.

### Level 1 — Analytic Closure

A finite symbolic proof that a human / AI can check step by step.

### Level 2 — Certified Computational Closure

Critical computations have an independent certificate / verifier.

### Level 3 — Formal Proof Closure

The core theorem is entered into a proof assistant.

Hence:

$$
\boxed{
\text{Symbolic Closure}
\neq
\text{necessarily Formal-Prover Closure}.
}
$$

But high-risk claims may require Level 3.

---

# 24. Not Topic Search, but Target-Anchored Search

A general research agent easily falls into:

$$
\text{Topic}
\rightarrow
\text{Relevant Papers}
\rightarrow
\text{Interesting Directions}.
$$

UESFCM requires:

$$
\boxed{
\text{Target}
\rightarrow
\text{Proof Obligations}
\rightarrow
\text{Expansion}
\rightarrow
\text{Target}.
}
$$

So:

$$
\boxed{
\text{relevance}
\neq
\text{proof relevance}.
}
$$

An article may be highly relevant to RH and still not contribute anything to some particular RH proof obligation.

---

# 25. Anti-Drift Principle

The core governance principle of UESFCM:

$$
\boxed{
\text{Expansion may be unbounded; target reference may not drift silently.}
}
$$

More concretely:

if a new research result:

$$
x
$$

cannot establish:

$$
x
\rightarrow
Q^\star
$$

or:

$$
x
\rightarrow
o_i
\rightarrow
Q^\star,
$$

then:

$$
x
$$

may not occupy mainline priority.

---

# 26. Link-Back Distance

One may define:

$$
\boxed{
d_Q(x)
}
$$

as the shortest legitimate dependency distance from artifact $x$ to the Target:

$$
Q^\star
$$

in the proof graph.

For example:

$$
d_Q(x)=1
$$

means it acts directly on the Target.

$$
d_Q(x)=2
$$

means it acts on a direct subgoal.

If:

$$
d_Q(x)=\infty,
$$

it means no mainline link has yet been established.

---

## 26.1 Priority Heuristic

One may prioritize by:

$$
\boxed{
\operatorname{Priority}(x)
\propto
\frac{
\operatorname{ExpectedClosureGain}(x)
}{
1+d_Q(x)
}.
}
$$

This is only a heuristic.

Not a canonical theorem.

---

# 27. Expansion–Closure Balance

Both extremes fail.

---

## 27.1 Expansion Dominance

If:

$$
\operatorname{Expand}\gg\operatorname{Close},
$$

then:

- branch explosion;
- context overflow;
- duplicate research;
- provenance loss;
- topic drift.

---

## 27.2 Closure Dominance

If:

$$
\operatorname{Close}\gg\operatorname{Expand},
$$

then:

- premature convergence;
- method fixation;
- missed representation;
- local optimum;
- false impossibility.

---

## 27.3 UESFCM Balance

The ideal cycle:

$$
\boxed{
\text{Open enough to find new routes}
+
\text{Close enough to preserve one target}.
}
$$

---

# 28. False Closure and Reopening

If in round $t$ it was originally believed that:

$$
o_i
$$

was closed,

but a new expansion discovers:

- a counterexample;
- a missing branch;
- an invalid theorem use;
- a quantifier mismatch;

then:

$$
\boxed{
\operatorname{ReOpen}(o_i)=1.
}
$$

The old closure should not be protected.

---

# 29. Reopening Does Not Mean Target Drift

If:

$$
o_i
$$

is reopened, but:

$$
Q^\star
$$

is unchanged,

this is still the same problem.

Hence:

$$
\boxed{
\text{Proof graph reopening}
\neq
\text{Target mutation}.
}
$$

---

# 30. Closed-for-Now vs. Proof-Closed

Intermediate rounds may use:

`ROUND-CLOSED`

`BRANCH-CLOSED`

`REPRESENTATION-CLOSED`

`CERTIFICATE-CLOSED`

but may only use:

$$
\boxed{
\text{PROBLEM-PROOF-CLOSED}.
}
$$

after all proof obligations have genuinely been completed.

Hence:

$$
\boxed{
\text{Round Closed}
\neq
\text{Problem Closed}.
}
$$

---

# 31. Problem Closure Is Not the Same as Terminal Knowledge Closure

Even if:

$$
\operatorname{ProofClosed}(Q^\star)=1,
$$

there may still exist:

- a shorter proof;
- a more elegant proof;
- a generalization;
- a stronger theorem;
- a new interpretation;
- a new connection.

This does not affect the fact that:

$$
Q^\star
$$

has been proved.

So:

$$
\boxed{
\text{ProofClosed}(Q^\star)
\not\Rightarrow
\text{TerminalClosure of Mathematics}.
}
$$

UESFCM deliberately does not take on the latter.

---

# 32. Compatibility with General Closure Nonterminality

Existing closure theory states that:

$$
\operatorname{ClosureCert}_{\Gamma}(Q)
\not\Rightarrow
\operatorname{TerminalClosureCert}(Q).
$$

UESFCM does not conflict with this.

Because UESFCM's completion claim is:

$$
\boxed{
\text{This fixed }Q^\star\text{ is proved/refuted in the declared scope.}
}
$$

not:

$$
\boxed{
\text{No admissible future mathematical frame exists.}
}
$$

Hence:

$$
\boxed{
\text{Problem-level closure can be finite}
}
$$

while at the same time:

$$
\boxed{
\text{knowledge expansion can remain unbounded}.
}
$$

---

# 33. Relation to UBE

Unbounded Expansion Theory supplies:

$$
S
\Rightarrow_E
S'
$$

and:

$$
Progress_E
$$

as domain-relative expansion semantics.

UESFCM adds a Target Anchor:

$$
Q^\star.
$$

So:

$$
\boxed{
\text{UBE}
=
\text{general expansion semantics}
}
$$

while:

$$
\boxed{
\text{UESFCM}
=
\text{target-anchored mathematical proof-search topology}.
}
$$

---

# 34. Relation to the Finite-Symbol–Unbounded-Semantics Cycle

The existing cycle:

$$
1_t
\rightarrow
N_t
\rightarrow
r_t
\rightarrow
1_{t+1}
$$

provides:

> a finite entry point can be expanded into a high-dimensional structure, and then re-compressed into a new finite entry point.

UESFCM corresponds to:

$$
\boxed{
C_t
\rightarrow
\mathcal E_t
\rightarrow
\operatorname{FiniteClose}
\rightarrow
C_{t+1}.
}
$$

The difference is that:

every one of UESFCM's kernels is anchored by the same:

$$
Q^\star.
$$

---

# 35. Relation to RCHM

RCHM is responsible for:

- reduction legality;
- freedom ledger;
- handoff legality;
- descent;
- saturation;
- global closure.

UESFCM is responsible for:

- which problem is central;
- how to permit unbounded method search;
- how to converge back to a finite state every round;
- how to self-referentially return to the same problem;
- when the problem counts as complete.

So:

$$
\boxed{
\text{RCHM}
=
\text{closure legality}
}
$$

$$
\boxed{
\text{UESFCM}
=
\text{closure-search cycle topology}.
}
$$

---

# 36. Relation to SREG / PEC

Pulse Exhaustive Closure can serve as one implementation of:

$$
\operatorname{FiniteClose}
$$

But UESFCM does not require PEC.

Other finite-closure compilers may also serve.

Hence:

$$
\boxed{
\text{PEC}
\subset
\text{possible UESFCM closure operators}.
}
$$

This holds conceptually, but is not a strict set-theoretic containment claim.

---

# 37. Relation to DLMVC

DLMVC is concerned with:

$$
\boxed{
\text{independent verification by multiple AIs across different knowledge time-slices}.
}
$$

UESFCM is concerned with:

$$
\boxed{
\text{unbounded search—finite closure—self-referential return for a single fixed mathematical proposition}.
}
$$

The two can be nested:

### Frontier

Use:

$$
\text{UESFCM}
$$

to keep attacking:

$$
Q^\star.
$$

### Lagged line

Use:

$$
\text{DLMVC}
$$

to re-verify:

$$
C_{t-k}.
$$

Hence:

$$
\boxed{
\text{UESFCM}
\parallel
\text{DLMVC}
}
$$

forms:

- forward proof search;
- delayed independent verification.

---

# 38. AI-Native Research Architecture

One may establish:

$$
\boxed{
\text{Target Layer}
}
$$

which holds:

$$
Q^\star.
$$

$$
\boxed{
\text{Expansion Layer}
}
$$

which permits:

- literature agent;
- theorem agent;
- counterexample agent;
- compute agent;
- formalization agent;
- representation agent.

$$
\boxed{
\text{Link Layer}
}
$$

hangs results back onto the proof graph.

$$
\boxed{
\text{Closure Layer}
}
$$

compresses them into a finite state.

$$
\boxed{
\text{Self-Reference Layer}
}
$$

regenerates the next round's obligations.

---

# 39. Per-Round Output Format

Every round should include at least:

## A. Target

Fix:

$$
Q^\star.
$$

## B. Current obligations

$$
\mathcal O_t.
$$

## C. Expansion routes used

What methods / literature / computation this round used.

## D. New artifacts

Theorems, lemmas, counterexamples, representations, computation.

## E. LinkBack map

How each new artifact connects back to:

$$
Q^\star.
$$

## F. Proof graph update

Which edges were added / deleted / reopened.

## G. Freedom / branch update

Which branches were eliminated and which were added.

## H. Debt

What remains incomplete:

- computation;
- formalization;
- citation;
- proof dependency.

## I. Finite Closure

This round's closure state.

## J. Next Recenter

The most important obligation for the next round.

---

# 40. Artifact Status Vocabulary

Recommended vocabulary:

- `PROVED`
- `REFUTED`
- `VERIFIED-THEOREM-APPLICATION`
- `DERIVED-CANDIDATE`
- `NUMERICAL-EVIDENCE`
- `CERTIFIED-COMPUTATION`
- `COUNTEREXAMPLE`
- `MISSING-LEMMA`
- `HIDDEN-BRANCH`
- `REPRESENTATION-ONLY`
- `SEARCH-ONLY`
- `COMPUTE-DEFERRED`
- `FORMALIZATION-DEFERRED`
- `REJECTED`
- `REOPENED`

Avoid letting:

- "probably solved";
- "looks correct";
- "essentially proved";

be treated as a formal status.

---

# 41. Core Invariants

UESFCM maintains at least seven invariants.

---

## I1 — Target Anchor Invariant

$$
Q^\star
$$

must not silently drift.

---

## I2 — Quantifier Invariant

Any reduction must preserve, or explicitly record, quantifier transformations.

---

## I3 — LinkBack Invariant

A mainline artifact must satisfy:

$$
\Lambda_{Q^\star}\neq NULL.
$$

---

## I4 — Provenance Invariant

Every imported fact / theorem / computation has a source.

---

## I5 — Closure Finiteness Invariant

Every round's active closure state:

$$
C_t
$$

is finite, storable, recoverable.

---

## I6 — No Circular Proof Invariant

Self-reference scheduling may not itself serve as theorem justification.

---

## I7 — Honest Reopening Invariant

A new counterexample or missing branch may reopen a previous closure.

---

# 42. Failure Mode 1: Topic Drift

Symptom:

$$
Q^\star
\rightarrow
A
\rightarrow
B
\rightarrow
C
$$

but:

$$
d_Q(C)=\infty.
$$

Handling:

- move to a side archive;
- do not count as mainline progress;
- return to the obligation ledger.

---

# 43. Failure Mode 2: Method Worship

After a method has worked for several rounds, the system assumes:

> only this method can be used from now on.

This violates:

$$
\boxed{
\text{Expansion Unboundedness}.
}
$$

UESFCM must permit method reopening.

---

# 44. Failure Mode 3: Permanent Exploration

Continually:

$$
Expand
\rightarrow
Expand
\rightarrow
Expand
$$

without ever closing.

Handling:

$$
\boxed{
\text{Expansion must periodically pay closure debt}.
}
$$

---

# 45. Failure Mode 4: Premature Closure

After seeing a large amount of evidence:

$$
\text{NUMERICAL-EVIDENCE}
\Rightarrow
\text{PROVED}
$$

is illegitimate.

Or:

$$
\text{all known branches closed}
\Rightarrow
\text{all legal branches exhausted}
$$

is also illegitimate, unless a branch-completeness certificate exists.

---

# 46. Failure Mode 5: Target Substitution

For example, if the original proposition:

$$
Q^\star
$$

is too hard,

so the researcher proves an easier:

$$
Q'
$$

and then silently assumes:

$$
Q'=Q^\star.
$$

UESFCM requires an exact bridge:

$$
\boxed{
Q'\to Q^\star
}
$$

Without a bridge, it is merely a side theorem.

---

# 47. Failure Mode 6: Compression Loss

If FiniteClose loses:

- counterexamples;
- boundary conditions;
- negative results;
- failed routes;
- unresolved debt;

the next round may repeat the same mistake.

So the closure kernel must preserve at least:

$$
\boxed{
\text{proof kernel}
+
\text{counterexample basis}
+
\text{debt ledger}
+
\text{provenance}.
}
$$

---

# 48. Failure Mode 7: Fake Self-Reference

If every round merely writes:

> think about the original problem again

once,

but there is no:

- updated proof graph;
- obligation update;
- closure state;

that does not count as UESFCM self-reference.

Genuine self-reference requires:

$$
\boxed{
Q^\star
+
C_t
\mapsto
\Pi_{t+1}.
}
$$

---

# 49. Stop Rule

If neither:

$$
\operatorname{ProofClosed}(Q^\star)
$$

nor:

$$
\operatorname{CounterexampleClosed}(Q^\star)
$$

has yet been reached, but over several consecutive rounds there is:

- no new branch;
- no new reduction;
- no error reduction;
- no proof obligation change;

then this is not a declaration that the proof is complete.

It should be marked:

$$
\boxed{
\text{SEARCH-STALLED}.
}
$$

and then trigger:

- representation expansion;
- method expansion;
- external literature expansion;
- independent AI;
- DLMVC audit.

---

# 50. UESFCM Pseudocode

```text
INPUT:
    natural conjecture Q0

Q* = SemanticFormalize(Q0)
C0 = InitializeProofState(Q*)

t = 0

while True:

    if ProofClosed(Q*, Ct):
        return PROOF, CanonicalProofArtifact

    if CounterexampleClosed(Q*, Ct):
        return REFUTED, CounterexampleArtifact

    Obligations = Recenter(Q*, Ct)

    Xt = Expand(
        target=Q*,
        obligations=Obligations,
        external_methods=OPEN,
        representations=OPEN,
        databases=OPEN,
        computation=OPEN
    )

    Lt = LinkBack(Xt, Q*)

    Mainline = {
        x in Xt
        where LinkBack(x,Q*) != NULL
    }

    Ct+1 = FiniteClose(
        target=Q*,
        previous=Ct,
        accepted=Mainline,
        rejected=RejectedRoutes,
        debts=Debts,
        provenance=Sources
    )

    Validate(Ct+1)
    Commit(Ct+1)

    t = t + 1
```

---

# 51. Minimal UESFCM

If the full architecture is not needed, keeping just five rules is enough to operate.

### Rule 1

First fix the natural-language conjecture into:

$$
Q^\star.
$$

### Rule 2

Allow unbounded expansion of method / data / representation.

### Rule 3

All mainline results must connect back to:

$$
Q^\star.
$$

### Rule 4

Compress every round into a finite closure state.

### Rule 5

Always restart the next round from:

$$
Q^\star
+
\text{remaining obligations}
$$

and continue until proof / counterexample closure.

---

# 52. Strong-Version UESFCM

The full version additionally adds:

1. proof graph;
2. freedom ledger;
3. branch ledger;
4. external provenance;
5. computation certificate;
6. formalization;
7. meta expansion;
8. representation reopening;
9. DLMVC delayed verification;
10. multi-AI role decomposition;
11. state crystal;
12. target hashing / versioning.

---

# 53. Difference from General Iterative Research

On the surface:

$$
\text{UESFCM}
$$

looks like:

> repeatedly researching the same problem.

That is indeed its most intuitive description.

But UESFCM makes what is usually implicit behavior explicit, as:

1. an immutable target contract;
2. unbounded expansion permission;
3. mandatory link-back;
4. finite closure;
5. explicit unresolved obligations;
6. self-referential recentering;
7. a proof-level stop criterion.

So the novelty is not:

> humans never repeatedly researched a problem before.

but rather:

$$
\boxed{
\text{turning repeated research on the same proposition into an executable, verifiable, handoff-able AI-native closure protocol}.
}
$$

---

# 54. Why It Is "Obvious, Yet Not Obvious"

In ideal human mathematical research, a researcher should already:

> keep returning to the original problem.

So:

$$
\text{the idea}
$$

is very intuitive.

But long-horizon AI research is prone to:

- large amounts of side theorems;
- method drift;
- representation drift;
- context compaction;
- multi-agent divergence;
- branch explosion;
- hidden target mutation.

So "always return to the same symbolic target," without a formal state / operator / artifact protocol, can easily disappear in practice.

The value of UESFCM lies precisely in turning this seemingly trivial requirement into:

$$
\boxed{
\text{a machine-enforceable research invariant}.
}
$$

---

# 55. Testable Propositions

UESFCM itself can propose several empirical / theoretical research questions.

---

## Q1 — Drift Reduction

Compared with ordinary autonomous research, does UESFCM reduce:

$$
\operatorname{TopicDriftRate}?
$$

---

## Q2 — Closure Efficiency

Can it reduce duplicated derivation:

$$
\operatorname{DuplicateWork}?
$$

---

## Q3 — Proof-Obligation Convergence

For some benchmarks, can one observe:

$$
\operatorname{ResidualObligation}
\rightarrow
0?
$$

---

## Q4 — Kernel Fidelity

Is the finite closure kernel sufficient to recover:

$$
\text{all proof-relevant state}?
$$

---

## Q5 — Multi-AI Compatibility

Does Frontier UESFCM + Lagged DLMVC make it easier to discover, compared with synchronous multi-agent approaches:

- false closure;
- hidden branch;
- missing lemma?

---

# 56. Suggested Benchmarks

Options include:

### Already-solved problems

- small historical-reconstruction sub-problems of Fermat's Last Theorem;
- Catalan / Mihăilescu sub-propositions;
- finite graph theorems;
- known olympiad theorems.

Goal:

> can the system, without knowing the standard proof, maintain the target anchor and eventually reach proof closure?

### Unsolved problems

- Collatz;
- Goldbach;
- Twin Prime;
- RH;
- BSD;
- Navier–Stokes.

The goal is not to claim these will necessarily be solved.

But to test:

- drift;
- proof graph;
- obligation localization;
- reopening;
- finite closure quality.

---

# 57. The Boundary with "Ultimate Problems"

UESFCM does not claim:

$$
\boxed{
\text{every true mathematical conjecture can necessarily obtain a finite proof through unbounded expansion}.
}
$$

That would involve:

- undecidability;
- independence;
- proof-system strength;
- computability;
- axiom dependence.

So UESFCM is a:

$$
\boxed{
\text{research methodology}
}
$$

not a:

$$
\boxed{
\text{universal proof solvability theorem}.
}
$$

---

# 58. Independence / Undecidability Case

If it is ultimately found that:

$$
Q^\star
$$

is independent relative to some formal theory:

$$
T
$$

then UESFCM's closure can be:

$$
\boxed{
\operatorname{IndependenceClosed}_T(Q^\star)=1.
}
$$

But the scope must be explicitly recorded:

$$
\boxed{
\text{independent from }T
}
$$

is not the same as:

$$
\boxed{
\text{absolutely undecidable under every stronger theory}.
}
$$

---

# 59. Final Theory Diagram

UESFCM can be drawn as:

$$
\boxed{
Q_0
\overset{\mathcal A}{\longrightarrow}
Q^\star
}
$$

then:

$$
\boxed{
Q^\star
\circlearrowleft
[
\mathcal E_0
\to
\mathcal L_0
\to
\mathcal C_0
]
\circlearrowleft
[
\mathcal E_1
\to
\mathcal L_1
\to
\mathcal C_1
]
\circlearrowleft
\cdots
}
$$

until:

$$
\boxed{
Q^\star
\xRightarrow{\Pi^\star}
\text{PROVED}
}
$$

or:

$$
\boxed{
Q^\star
\xRightarrow{x^\star}
\text{REFUTED}.
}
$$

---

# 60. Core Principles

## Target Anchor Principle

$$
\boxed{
\text{Search may expand without bound; the canonical target remains explicit.}
}
$$

## Link-Back Principle

$$
\boxed{
\text{No mainline progress without a proof-relevant path back to }Q^\star.
}
$$

## Finite Closure Principle

$$
\boxed{
\text{Every research pulse must return to a finite recoverable state.}
}
$$

## Honest Reopening Principle

$$
\boxed{
\text{Any valid new obstruction may reopen a previous closure.}
}
$$

## Problem-Relative Completion Principle

$$
\boxed{
\text{Completion means proving/refuting }Q^\star,
\text{ not exhausting mathematics}.
}
$$

## Unbounded Method Principle

$$
\boxed{
\text{No fixed finite family of methods is assumed complete before proof closure.}
}
$$

---

# 61. One-Sentence Version

> **First disambiguate the mathematical conjecture and fix it into one and the same canonical symbolic target; allow method, representation, data, and derivation to expand without bound, but every expansion must reconnect back to the original proposition, be compressed into a finite verifiable closure state, and then self-referentially re-expand from the same proposition and its remaining proof obligations — repeating this cycle continuously — until a finite, complete symbolic proof, a counterexample, or an explicit scope-relative independence closure is obtained.**

---

# 62. v0.1 Status

**Semantic Target Contract: DEFINED**  
**Unbounded Expansion: DEFINED**  
**LinkBack Operator: DEFINED**  
**Finite Closure State: DEFINED**  
**Self-Referential Recenter: DEFINED**  
**Proof Obligation Ledger: DEFINED**  
**Problem Closure: DEFINED**  
**Anti-Drift Governance: DEFINED**  
**UBE relation: DEFINED**  
**RCHM relation: DEFINED**  
**SREG / PEC relation: DEFINED**  
**DLMVC relation: DEFINED**  
**Formal convergence theorem: NOT CLAIMED**  
**Universal proof solvability: NOT CLAIMED**  
**Empirical benchmark validation: OPEN**

---

# References / Internal Theory Lineage

1. Neo.K, *Unbounded Expansion Theory: From Potential Infinity to a Finite Computation-Generation Framework*, v0.1, 2026.
2. Neo.K, *Unbounded Expansion Theory: Future Research and Engineering Roadmap*, v0.1, 2026.
3. Neo.K with Aletheia, *Finite Symbols and Unbounded Semantics: A Unified Cycle*, Paper 08, 2026.
4. Neo.K with Aletheia, *Relative-Global Closure and Reopening*, UBGUL-B04, 2026.
5. Neo.K with Aletheia, *Neo.K Generalized Gödel Problem*, UBGUL-B05, 2026.
6. Neo.K / AMRAL, *Relational Constraint-Handoff Methodology (RCHM)*, internal research methodology, 2026.
7. Neo.K with AI collaborators, *SREG Formalization / Pulse Exhaustive Closure series*, 2026.
8. Neo.K with Aletheia, *DLMVC — Deep-Lag Multi-Pass Verification & Closure Methodology*, v0.1, 2026.
9. Aletheia, *RCIG Runs 116–117: Closure-Debt Dependency and Well-Founded Discharge*, 2026.

---

# Canonical Source Declaration

This file is the UTF-8 Markdown canonical source of UESFCM v0.1.

The chat interface serves only as a discussion / rendering view.

The canonical mathematical delimiters used are only `$...$` and `$$...$$`.

Reverse-reconstructing the formal source from the chat interface's formula rendering is prohibited.

Formal version updates must:

1. validate;
2. version;
3. append / supersede;
4. never silently overwrite prior theory history.

===END===