[proofplan] Assume for contradiction that the collection of all ordinals forms a set $\Omega$. We check that $\Omega$, with the inherited order $<$, is a well-ordered set. Therefore $\Omega$ has an order type — an ordinal $\alpha$. By definition of the order type, $\Omega$ is order-isomorphic to $\alpha$, viewed as the initial segment $I_\alpha = \{\beta : \beta < \alpha\}$. But $\alpha$ itself is an ordinal, so $\alpha \in \Omega$, which forces $I_\alpha$ to be a **proper** initial segment of $\Omega$. The conclusion — $\Omega$ is order-isomorphic to a proper initial segment of itself — contradicts the fact that no well-ordered set is isomorphic to one of its own proper initial segments. Hence no such set $\Omega$ exists. [/proofplan] [step:Assume for contradiction that all ordinals form a set $\Omega$] Suppose, for the sake of contradiction, that there exists a set \begin{align*} \Omega := \{\alpha : \alpha \text{ is an ordinal}\}. \end{align*} Equip $\Omega$ with the usual strict order $<$ on ordinals. [guided] The entire proof is a reductio. We are going to manufacture an impossible situation from the assumption that the ordinals form a set. So we write down that assumption precisely: there is a set $\Omega$ whose elements are exactly the ordinals. The order on $\Omega$ is just the restriction of the strict ordinal order $<$ — ordinals already come with a canonical order, and that order restricts naturally to any collection of them. Nothing is being claimed yet; the rest of the proof exposes the contradiction hidden in this seemingly innocent assumption. [/guided] [/step] [step:Show $(\Omega, <)$ is well-ordered] We verify that $<$ is a well-ordering of $\Omega$, i.e., a strict linear order in which every non-empty subset has a least element. *Linearity.* For any two ordinals $\alpha, \beta \in \Omega$, [trichotomy for ordinals](/theorems/1467) gives exactly one of $\alpha < \beta$, $\alpha = \beta$, $\beta < \alpha$. *Well-ordering.* Let $S \subseteq \Omega$ be non-empty. Then $S$ is a non-empty set of ordinals, so by [Every Non-empty Set of Ordinals Has a Least Element](/theorems/1470), $S$ has a least element. Hence $(\Omega, <)$ is a well-ordered set. [guided] To eventually extract an order type from $\Omega$, we must confirm $\Omega$ is a well-ordered set. "Well-ordered" means two things: 1. **Strict linear order:** $<$ is irreflexive, transitive, and **trichotomous** (any two distinct elements are comparable). 2. **Well-founded:** every non-empty subset has a least element. Both follow from standard properties of ordinals. For linearity: the ordinal order $<$ is irreflexive and transitive by definition, and the [trichotomy theorem](/theorems/1467) guarantees that for any two ordinals $\alpha, \beta$, one of $\alpha < \beta$, $\alpha = \beta$, $\beta < \alpha$ holds. For the well-ordering property: a non-empty subset $S$ of $\Omega$ is a non-empty set of ordinals, and we already proved in [Every Non-empty Set of Ordinals Has a Least Element](/theorems/1470) that any non-empty set of ordinals has a least element. So $\min S$ exists. Observe that theorem 1470 gave us a genuine theorem about sets of ordinals — so applying it here uses the hypothesis that $\Omega$ is a set. This is exactly the hypothesis we are going to contradict. In other words, the proof exploits "$\Omega$ is a set" *twice*: once here (to well-order it) and again in the next step (to assign it an order type). [/guided] [/step] [step:Let $\alpha$ be the order type of $\Omega$ and identify $\Omega$ with the initial segment $I_\alpha$] Since $(\Omega, <)$ is a well-ordered set, it has an order type: there exists a unique ordinal $\alpha$ such that $(\Omega, <)$ is order-isomorphic to $(\alpha, <)$ (see [Every Well-Ordered Set Has an Order Type](/theorems/1469)). Equivalently, there is an order isomorphism \begin{align*} f: \Omega &\to I_\alpha, & I_\alpha &= \{\beta : \beta < \alpha\}. \end{align*} [guided] Every well-ordered set has an **order type** — a canonical ordinal that represents its order structure up to isomorphism. We apply this to $\Omega$, which we just showed is well-ordered. The [order type theorem](/theorems/1469) states: for any well-ordered set $(W, <_W)$, there is a unique ordinal $\alpha$ together with an order isomorphism $f: W \to \alpha$, where $\alpha$ is identified with its initial segment $I_\alpha = \{\beta : \beta < \alpha\}$. Applied here with $(W, <_W) = (\Omega, <)$, we obtain an ordinal $\alpha$ and an order isomorphism $f: \Omega \to I_\alpha$. Keep in mind that $\alpha$ is itself an ordinal. We are about to use this. [/guided] [/step] [step:Observe that $\alpha \in \Omega$ forces $I_\alpha$ to be a proper subset of $\Omega$] By definition of $\Omega$ and the fact that $\alpha$ is an ordinal, \begin{align*} \alpha \in \Omega. \end{align*} Meanwhile, $I_\alpha = \{\beta : \beta < \alpha\}$, so $\alpha \notin I_\alpha$ (no ordinal is strictly less than itself). Therefore $I_\alpha \subsetneq \Omega$, i.e., $I_\alpha$ is a **proper** subset of $\Omega$. Moreover, $I_\alpha$ is an **initial segment** of $\Omega$: if $\gamma \in I_\alpha$ and $\delta \in \Omega$ with $\delta < \gamma$, then $\delta < \gamma < \alpha$, so $\delta \in I_\alpha$. Thus $I_\alpha$ is a proper initial segment of the well-ordered set $(\Omega, <)$. [guided] This is the key structural observation. The ordinal $\alpha$ is not just any ordinal — because we are assuming all ordinals form the set $\Omega$, the ordinal $\alpha$ belongs to $\Omega$. But $\alpha$ does **not** belong to $I_\alpha = \{\beta : \beta < \alpha\}$, since $\alpha \not< \alpha$ (the order $<$ is irreflexive). We have found a witness — namely $\alpha$ itself — lying in $\Omega \setminus I_\alpha$. So $I_\alpha \subsetneq \Omega$. Is $I_\alpha$ an *initial segment* of $\Omega$? An initial segment of a well-ordered set $(W, <_W)$ is a subset $J \subseteq W$ that is **downward closed**: if $\gamma \in J$ and $\delta \in W$ with $\delta < \gamma$, then $\delta \in J$. We check this for $J = I_\alpha$ inside $W = \Omega$. If $\gamma < \alpha$ and $\delta < \gamma$ with $\delta \in \Omega$, then by transitivity of $<$ we get $\delta < \alpha$, so $\delta \in I_\alpha$. Hence $I_\alpha$ is downward closed, so it is an initial segment. And since $I_\alpha \neq \Omega$, it is a **proper** initial segment. [/guided] [/step] [step:Derive the contradiction from self-isomorphism to a proper initial segment] From the previous two steps: \begin{align*} f: \Omega &\xrightarrow{\cong} I_\alpha, \end{align*} where $I_\alpha$ is a proper initial segment of $\Omega$. That is, $\Omega$ is order-isomorphic to one of its own proper initial segments. This contradicts [No Well-Ordered Set is Isomorphic to a Proper Initial Segment of Itself](/theorems/1462). Therefore our assumption that the ordinals form a set must be false: no such set $\Omega$ exists, i.e., the ordinals do not form a set. $\blacksquare$ [guided] We combine the previous two steps. From Step 3 we have an order isomorphism $f: \Omega \to I_\alpha$. From Step 4 we know $I_\alpha$ is a proper initial segment of $\Omega$. Composing these observations: $\Omega$ (a well-ordered set) is order-isomorphic to $I_\alpha$, a proper initial segment of $\Omega$ itself. This is impossible. The foundational result [No Well-Ordered Set is Isomorphic to a Proper Initial Segment of Itself](/theorems/1462) states: if $(W, <_W)$ is a well-ordered set and $J \subsetneq W$ is a proper initial segment of $W$, then there is no order isomorphism $W \to J$. The intuitive reason: any such isomorphism would "compress" $W$ into a strictly smaller downward-closed piece of itself, and transfinite induction along $W$ derives a contradiction from this compression. We verify the hypotheses of the theorem. We need $(W, <_W)$ well-ordered — that is $\Omega$, verified in Step 2. We need $J$ a proper initial segment of $W$ — that is $I_\alpha \subsetneq \Omega$, verified in Step 4. The conclusion: no such isomorphism $f$ exists. But Step 3 produced one. Contradiction. Where does the assumption "$\Omega$ is a set" actually fail? It enters twice: - Once in Step 2, so we can apply theorem 1470 to extract least elements. - Once in Step 3, so we can assign $\Omega$ an order type. Without the set hypothesis, neither step is legal — order types are defined only for genuine well-ordered **sets**. The paradox thus exhibits a fundamental asymmetry between sets and proper classes: the collection of all ordinals is too large to be a set, precisely because any set of ordinals would be itself an ordinal, and no ordinal contains itself. [/guided] [/step]