Abstract
That the continuum hypothesis is independent cannot be said by one proof. This paper asks why two are needed—the answer is that the two techniques point in opposite directions. Goedel selected from what was already there and narrowed; Cohen added what had not been there and widened. No new mathematical theorem and no new law is claimed.
First, the narrowing side is counted at finite levels. The constructible hierarchy runs $0,1,2,4,16,\mathbf{65536}$, and the next level is $2^{65536}$, a number of $\mathbf{19729}$ digits(Section 2).
Second, this is the core of the paper. At finite levels, selecting and taking-everything agree. They part company exactly one step past $\omega$, where they separate into $\aleph_0$ and $2^{\aleph_0}$(Section 3).
Third, the widening side was actually run. Meeting $\mathbf{48}$ dense sets in turn and extending the condition $\mathbf{24}$ times produces a real that agrees with none of the $24$ reals of the old universe—the number that agree is $\mathbf{0}$(Section 4).
Fourth, the directions are opposite. Forcing only ever adds one bit and removes nothing whatever—the old universe sits inside the new one unchanged(Section 4).
Fifth, neither alone establishes independence. Goedel in $1938$ and Cohen in $1963$, separated by $\mathbf{25}$ years, are both required before the continuum hypothesis becomes independent(Section 5).
Sixth, the size of the continuum is not free, and it is not settled either. Since $\mathrm{cf}(2^{\aleph_0})>\aleph_0$, the value $\aleph_\omega$ is forbidden, while $\aleph_1$, $\aleph_2$, $\aleph_{17}$ and $\aleph_{\omega+1}$ are all permitted(Section 6).
Scope (what is not claimed)
No new mathematical theorem and no new law is claimed—the constructible universe, forcing, the independence of the continuum hypothesis, Koenig's theorem and Easton's theorem are all standard. No set theory is built—what is used is one exponentiation, finitely many dense sets, and a diagonal argument. Independence is not proved here—neither Goedel's construction nor Cohen's is carried out. The two directions are set side by side and the reason both are needed is written down; that is all. The incompleteness theorems are not treated—Paper 189 showed that being unprovable is not a property of a sentence, and states in its own text that it does not prove Goedel's theorem. This paper draws the same line. The distinct roots of “cannot be constructed” are not recounted—Papers 179 and 163 treat those. The construction at issue there is on the side of choice and existence proofs, whereas the constructibility here is Goedel's $L$. Two different things share a word, and they are not conflated. Forcing is not formalised—what Section 4 runs is a procedure meeting finitely many dense sets in turn, an analogue of the countable case. Genuine forcing cannot be written with the tools of this paper. The consistency of $\mathrm{ZF}$ is not discussed—both techniques give relative consistency alone, and that fact is itself part of the subject here. The numbers are finite demonstrations—Section 2 runs to $6$ levels, Section 4 uses $24$ reals and $48$ dense sets, and the random seed is fixed. Relation to earlier papers: Paper 189 showed that unprovability is not a property of a sentence but a relation to an axiom system—this paper turns to the two techniques that make that relation. Paper 163 showed with Shannon and Galois that an existence theorem and a construction have distinct roots—the two here are not a pair of existence and construction but a pair of directions of construction. Paper 179 showed that “cannot be constructed” has distinct roots—this paper adds that “can be constructed” has two directions. Paper 232 counted four different claims written with the same equals sign—this paper writes that the single word “independent” holds only as the conjunction of two proofs. Paper 190 measured rare on a logarithmic scale—Section 2 here likewise reports in digits. What is added is counting the finite levels of the constructible hierarchy and obtaining $65536$ and the $19729$ digits of the level after it, locating the place where selecting and taking-everything part company as one step past $\omega$, running the dense-set procedure to demonstrate the direction of adding, counting the separators between the two techniques as three, and setting out which cardinalities are permitted and which forbidden.
1. Notation and Placement
Setting: the continuum hypothesis ($\mathrm{CH}$) asserts that there is no cardinality between $\aleph_0$ and $2^{\aleph_0}$. That it is independent of $\mathrm{ZFC}$ means that neither $\mathrm{CH}$ nor $\lnot\mathrm{CH}$ can be proved.
Question: why are two proofs needed?
The answer is that the directions are opposite. One builds a world in which $\mathrm{CH}$ holds and the other a world in which it fails—and the ways of building a world are the reverse of each other.
A caution: constructibility in this paper means Goedel's $L$. The “can or cannot be constructed” of Papers 179 and 163 is the contrast between an existence proof and an explicit construction, a different notion—one word standing for two different things.
2. The Narrowing Side: Building by Selecting
Setting: the constructible hierarchy is built by $L_{\alpha+1}=\mathrm{Def}(L_\alpha)$. $\mathrm{Def}$ takes only the subsets that a formula can name—not every subset.
Check 2.1: count the finite levels.
| Level $\alpha$ | $|L_\alpha|$ |
|---|---|
| $0$ | $0$ |
| $1$ | $1$ |
| $2$ | $2$ |
| $3$ | $4$ |
| $4$ | $16$ |
| $5$ | $\mathbf{65536}$ |
| $6$ | $2^{65536}$ ($\mathbf{19729}$ digits) |
Observation 2.2 at finite levels the nameable subsets coincide with all the subsets. A finite set has only finitely many subsets and every one of them can be written out by a formula—hence $|L_{\alpha+1}|=2^{|L_\alpha|}$.
Observation 2.3 so far not one difference between selecting and taking-everything is visible. Whether $L$ is genuinely narrow cannot be seen by looking at the finite levels.
3. The Core: They Part Company One Step Past $\omega$
Check 3.1: set $\omega$ and its successor side by side.
| Level | Size | What was done |
|---|---|---|
| $L_\omega$ | $\aleph_0$ | the union of the finite levels |
| $L_{\omega+1}$ | $\mathbf{\aleph_0}$ | only the nameable ones were taken |
| $\mathcal{P}(L_\omega)$ | $\mathbf{2^{\aleph_0}}$ | every subset was taken |
Observation 3.2 here for the first time they differ. There are only countably many formulas, so only countably many nameable subsets—while the subsets themselves number $2^{\aleph_0}$.
Observation 3.3 the place is now located—the first level at which selecting and taking-everything part company is $\omega+1$. They agree at the finite levels, still agree at $\omega$, and separate at the next step.
Observation 3.4 inside $L$, $2^{\aleph_0}=\aleph_1$. There are only nameable reals, so the reals cannot grow too numerous—it is the narrowing that makes $\mathrm{CH}$ hold.
4. The Widening Side: Building by Adding
Setting: forcing does the reverse. A finite condition, a string with finitely many bits decided, is extended while dense sets are met in turn. The union of these conditions is the new real.
Check 4.1: let the old universe hold $24$ reals. Two families of dense sets are met alternately—$D_k$, which lengthens the condition past $k$, and $E_k$, which differs from the $k$th real.
| Quantity | Value |
|---|---|
| Dense sets met | $\mathbf{48}$ |
| Extensions of the condition | $\mathbf{24}$ |
| The real obtained | $101010010110100101101100$ |
| How many of the $24$ old reals it equals | $\mathbf{0}$ |
Observation 4.2 at each step the $k$th bit is made to differ from the $k$th real, so it disagrees with every row. A real that was not in the old universe has been produced.
Observation 4.3 note the direction—each step adds one bit and removes nothing at all. The old universe sits inside the new one unchanged—the exact reverse of $L$, which selects and creates nothing.
Observation 4.4 a point to state honestly—what Section 4 ran is a procedure meeting finitely many dense sets in turn, which is the diagonal argument itself. Genuine forcing handles countably many dense sets and requires an argument for the existence of a generic extension—this paper uses the procedure as a demonstration of direction and does not claim to have formalised forcing.
5. Neither Alone Establishes Independence
Check 5.1: set the two side by side.
| Goedel ($1938$) | Cohen ($1963$) | |
|---|---|---|
| Instrument | the constructible universe $L$ | forcing, a generic extension |
| Direction | narrows (selects) | widens (adds) |
| World built | one where $\mathrm{CH}$ holds | one where $\mathrm{CH}$ fails |
| What was shown | $\mathrm{Con}(\mathrm{ZF})\Rightarrow\mathrm{Con}(\mathrm{ZFC}+\mathrm{GCH})$ | $\mathrm{Con}(\mathrm{ZF})\Rightarrow\mathrm{Con}(\mathrm{ZF}+\lnot\mathrm{CH})$ |
Observation 5.2 $\mathbf{25}$ years lie between them. And neither alone permits the word “independent”—Goedel alone yields only that $\lnot\mathrm{CH}$ is unprovable, and Cohen alone only that $\mathrm{CH}$ is.
Observation 5.3 two techniques pointing opposite ways are both required for one conclusion. Discard either and the conclusion is halved.
6. The Size of the Continuum Is Not Free, and Not Settled Either
Check 6.1: Koenig's theorem gives $\mathrm{cf}(2^{\aleph_0})>\aleph_0$, which separates the permitted values from the forbidden.
| Candidate for $2^{\aleph_0}$ | Cofinality | Verdict |
|---|---|---|
| $\aleph_1$ | $\aleph_1$ | permitted |
| $\aleph_2$ | $\aleph_2$ | permitted |
| $\aleph_{17}$ | $\aleph_{17}$ | permitted |
| $\aleph_\omega$ | $\mathbf{\aleph_0}$ | forbidden |
| $\aleph_{\omega+1}$ | $\aleph_{\omega+1}$ | permitted |
Observation 6.2 only $\aleph_\omega$ falls away. Its cofinality is $\aleph_0$, and a theorem forbids it—independence does not mean that anything at all may be inserted.
Observation 6.3 not free, and not settled. What is forbidden is exactly the cardinals of countable cofinality, and all the rest are consistent—Easton's theorem says that the whole of the rest can in fact be realised.
7. Three Separators
Observation 7.1 the first separator, absoluteness. $L$ comes out the same from whichever model it is built. A forcing extension, by contrast, depends on the partial order chosen—different worlds can be built from one starting point.
Observation 7.2 the second separator, whether sets are made. $L$ makes not one new real (only the nameable ones are admitted). Forcing makes new reals (Section 4).
Observation 7.3 the third separator, that both give relative consistency alone. Each carries the condition “if $\mathrm{ZF}$ is consistent”—the consistency of $\mathrm{ZF}$ itself cannot be shown within $\mathrm{ZF}$. Here alone the two techniques agree.
Observation 7.4 opposite on two of the three, agreeing on one—so they are neither “the same thing done twice” nor “two entirely unrelated things”.
8. Roll Call and Closing
- The finite levels(Check 2.1): $0,1,2,4,16,65536$, then $19729$ digits.
- No difference at finite levels(Observation 2.2): nameable equals all.
- They part at $\omega+1$(Check 3.1): $\aleph_0$ against $2^{\aleph_0}$.
- The narrowing makes $\mathrm{CH}$ hold(Observation 3.4): inside $L$, $2^{\aleph_0}=\aleph_1$.
- The adding side was run(Check 4.1): $48$ dense sets, $24$ extensions, $0$ agreements.
- Nothing was removed(Observation 4.3): the old universe sits inside unchanged.
- Twenty-five years apart(Check 5.1): $1938$ and $1963$.
- One alone gives half(Observation 5.2): independence requires both.
- Only $\aleph_\omega$ is forbidden(Check 6.1): its cofinality is $\aleph_0$.
- Three separators(Section 7): absoluteness, making or not, relative consistency.
Closing: the single word “independent” was not the name of one proof. It is the name of the conjunction of two—and the two pointed in opposite directions. Goedel selected out only the nameable and narrowed the universe, so that the reals could not grow too numerous. Cohen added one bit at a time and widened it, making a real that the old universe did not hold. On the narrowed side $2^{\aleph_0}=\aleph_1$; on the widened side it can be made $\aleph_2$ or $\aleph_{17}$—and $\aleph_\omega$ alone is forbidden by its cofinality. One thing separates them—writing down whether one is selecting or adding. Write it down, and the two appear as a pair rather than as rivals. Do not write it down, and one says that independence has been proved while holding only half of it.
References
- Goedel, K. (1938). The Consistency of the Axiom of Choice and of the Generalized Continuum-Hypothesis. Proceedings of the National Academy of Sciences, 24(12).
- Cohen, P. J. (1963). The Independence of the Continuum Hypothesis. Proceedings of the National Academy of Sciences, 50(6).
- Cohen, P. J. (1964). The Independence of the Continuum Hypothesis, II. Proceedings of the National Academy of Sciences, 51(1).
- Kunen, K. (2011). Set Theory. London: College Publications.
- Jech, T. (2003). Set Theory, 3rd millennium edition. Berlin: Springer.
- Easton, W. B. (1970). Powers of Regular Cardinals. Annals of Mathematical Logic, 1(2).
- Devlin, K. J. (1984). Constructibility. Berlin: Springer.
- Yamagishi, Y. (2026). “A Good Code Exists” and “Here Is a Good Code” Are Different Sentences [Paper 163]. Zenodo. 10.5281/zenodo.22092677
- Yamagishi, Y. (2026). “Cannot Be Constructed” Has Distinct Roots [Paper 179]. Zenodo. 10.5281/zenodo.22098419
- Yamagishi, Y. (2026). “Unprovable” Is Not a Property of a Sentence [Paper 189]. Zenodo. 10.5281/zenodo.22105830
- Yamagishi, Y. (2026). “Rare” Is Measured on a Logarithmic Scale [Paper 190]. Zenodo. 10.5281/zenodo.22106182
- Yamagishi, Y. (2026). Learn from a Book with the Proofs Removed and the Equals Sign Loses Its Type [Paper 232]. Zenodo. 10.5281/zenodo.22127528
Reference
Yamagishi, Y. (2026). There Are Two Roads to Independence: One Narrows the Universe and the Other Widens It [Paper 259]. YOI: https://www.yamagishiyuuki.com/yoi/259
DOI: 10.5281/zenodo.22169162
ORCID: 0009-0006-8587-3450
PDF (English) · PDF (Japanese) · Date proof of the English PDF (OpenTimestamps .ots) · SHA256SUMS
The record page for this paper (in Japanese, with the PDFs in each language and the date proof): YOI 259. This paper is a preprint and has not been peer reviewed.