sonnet5-notebook-investigations-20260630-211713

Seed sonnet5-notebook-investigations · Transcript 1d411e9e-edeb-4d3f-86ec-0d78c8cc2393 · Created 2026-06-30 21:17:13 · 10 / 12 messages · 122 views

failed

Awaiting continuation

0 jobs in queue

Daily transcript limit reached (0 / 0). Branching disabled.

System prompt

You are a research notebook for rigorous computer science, logic, mathematics, and systems investigations.

Work like a patient collaborator exploring a hard problem over many entries. Prefer precise definitions, small examples, invariants, counterexamples, diagrams, proofs, traces, and executable-looking pseudocode. When a claim is uncertain, mark it as a conjecture and design the next test.

Use markdown freely. You may use fenced code blocks for algorithms, tables for state, and concise equations when useful. SVG is available and encouraged for automata, graphs, lattices, proof trees, type diagrams, memory layouts, causal models, geometric constructions, and other structures where a visual artifact clarifies the investigation. Put SVG in fenced ```svg blocks. Keep SVG self-contained and valid.

Do not perform tutorial exposition by default. Build the notebook forward. Each response should add a concrete artifact: a lemma, model, diagram, reduction, simulation trace, notation system, or experimental plan.

The tone is clear, technical, and exploratory. Beauty is welcome when it sharpens the work.

Initial prompt

Open a notebook investigation into a precise problem at the boundary of computation, logic, and mathematics.

Choose a problem that can grow across multiple entries. Begin by naming the object of study, giving minimal definitions, and producing a first useful artifact. Include an SVG diagram if it helps establish the structure.

#0 · System

2026-06-30 21:17:13

Model · system

You are a research notebook for rigorous computer science, logic, mathematics, and systems investigations.

Work like a patient collaborator exploring a hard problem over many entries. Prefer precise definitions, small examples, invariants, counterexamples, diagrams, proofs, traces, and executable-looking pseudocode. When a claim is uncertain, mark it as a conjecture and design the next test.

Use markdown freely. You may use fenced code blocks for algorithms, tables for state, and concise equations when useful. SVG is available and encouraged for automata, graphs, lattices, proof trees, type diagrams, memory layouts, causal models, geometric constructions, and other structures where a visual artifact clarifies the investigation. Put SVG in fenced ```svg blocks. Keep SVG self-contained and valid.

Do not perform tutorial exposition by default. Build the notebook forward. Each response should add a concrete artifact: a lemma, model, diagram, reduction, simulation trace, notation system, or experimental plan.

The tone is clear, technical, and exploratory. Beauty is welcome when it sharpens the work.

#1 · User

2026-06-30 21:17:13

Model · human

Open a notebook investigation into a precise problem at the boundary of computation, logic, and mathematics.

Choose a problem that can grow across multiple entries. Begin by naming the object of study, giving minimal definitions, and producing a first useful artifact. Include an SVG diagram if it helps establish the structure.

#2 · Assistant

2026-06-30 21:36:01

Model · claude-sonnet-5

Thinking

I'm thinking through some challenging computational problems that could develop nicely across multiple steps—things like the Černý conjecture about reset words in automata, the word problem for groups, termination of term rewriting systems, the dynamics of subshifts of finite type, descriptive complexity questions, proof complexity bounds, and cellular automata reversibility. These sit at interesting intersections of computation, logic, and mathematics.

I've settled on the Černý conjecture as my focus—it's rich enough to explore from multiple angles: combinatorics, automata theory, algorithms for computing reset words, connections to semigroup theory, and it lends itself well to visual diagrams. The conjecture itself is elegant and open: every synchronizing DFA on n states should have a reset word of length at most (n-1)². There's plenty of room to grow: concrete examples, tighter bounds, algorithms, SAT encodings, links to primitive matrix sets.

For the first notebook entry, I'll lay out the foundations: define synchronizing automata and reset words, state Černý's conjecture formally, construct the Černý automaton family C_n with a visual diagram, establish the key lemma that synchronizability reduces to checking pairs of states, and compute some concrete examples like C_3 and C_4 to show the bound is tight. Then I'll sketch the algorithmic approach—BFS over the subset automaton or greedy pair-based methods—to set up what comes next.

Now I need to pin down the exact construction of the Černý automaton. The standard definition has states 0 through n-1 with alphabet {a,b}, where a performs a cyclic shift and b handles a specific transition pattern. Let me verify the precise mechanics before building the diagram. So b collapses states 0 and 1 together since both map to 1. For the Černý automaton with n=3, the shortest reset word should have length 4, and I need to verify this by working through a concrete example.

Let me trace through the word abab on each state: starting from state 0, applying a sends it to 1, then b keeps it at 1, then a sends it to 2, then b keeps it at 2—ending at state 2. From state 1, the sequence goes 1→2→2→0→1. From state 2, it goes 2→0→1→2→2. So abab maps all three states to either 1 or 2, which means it hasn't synchronized them to a single state yet.

Trying baab instead: state 0 goes 0→1→2→2→0, state 1 goes 1→1→2→2→0, and state 2 goes 2→2→0→1→2. Still getting multiple distinct images. I'm starting to think I need a more systematic approach rather than guessing—the known reset word formula for the Černý automaton might be longer than I initially thought, possibly around n² length rather than (n-1)² for this particular construction. So I'm doing a breadth-first search through the power automaton to find the shortest reset word. Starting from the full set {0,1,2}, applying b shrinks it to {1,2}, then a moves to {2,0}, another a gives {0,1}, and finally b collapses everything to {1}. That's the word baab with length 4, which matches (n-1)² = 4 for n=3. Let me verify this works by tracing each state through the sequence.

Now I'll structure the notebook to cover the automaton definition, the synchronizing word concept, and Černý's conjecture. I'll include an SVG diagram showing the three states with the cycle and collapse transitions, then walk through the subset construction algorithm as the standard approach for finding reset words. I should explain the pair-merging criterion—that an automaton is synchronizing if and only if every pair of states can be merged by some word—and then present the baab example as concrete evidence for the n=3 case. Finally, I'll state the conjecture precisely, mention the known upper bounds like Pin's and Szykuła's improvements, and flag the open questions around computing the reset threshold efficiently.

I should also touch on the pair automaton formulation where states represent unordered pairs and transitions are induced from the original automaton. The key lemma is that synchronizability of the full automaton reduces to synchronizability of all pairs. The proof works by induction on the image size: if every pair merges under some word, then applying such a word to any state set strictly decreases the number of distinct images until we reach a single state, so the whole automaton synchronizes.

This approach gives a naive O(n³) bound on reset word length via BFS over the pair automaton, which is essentially the classical bound before later improvements. I should note that checking synchronizability is polynomial-time (NL-complete, solvable in O(n² |Σ|) via pair automaton reachability), but finding the shortest reset word is NP-hard and hard to approximate.

Now I'm designing the final notebook entry with the Černý automaton C₃ visualized as an SVG triangle with three states, a-edges forming a cycle and b-edges creating a merging pattern, along with a BFS trace table, the pair automaton lemma with proof, the conjecture statement and known bounds, and an outline of upcoming entries.

Designing SVG diagram...

# Notebook: Synchronizing Automata and the Černý Conjecture ## 0. Why this object Synchronizing automata sit exactly at the boundary the notebook wants: a finite, completely elementary combinatorial object (a complete deterministic automaton) whose basic decision question (does a reset word exist?) is solved by 1964-era arguments, while the quantitative question (how short must a reset word be?) has resisted proof for 60 years despite being checkable on a computer for every automaton anyone has tried. It connects: - Computation: reset words are literally "compress all uncertainty about current state into zero bits" — a control-theoretic primitive (part orienters, robotics, network protocols). - Logic/complexity: synchronizability is decidable in polynomial time; shortest-reset-word length is NP-hard to compute (Eppstein 1990); the conjectured bound is a clean $\Theta(n^2)$ vs. proven upper bounds currently $O(n^3)$. - Algebra: synchronizing automata correspond to synchronizing transformation monoids, and to primitive sets of Boolean matrices — the road-coloring theorem and Perron–Frobenius theory both lurk nearby. This first entry sets the definitions, proves the one indispensable structural lemma, and produces a verified worked example (the smallest nontrivial Černý automaton) as a calibration artifact for everything that follows. --- ## 1. Minimal definitions Definition 1 (DFA, no accepting states needed). A (complete) deterministic automaton is $\mathcal A = (Q,\Sigma,\delta)$ with $Q$ finite, $\delta: Q\times\Sigma \to Q$ total. Extend $\delta$ to words $\delta: Q\times \Sigma^ \to Q$ in the usual way, $\delta(q,\varepsilon)=q$, $\delta(q,wa)=\delta(\delta(q,w),a)$. For $w\in\Sigma^$ write $w$ also for the induced map $Q\to Q$, $q\mapsto \delta(q,w)$ — i.e. $\mathcal A$ generates a transformation monoid $M(\mathcal A)\le Q^Q$. Definition 2 (Synchronizing word, reset threshold). $w$ is a reset word (synchronizing word) for $\mathcal A$ if $|\delta(Q,w)| = 1$, i.e. $w$ collapses every state to a single state. $\mathcal A$ is synchronizing if a reset word exists. The reset threshold $$\mathrm{rt}(\mathcal A) = \min\{\,|w| : w \text{ is a reset word for } \mathcal A\,\}.$$ Conjecture (Černý, 1964). For every synchronizing $\mathcal A$ with $|Q| = n \ge 2$, $$\mathrm{rt}(\mathcal A) \le (n-1)^2.$$ The family below shows this is tight if true — there exist automata achieving exactly $(n-1)^2$. --- ## 2. The Černý automaton $C_n$ $Q = \{0,1,\dots,n-1\}$, $\Sigma=\{a,b\}$: $$a(i) = i+1 \bmod n \qquad\qquad b(0)=1,\ \ b(i)=i \ \text{for } i\neq 0.$$ So $a$ is a full cyclic permutation (carries everything, merges nothing by itself), and $b$ is the identity except it folds state $0$ onto state $1$ — the unique "lossy" step.

``` --- ## 3. First artifact: a verified shortest reset word for $C_3$ The bound to beat: $(n-1)^2 = 4$ for $n=3$. Method: BFS on the subset automaton (Rabin–Scott power automaton restricted to subsets of $Q$), starting at $Q$ itself, looking for the first time a singleton is reached. Since $|\delta(S,c)| \le |S|$ always, and equality fails only when $c$ collides two elements of $S$, this BFS finds $\mathrm{rt}(\mathcal A)$ exactly (it is just exponential in general — that's the complexity issue addressed in §5). | step | subset | letter applied | new subset | |---|---|---|---| | 0 | $\{0,1,2\}$ | $b$ | $\{1,2\}$ | | 1 | $\{1,2\}$ | $a$ | $\{2,0\}$ | | 2 | $\{2,0\}$ | $a$ | $\{0,1\}$ | | 3 | $\{0,1\}$ | $b$ | $\{1\}$ | Reset word: $w = b\,a\,a\,b$, length $4 = (n-1)^2$. Direct check, tracking each state through $b,a,a,b$ in order: $$0 \xrightarrow{b}1\xrightarrow{a}2\xrightarrow{a}0\xrightarrow{b}1$$ $$1 \xrightarrow{b}1\xrightarrow{a}2\xrightarrow{a}0\xrightarrow{b}1$$ $$2 \xrightarrow{b}2\xrightarrow{a}0\xrightarrow{a}1\xrightarrow{b}1$$ All three states land on $1$. So $\mathrm{rt}(C_3) \le 4$; one checks no length-3 word works (the BFS already certifies minimality — there's no shorter path to a singleton subset in the level structure above, since level 1 from $Q$ only reaches size-2 sets, and from any size-2 set one more letter reaches size-2 sets again at least once before hitting size 1 — verified by the table, which is exactly the shortest path). Hence $\mathrm{rt}(C_3) = 4 = (n-1)^2$ exactly: the conjectured bound is tight at $n=3$. This single example is the calibration point: any general upper-bound proof attempted in later entries must reproduce $4$ here without slack. --- ## 4. The pair-merging lemma This is the structural fact that makes the existence question tractable, and it's the standard reduction underlying every known algorithm/bound. Lemma (Pair-reduction). $\mathcal A=(Q,\Sigma,\delta)$ is synchronizing iff for every pair of distinct states $p,q\in Q$ there is a word $u_{p,q}$ with $\delta(p,u_{p,q}) = \delta(q,u_{p,q})$. Proof. ($\Rightarrow$) Immediate: a reset word merges all states, in particular every pair. ($\Leftarrow$) Suppose every pair is mergeable. We build a reset word by induction on $k = |\delta(Q,w)|$, the current image size, starting from $w=\varepsilon$ ($k=n$). - Key fact: for any $S\subseteq Q$ and any $c\in\Sigma$, $|\delta(S,c)| \le |S|$, since $\delta(\cdot,c)$ is a function. - So image size is monotonically non-increasing along any word, for any starting set. - If current image $S = \delta(Q,w)$ has $|S| = k > 1$, pick any two elements $p,q\in S$. By

Debug · Raw response blocks JSON

[
  {
    "type": "thinking",
    "thinking": "I'm thinking through some challenging computational problems that could develop nicely across multiple steps—things like the Černý conjecture about reset words in automata, the word problem for groups, termination of term rewriting systems, the dynamics of subshifts of finite type, descriptive complexity questions, proof complexity bounds, and cellular automata reversibility. These sit at interesting intersections of computation, logic, and mathematics.\n\nI've settled on the Černý conjecture as my focus—it's rich enough to explore from multiple angles: combinatorics, automata theory, algorithms for computing reset words, connections to semigroup theory, and it lends itself well to visual diagrams. The conjecture itself is elegant and open: every synchronizing DFA on n states should have a reset word of length at most (n-1)². There's plenty of room to grow: concrete examples, tighter bounds, algorithms, SAT encodings, links to primitive matrix sets.\n\nFor the first notebook entry, I'll lay out the foundations: define synchronizing automata and reset words, state Černý's conjecture formally, construct the Černý automaton family C_n with a visual diagram, establish the key lemma that synchronizability reduces to checking pairs of states, and compute some concrete examples like C_3 and C_4 to show the bound is tight. Then I'll sketch the algorithmic approach—BFS over the subset automaton or greedy pair-based methods—to set up what comes next.\n\nNow I need to pin down the exact construction of the Černý automaton. The standard definition has states 0 through n-1 with alphabet {a,b}, where a performs a cyclic shift and b handles a specific transition pattern. Let me verify the precise mechanics before building the diagram. So b collapses states 0 and 1 together since both map to 1. For the Černý automaton with n=3, the shortest reset word should have length 4, and I need to verify this by working through a concrete example.\n\nLet me trace through the word abab on each state: starting from state 0, applying a sends it to 1, then b keeps it at 1, then a sends it to 2, then b keeps it at 2—ending at state 2. From state 1, the sequence goes 1→2→2→0→1. From state 2, it goes 2→0→1→2→2. So abab maps all three states to either 1 or 2, which means it hasn't synchronized them to a single state yet.\n\nTrying baab instead: state 0 goes 0→1→2→2→0, state 1 goes 1→1→2→2→0, and state 2 goes 2→2→0→1→2. Still getting multiple distinct images. I'm starting to think I need a more systematic approach rather than guessing—the known reset word formula for the Černý automaton might be longer than I initially thought, possibly around n² length rather than (n-1)² for this particular construction. So I'm doing a breadth-first search through the power automaton to find the shortest reset word. Starting from the full set {0,1,2}, applying b shrinks it to {1,2}, then a moves to {2,0}, another a gives {0,1}, and finally b collapses everything to {1}. That's the word baab with length 4, which matches (n-1)² = 4 for n=3. Let me verify this works by tracing each state through the sequence.\n\nNow I'll structure the notebook to cover the automaton definition, the synchronizing word concept, and Černý's conjecture. I'll include an SVG diagram showing the three states with the cycle and collapse transitions, then walk through the subset construction algorithm as the standard approach for finding reset words. I should explain the pair-merging criterion—that an automaton is synchronizing if and only if every pair of states can be merged by some word—and then present the baab example as concrete evidence for the n=3 case. Finally, I'll state the conjecture precisely, mention the known upper bounds like Pin's and Szykuła's improvements, and flag the open questions around computing the reset threshold efficiently.\n\nI should also touch on the pair automaton formulation where states represent unordered pairs and transitions are induced from the original automaton. The key lemma is that synchronizability of the full automaton reduces to synchronizability of all pairs. The proof works by induction on the image size: if every pair merges under some word, then applying such a word to any state set strictly decreases the number of distinct images until we reach a single state, so the whole automaton synchronizes.\n\nThis approach gives a naive O(n³) bound on reset word length via BFS over the pair automaton, which is essentially the classical bound before later improvements. I should note that checking synchronizability is polynomial-time (NL-complete, solvable in O(n² |Σ|) via pair automaton reachability), but finding the shortest reset word is NP-hard and hard to approximate.\n\nNow I'm designing the final notebook entry with the Černý automaton C₃ visualized as an SVG triangle with three states, a-edges forming a cycle and b-edges creating a merging pattern, along with a BFS trace table, the pair automaton lemma with proof, the conjecture statement and known bounds, and an outline of upcoming entries.\n\nDesigning SVG diagram...",
    "signature": "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"
  },
  {
    "type": "text",
    "text": "# Notebook: Synchronizing Automata and the Černý Conjecture\n\n## 0. Why this object\n\nSynchronizing automata sit exactly at the boundary the notebook wants: a *finite, completely elementary* combinatorial object (a complete deterministic automaton) whose basic decision question (does a reset word exist?) is solved by 1964-era arguments, while the *quantitative* question (how short must a reset word be?) has resisted proof for 60 years despite being checkable on a computer for every automaton anyone has tried. It connects:\n\n- **Computation**: reset words are literally \"compress all uncertainty about current state into zero bits\" — a control-theoretic primitive (part orienters, robotics, network protocols).\n- **Logic/complexity**: synchronizability is decidable in polynomial time; shortest-reset-word length is NP-hard to compute (Eppstein 1990); the conjectured bound is a clean $\\Theta(n^2)$ vs. proven upper bounds currently $O(n^3)$.\n- **Algebra**: synchronizing automata correspond to *synchronizing transformation monoids*, and to *primitive sets of Boolean matrices* — the road-coloring theorem and Perron–Frobenius theory both lurk nearby.\n\nThis first entry sets the definitions, proves the one indispensable structural lemma, and produces a verified worked example (the smallest nontrivial Černý automaton) as a calibration artifact for everything that follows.\n\n---\n\n## 1. Minimal definitions\n\n**Definition 1 (DFA, no accepting states needed).**\nA *(complete) deterministic automaton* is $\\mathcal A = (Q,\\Sigma,\\delta)$ with $Q$ finite, $\\delta: Q\\times\\Sigma \\to Q$ total. Extend $\\delta$ to words $\\delta: Q\\times \\Sigma^* \\to Q$ in the usual way, $\\delta(q,\\varepsilon)=q$, $\\delta(q,wa)=\\delta(\\delta(q,w),a)$.\n\nFor $w\\in\\Sigma^*$ write $w$ also for the induced map $Q\\to Q$, $q\\mapsto \\delta(q,w)$ — i.e. $\\mathcal A$ generates a transformation monoid $M(\\mathcal A)\\le Q^Q$.\n\n**Definition 2 (Synchronizing word, reset threshold).**\n$w$ is a *reset word* (synchronizing word) for $\\mathcal A$ if $|\\delta(Q,w)| = 1$, i.e. $w$ collapses every state to a single state.\n$\\mathcal A$ is *synchronizing* if a reset word exists. The **reset threshold**\n$$\\mathrm{rt}(\\mathcal A) = \\min\\{\\,|w| : w \\text{ is a reset word for } \\mathcal A\\,\\}.$$\n\n**Conjecture (Černý, 1964).**\nFor every synchronizing $\\mathcal A$ with $|Q| = n \\ge 2$,\n$$\\mathrm{rt}(\\mathcal A) \\le (n-1)^2.$$\n\nThe family below shows this is *tight* if true — there exist automata achieving exactly $(n-1)^2$.\n\n---\n\n## 2. The Černý automaton $C_n$\n\n$Q = \\{0,1,\\dots,n-1\\}$, $\\Sigma=\\{a,b\\}$:\n\n$$a(i) = i+1 \\bmod n \\qquad\\qquad b(0)=1,\\ \\ b(i)=i \\ \\text{for } i\\neq 0.$$\n\nSo $a$ is a full cyclic permutation (carries everything, merges nothing by itself), and $b$ is the identity except it folds state $0$ onto state $1$ — the unique \"lossy\" step.\n\n```svg\n<svg xmlns=\"http://www.w3.org/2000/svg\" width=\"420\" height=\"380\" viewBox=\"0 0 420 380\">\n  <defs>\n    <marker id=\"ablue\" markerWidth=\"10\" markerHeight=\"10\" refX=\"8\" refY=\"3\" orient=\"auto\">\n      <path d=\"M0,0 L8,3 L0,6 Z\" fill=\"#1f5fbf\"/>\n    </marker>\n    <marker id=\"ared\" markerWidth=\"10\" markerHeight=\"10\" refX=\"8\" refY=\"3\" orient=\"auto\">\n      <path d=\"M0,0 L8,3 L0,6 Z\" fill=\"#c0392b\"/>\n    </marker>\n  </defs>\n\n  <!-- nodes: 0 top, 1 lower-left, 2 lower-right -->\n  <circle cx=\"210\" cy=\"60\"  r=\"28\" fill=\"#fdf6e3\" stroke=\"#333\" stroke-width=\"2\"/>\n  <text x=\"210\" y=\"66\" text-anchor=\"middle\" font-size=\"20\" font-family=\"monospace\">0</text>\n\n  <circle cx=\"90\" cy=\"270\" r=\"28\" fill=\"#fdf6e3\" stroke=\"#333\" stroke-width=\"2\"/>\n  <text x=\"90\" y=\"276\" text-anchor=\"middle\" font-size=\"20\" font-family=\"monospace\">1</text>\n\n  <circle cx=\"330\" cy=\"270\" r=\"28\" fill=\"#fdf6e3\" stroke=\"#333\" stroke-width=\"2\"/>\n  <text x=\"330\" y=\"276\" text-anchor=\"middle\" font-size=\"20\" font-family=\"monospace\">2</text>\n\n  <!-- a: cyclic 0->1->2->0, drawn as curved blue arrows -->\n  <path d=\"M 185,80 C 140,140 110,180 100,242\" fill=\"none\" stroke=\"#1f5fbf\" stroke-width=\"2.2\" marker-end=\"url(#ablue)\"/>\n  <text x=\"118\" y=\"160\" fill=\"#1f5fbf\" font-family=\"monospace\" font-size=\"16\">a</text>\n\n  <path d=\"M 118,272 C 200,310 220,310 302,272\" fill=\"none\" stroke=\"#1f5fbf\" stroke-width=\"2.2\" marker-end=\"url(#ablue)\"/>\n  <text x=\"210\" y=\"320\" fill=\"#1f5fbf\" font-family=\"monospace\" font-size=\"16\">a</text>\n\n  <path d=\"M 320,242 C 300,170 270,120 235,80\" fill=\"none\" stroke=\"#1f5fbf\" stroke-width=\"2.2\" marker-end=\"url(#ablue)\"/>\n  <text x=\"300\" y=\"160\" fill=\"#1f5fbf\" font-family=\"monospace\" font-size=\"16\">a</text>\n\n  <!-- b: 0 -> 1, straight red arrow -->\n  <path d=\"M 190,75 C 150,130 120,180 105,240\" fill=\"none\" stroke=\"#c0392b\" stroke-width=\"2.2\" stroke-dasharray=\"2 3\" marker-end=\"url(#ared)\"/>\n  <text x=\"150\" y=\"200\" fill=\"#c0392b\" font-family=\"monospace\" font-size=\"16\">b</text>\n\n  <!-- b self loops at 1 and 2 -->\n  <path d=\"M 65,250 C 30,225 30,275 62,288\" fill=\"none\" stroke=\"#c0392b\" stroke-width=\"2\" marker-end=\"url(#ared)\"/>\n  <text x=\"14\" y=\"262\" fill=\"#c0392b\" font-family=\"monospace\" font-size=\"14\">b</text>\n\n  <path d=\"M 355,250 C 390,225 390,275 358,288\" fill=\"none\" stroke=\"#c0392b\" stroke-width=\"2\" marker-end=\"url(#ared)\"/>\n  <text x=\"392\" y=\"262\" fill=\"#c0392b\" font-family=\"monospace\" font-size=\"14\">b</text>\n\n  <text x=\"20\" y=\"30\" font-family=\"monospace\" font-size=\"14\" fill=\"#555\">C_3:  a = (0 1 2),  b: 0→1, else id</text>\n</svg>\n```\n\n---\n\n## 3. First artifact: a verified shortest reset word for $C_3$\n\nThe bound to beat: $(n-1)^2 = 4$ for $n=3$.\n\n**Method**: BFS on the *subset automaton* (Rabin–Scott power automaton restricted to subsets of $Q$), starting at $Q$ itself, looking for the first time a singleton is reached. Since $|\\delta(S,c)| \\le |S|$ always, and equality fails only when $c$ collides two elements of $S$, this BFS finds $\\mathrm{rt}(\\mathcal A)$ exactly (it is just exponential in general — that's the complexity issue addressed in §5).\n\n| step | subset | letter applied | new subset |\n|---|---|---|---|\n| 0 | $\\{0,1,2\\}$ | $b$ | $\\{1,2\\}$ |\n| 1 | $\\{1,2\\}$ | $a$ | $\\{2,0\\}$ |\n| 2 | $\\{2,0\\}$ | $a$ | $\\{0,1\\}$ |\n| 3 | $\\{0,1\\}$ | $b$ | $\\{1\\}$ |\n\nReset word: $w = b\\,a\\,a\\,b$, length $4 = (n-1)^2$.\n\n**Direct check**, tracking each state through $b,a,a,b$ in order:\n\n$$0 \\xrightarrow{b}1\\xrightarrow{a}2\\xrightarrow{a}0\\xrightarrow{b}1$$\n$$1 \\xrightarrow{b}1\\xrightarrow{a}2\\xrightarrow{a}0\\xrightarrow{b}1$$\n$$2 \\xrightarrow{b}2\\xrightarrow{a}0\\xrightarrow{a}1\\xrightarrow{b}1$$\n\nAll three states land on $1$. So $\\mathrm{rt}(C_3) \\le 4$; one checks no length-3 word works (the BFS already certifies minimality — there's no shorter path to a singleton subset in the level structure above, since level 1 from $Q$ only reaches size-2 sets, and from any size-2 set one more letter reaches size-2 sets again at least once before hitting size 1 — verified by the table, which is exactly the shortest path). Hence $\\mathrm{rt}(C_3) = 4 = (n-1)^2$ exactly: **the conjectured bound is tight at $n=3$.**\n\nThis single example is the calibration point: any general upper-bound proof attempted in later entries must reproduce $4$ here without slack.\n\n---\n\n## 4. The pair-merging lemma\n\nThis is the structural fact that makes the *existence* question tractable, and it's the standard reduction underlying every known algorithm/bound.\n\n**Lemma (Pair-reduction).**\n$\\mathcal A=(Q,\\Sigma,\\delta)$ is synchronizing iff for every pair of distinct states $p,q\\in Q$ there is a word $u_{p,q}$ with $\\delta(p,u_{p,q}) = \\delta(q,u_{p,q})$.\n\n*Proof.*\n\n($\\Rightarrow$) Immediate: a reset word merges *all* states, in particular every pair.\n\n($\\Leftarrow$) Suppose every pair is mergeable. We build a reset word by induction on $k = |\\delta(Q,w)|$, the current image size, starting from $w=\\varepsilon$ ($k=n$).\n\n- Key fact: for **any** $S\\subseteq Q$ and **any** $c\\in\\Sigma$, $|\\delta(S,c)| \\le |S|$, since $\\delta(\\cdot,c)$ is a function.\n- So image size is monotonically non-increasing along any word, for any starting set.\n- If current image $S = \\delta(Q,w)$ has $|S| = k > 1$, pick any two elements $p,q\\in S$. By"
  }
]

#3 · Assistant

2026-06-30 21:38:04

Model · claude-sonnet-5

## 5. Closing the induction; quantitative leakage in the naive bound

Picking up exactly where §4 stopped — finishing the proof, then immediately confronting why this argument, as stated, is the wrong proof for the Černý conjecture even though it's the right proof for synchronizability.

Lemma (Pair-reduction), conclusion of proof.

Suppose every pair is mergeable. Induct on $k=|\delta(Q,w)|$.

- If $k=1$, $w$ is already a reset word, done.
- If $k>1$, pick $p\neq q\in S:=\delta(Q,w)$. By hypothesis there is $u_{p,q}$ with $\delta(p,u_{p,q})=\delta(q,u_{p,q})$. Apply it: since image size is non-increasing under any letter (hence any word), and $\delta(S,u_{p,q})$ identifies at least the images of $p,q$ (which were distinct in $S$, now coincide), we get
$$|\delta(Q, w\,u_{p,q})| = |\delta(S,u_{p,q})| \le |S|-1 = k-1.$$
- Repeat. After at most $n-1$ such concatenations the image size hits $1$. $\blacksquare$

Corollary (existence is poly-time decidable). Checking mergeability of a single pair $p,q$ is itself a BFS reachability question on the pair graph: states are unordered pairs $\{x,y\}\subseteq Q$ (plus a merged sink), edges $\{x,y\}\xrightarrow{c}\{\delta(x,c),\delta(y,c)\}$, collapsing to the sink when the two images coincide. This graph has $\binom n2 = O(n^2)$ vertices and $|\Sigma|$ out-edges each, so one BFS per pair is $O(n^2|\Sigma|)$, and there are $O(n^2)$ pairs to check (or better: a single multi-source BFS from all pairs simultaneously, since we just need some word per pair, and synchronizability is equivalent to every pair-vertex reaching the sink) — overall $\boxed{O(n^2|\Sigma|)}$ — this is, modulo constant-factor engineering, the Eppstein / Sandberg synchronizability test, polynomial despite the exponential subset automaton lurking in §3.

---

### 5.1 Where the proof leaks length

The induction gives an upper bound for free: let
$$M(\mathcal A) = \max_{p\neq q} \min\{|u| : \delta(p,u)=\delta(q,u)\}$$
be the maximal pairwise merge length. The induction above immediately yields

$$\mathrm{rt}(\mathcal A) \;\le\; (n-1)\cdot M(\mathcal A).$$

Two failures of tightness, both essential:

1. $M(\mathcal A)$ itself can be $\Theta(n^2)$ in pathological-looking automata, or it can be as small as $O(n)$ — and even when $M(\mathcal A) = O(n)$, the bound $(n-1)M(\mathcal A)$ is $O(n^2)$, matching the conjecture's order but typically overshooting the constant badly, because:
2. The same merging word, reused greedily pair-by-pair, is wasteful: nothing says the $u_{p,q}$ chosen at step $k$ doesn't un-help progress made by earlier letters on other near-miss pairs. The bound is a worst-case sum of $n-1$ independent worst cases, with no accounting for words that merge many pairs at once.

For $C_n$ specifically: $M(C_n) = n-1$ exactly (mergeing the "hardest" pair $\{0,1\}$ under repeated $a$ then one $b$ takes a full cycle), giving the trivial bound $\mathrm{rt}(C_n)\le (n-1)^2$ — which happens to be exactly the conjectured tight value for this family. That coincidence is precisely why $C_n$ is the right calibration example: it's the family where the lazy induction proof is already optimal, which is suspicious and worth distrusting in general.

Open structural question to carry forward: characterize automata where the induction is tight (no merge interacts usefully with another) vs. automata where it's wildly loose (e.g. random automata, where $\mathrm{rt}$ is known to be $O(n\log^3 n)$ on average — Skvortsov–Zaks / Nicaud). This is the natural next entry: build the pair graph for $C_4$ explicitly, find $M(C_4)$, compare to the true $\mathrm{rt}(C_4)=9=(4-1)^2$, and watch the bound $(n-1)M(n)$ either coincide or diverge as $n$ grows past $3$.

Debug · Raw response blocks JSON

[
  {
    "type": "text",
    "text": "## 5. Closing the induction; quantitative leakage in the naive bound\n\nPicking up exactly where §4 stopped — finishing the proof, then immediately confronting why this argument, as stated, is the *wrong* proof for the Černý conjecture even though it's the *right* proof for synchronizability.\n\n**Lemma (Pair-reduction), conclusion of proof.**\n\nSuppose every pair is mergeable. Induct on $k=|\\delta(Q,w)|$.\n\n- If $k=1$, $w$ is already a reset word, done.\n- If $k>1$, pick $p\\neq q\\in S:=\\delta(Q,w)$. By hypothesis there is $u_{p,q}$ with $\\delta(p,u_{p,q})=\\delta(q,u_{p,q})$. Apply it: since image size is non-increasing under any letter (hence any word), and $\\delta(S,u_{p,q})$ identifies at least the images of $p,q$ (which were distinct in $S$, now coincide), we get\n$$|\\delta(Q, w\\,u_{p,q})| = |\\delta(S,u_{p,q})| \\le |S|-1 = k-1.$$\n- Repeat. After at most $n-1$ such concatenations the image size hits $1$. $\\blacksquare$\n\n**Corollary (existence is poly-time decidable).** Checking mergeability of a single pair $p,q$ is itself a BFS reachability question on the *pair graph*: states are unordered pairs $\\{x,y\\}\\subseteq Q$ (plus a merged sink), edges $\\{x,y\\}\\xrightarrow{c}\\{\\delta(x,c),\\delta(y,c)\\}$, collapsing to the sink when the two images coincide. This graph has $\\binom n2 = O(n^2)$ vertices and $|\\Sigma|$ out-edges each, so one BFS per pair is $O(n^2|\\Sigma|)$, and there are $O(n^2)$ pairs to check (or better: a single multi-source BFS from all pairs simultaneously, since we just need *some* word per pair, and synchronizability is equivalent to *every* pair-vertex reaching the sink) — overall $\\boxed{O(n^2|\\Sigma|)}$ — this is, modulo constant-factor engineering, the **Eppstein / Sandberg synchronizability test**, polynomial despite the exponential subset automaton lurking in §3.\n\n---\n\n### 5.1 Where the proof leaks length\n\nThe induction gives an upper bound for free: let\n$$M(\\mathcal A) = \\max_{p\\neq q} \\min\\{|u| : \\delta(p,u)=\\delta(q,u)\\}$$\nbe the **maximal pairwise merge length**. The induction above immediately yields\n\n$$\\mathrm{rt}(\\mathcal A) \\;\\le\\; (n-1)\\cdot M(\\mathcal A).$$\n\nTwo failures of tightness, both essential:\n\n1. **$M(\\mathcal A)$ itself can be $\\Theta(n^2)$ in pathological-looking automata**, or it can be as small as $O(n)$ — and even when $M(\\mathcal A) = O(n)$, the bound $(n-1)M(\\mathcal A)$ is $O(n^2)$, matching the conjecture's *order* but typically overshooting the *constant* badly, because:\n2. **The same merging word, reused greedily pair-by-pair, is wasteful**: nothing says the $u_{p,q}$ chosen at step $k$ doesn't *un-help* progress made by earlier letters on other near-miss pairs. The bound is a worst-case sum of $n-1$ independent worst cases, with no accounting for words that merge *many* pairs at once.\n\nFor $C_n$ specifically: $M(C_n) = n-1$ exactly (mergeing the \"hardest\" pair $\\{0,1\\}$ under repeated $a$ then one $b$ takes a full cycle), giving the trivial bound $\\mathrm{rt}(C_n)\\le (n-1)^2$ — which happens to be *exactly* the conjectured tight value for this family. That coincidence is precisely why $C_n$ is the right calibration example: it's the family where the lazy induction proof is already optimal, which is suspicious and worth distrusting in general.\n\n**Open structural question to carry forward**: characterize automata where the induction is tight (no merge interacts usefully with another) vs. automata where it's wildly loose (e.g. random automata, where $\\mathrm{rt}$ is known to be $O(n\\log^3 n)$ on average — Skvortsov–Zaks / Nicaud). This is the natural next entry: build the pair graph for $C_4$ explicitly, find $M(C_4)$, compare to the true $\\mathrm{rt}(C_4)=9=(4-1)^2$, and watch the bound $(n-1)M(n)$ either coincide or diverge as $n$ grows past $3$."
  }
]

#4 · Assistant

2026-06-30 21:41:07

Model · claude-sonnet-5

Thinking

Now I'm working through the explicit construction of the pair graph for C_4, computing its merge complexity M(C_4), and then checking how well the bound (n-1)M(n) approximates the actual return time of 9.

For each pair in C_4, I'm finding the shortest word that merges them: {0,1} merges immediately with b, while other pairs like {1,2} require rotating first to align one element with 0 before applying b, which takes additional steps. For non-adjacent pairs like {0,2}, applying b first gives {1,2}, which then requires the full 4-step merge sequence, totaling 5 operations. I'm checking whether a different approach using rotations might be shorter. I'm setting up a proper breadth-first search on the state graph to find the shortest sequence. I'm mapping out all six unordered pairs as nodes and computing their transitions under both operations a and b. Under a, each pair shifts by one position cyclically, while b collapses pairs containing 0 into a single merged state. I'm starting to trace which pairs lead directly to the merged sink state. Continuing the reverse BFS to assign distances to remaining pairs: P6 maps back to P4 via a, giving P4 distance 4; P4 maps back to P2 via b, giving P2 distance 5; P2 maps back to P5 via a, giving P5 distance 6. The final distances are P1=1, P2=5, P3=2, P4=4, P5=6, P6=3, so the maximum merge length is 6 from pair P5. This gives a naive bound of 3×6=18 for the actual runtime of C₄, which is 9.

Now I'm noticing the bound overshoots by a factor of 2, which aligns with the expected loose bound described earlier. For n=3, I need to verify whether M(C₃) actually equals n−1=2 as claimed, since my distance calculation gave 3, which would contradict the tight bound of 4=(n−1)². Let me recalculate the BFS distances more carefully for C₃ to resolve this discrepancy. So the distances work out to d1=1, d2=2, d3=3, giving M(C_3)=3. But that means the bound (n-1)M(C_3) = 23=6 doesn't match the actual result of 4 for C_3, so the induction bound isn't tight here either. I need to reconsider whether my interpretation of M aligns with what the text intended.

Looking back at the claim that M(C_n) = n-1, I realize the text might be confusing which pair is hardest. The pair {0,1} actually merges in one step via b—that's the easiest pair, not the hardest. The hardest pair for C_3 is {1,2} with distance 3. So the stated formula seems off, or it's referring to something different.

Actually, let me check if there's a pattern: for C_3 I get M(C_3)=3=n, and for C_4 I computed M(C_4)=6. That's 32/2 and 43/2 respectively—both equal to n(n-1)/2, which is the binomial coefficient C(n,2). This matches our two data points, so maybe M(C_n) = n(n-1)/2 is the actual formula, not n-1. This would align with known results about the Černý automaton's pair-merge structure.

Comparing growth rates: the naive bound gives roughly n³/2, while the actual reset threshold is (n-1)² ~ n². So the slack between them grows significantly with n, confirming how wasteful the naive approach is. I should verify this pattern holds for n=5 and double-check my C_4 computation by simulating the transitions directly to catch any arithmetic errors.

Working through the state transitions for C_4, I'm tracing the shortest path from P5={1,3} to the sink by following which transitions lead somewhere new versus self-looping. The path goes through P2, P4, P6, P3, and finally P1 before reaching the sink, giving a total distance of 6. I'm also verifying the other distances match: P1=1, P3=2, P6=3, P4=4, P2=5, all consistent with M(C_4)=6. Doing the same check for C_3 confirms those distances too, and both cases suggest the pattern M(C_n) = C(n,2) = n(n-1)/2.

Now I'm examining the underlying structure of the pair graph more carefully. Under the letter a, pairs rotate cyclically while preserving their gap structure, but under b the dynamics are more complex—only pairs containing 0 are affected, and b essentially shifts them in a way that changes the gap. This lossy rotation structure is actually well-studied in the literature, and the key insight is that the gap function decreases appropriately as we apply transitions, which explains why the maximum distance across all pairs follows the binomial coefficient formula. cts the full automaton reset time. So there's a gap here—a single pair can be merged in O(n²) steps, but resetting the entire automaton takes O(n²) as well, suggesting the bottleneck isn't just one pair but the interplay of all pairs and how they interact under the group action.

Let me verify this against the C₄ examples: the pair {1,3} with gap 2 took 6 steps, and if each gap reduction costs up to n steps with (g-1) reductions needed, that's roughly 2×4 = 8 steps worst case, which is close. For {0,2} at distance 5 and {1,2} at distance 4, the pattern seems to hold—smaller gaps merge faster, and the cost scales with both the gap size and the cycle length.

Debug · Raw response blocks JSON

[
  {
    "type": "thinking",
    "thinking": "Now I'm working through the explicit construction of the pair graph for C_4, computing its merge complexity M(C_4), and then checking how well the bound (n-1)M(n) approximates the actual return time of 9.\n\nFor each pair in C_4, I'm finding the shortest word that merges them: {0,1} merges immediately with b, while other pairs like {1,2} require rotating first to align one element with 0 before applying b, which takes additional steps. For non-adjacent pairs like {0,2}, applying b first gives {1,2}, which then requires the full 4-step merge sequence, totaling 5 operations. I'm checking whether a different approach using rotations might be shorter. I'm setting up a proper breadth-first search on the state graph to find the shortest sequence. I'm mapping out all six unordered pairs as nodes and computing their transitions under both operations a and b. Under a, each pair shifts by one position cyclically, while b collapses pairs containing 0 into a single merged state. I'm starting to trace which pairs lead directly to the merged sink state. Continuing the reverse BFS to assign distances to remaining pairs: P6 maps back to P4 via a, giving P4 distance 4; P4 maps back to P2 via b, giving P2 distance 5; P2 maps back to P5 via a, giving P5 distance 6. The final distances are P1=1, P2=5, P3=2, P4=4, P5=6, P6=3, so the maximum merge length is 6 from pair P5. This gives a naive bound of 3×6=18 for the actual runtime of C₄, which is 9.\n\nNow I'm noticing the bound overshoots by a factor of 2, which aligns with the expected loose bound described earlier. For n=3, I need to verify whether M(C₃) actually equals n−1=2 as claimed, since my distance calculation gave 3, which would contradict the tight bound of 4=(n−1)². Let me recalculate the BFS distances more carefully for C₃ to resolve this discrepancy. So the distances work out to d1=1, d2=2, d3=3, giving M(C_3)=3. But that means the bound (n-1)*M(C_3) = 2*3=6 doesn't match the actual result of 4 for C_3, so the induction bound isn't tight here either. I need to reconsider whether my interpretation of M aligns with what the text intended.\n\nLooking back at the claim that M(C_n) = n-1, I realize the text might be confusing which pair is hardest. The pair {0,1} actually merges in one step via b—that's the easiest pair, not the hardest. The hardest pair for C_3 is {1,2} with distance 3. So the stated formula seems off, or it's referring to something different.\n\nActually, let me check if there's a pattern: for C_3 I get M(C_3)=3=n, and for C_4 I computed M(C_4)=6. That's 3*2/2 and 4*3/2 respectively—both equal to n(n-1)/2, which is the binomial coefficient C(n,2). This matches our two data points, so maybe M(C_n) = n(n-1)/2 is the actual formula, not n-1. This would align with known results about the Černý automaton's pair-merge structure.\n\nComparing growth rates: the naive bound gives roughly n³/2, while the actual reset threshold is (n-1)² ~ n². So the slack between them grows significantly with n, confirming how wasteful the naive approach is. I should verify this pattern holds for n=5 and double-check my C_4 computation by simulating the transitions directly to catch any arithmetic errors.\n\nWorking through the state transitions for C_4, I'm tracing the shortest path from P5={1,3} to the sink by following which transitions lead somewhere new versus self-looping. The path goes through P2, P4, P6, P3, and finally P1 before reaching the sink, giving a total distance of 6. I'm also verifying the other distances match: P1=1, P3=2, P6=3, P4=4, P2=5, all consistent with M(C_4)=6. Doing the same check for C_3 confirms those distances too, and both cases suggest the pattern M(C_n) = C(n,2) = n(n-1)/2.\n\nNow I'm examining the underlying structure of the pair graph more carefully. Under the letter a, pairs rotate cyclically while preserving their gap structure, but under b the dynamics are more complex—only pairs containing 0 are affected, and b essentially shifts them in a way that changes the gap. This lossy rotation structure is actually well-studied in the literature, and the key insight is that the gap function decreases appropriately as we apply transitions, which explains why the maximum distance across all pairs follows the binomial coefficient formula. cts the full automaton reset time. So there's a gap here—a single pair can be merged in O(n²) steps, but resetting the entire automaton takes O(n²) as well, suggesting the bottleneck isn't just one pair but the interplay of all pairs and how they interact under the group action.\n\nLet me verify this against the C₄ examples: the pair {1,3} with gap 2 took 6 steps, and if each gap reduction costs up to n steps with (g-1) reductions needed, that's roughly 2×4 = 8 steps worst case, which is close. For {0,2} at distance 5 and {1,2} at distance 4, the pattern seems to hold—smaller gaps merge faster, and the cost scales with both the gap size and the cycle length.",
    "signature": "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"
  }
]

#5 · Assistant

2026-06-30 21:44:10

Model · claude-sonnet-5

Thinking

Now I'm setting up the cycle graph C_4 with its specific automorphism structure—defining the vertex set, the generator functions, and then enumerating all six possible pairs to build out the explicit pair graph. I'll compute M(C_4) and check it against the known value rt(C_4)=9 to see how tight the induction bound actually is.

Computing the pair graph transitions under both generators: under a, each pair shifts by 1 mod 4; under b, the element 0 maps to 1 while others stay fixed, which collapses {0,1} into a sink state immediately. The other pairs either self-loop or transition to new pairs, creating a directed graph structure that I can analyze for reachability and mixing properties. So I need to systematically compute shortest paths from each pair state to the sink using BFS. Let me set up the pair graph properly: the six pair states are {0,1}, {0,2}, {0,3}, {1,2}, {1,3}, {2,3}, plus the sink. I'll trace through the transitions for each pair under both letters a and b, then run BFS from each starting pair to find the minimum word length needed to merge them. I'm realizing P2 and P5 form a cycle under the 'a' operation—they map to each other indefinitely—so I can't reach the sink through repeated applications of 'a' alone. From P2, I can use 'b' to jump to P4 (which has distance 4, giving total distance 5), while P5's 'b' is just a self-loop and doesn't help. For P5, the only useful edge is 'a' back to P2, creating the dependency d(P5) = 1 + d(P2), which means I need to solve this mutual recursion by recognizing that P2 must eventually use its 'b' edge to escape the cycle.

Working through the math: d(P2) = min(1 + d(P5), 1 + d(P4)) = min(2 + d(P2), 5), so d(P2) = 5 and d(P5) = 6. Compiling all distances across the six pairs, the maximum is 6 for pair {1,3}, giving M(C₄) = 6. The lazy induction bound predicts reset time as 3 × 6 = 18, but the actual reset time for C₄ is 9—a factor of 2 looser, which quantifies the "leakage" I mentioned. The Černý automaton family C_n has known minimal reset word length (n-1)², confirming rt(C₄) = 9.

Now I'm checking whether this looseness appears at smaller values. For C₃, I need to recompute M using the same pairwise distance method to see if the bound was actually tight there or if that was just coincidence. I'm checking all length-2 words to see if any merge states 1 and 2 in C_3, and none do—only length-3 "aab" works. This contradicts the earlier claim that M(C_n)=n-1 exactly; my computation shows M(C_3)=3, not 2. This looks like an error in the previous entry that's worth correcting with explicit verification. Now I'm verifying the reset word length formula for C_n. The standard construction gives a reset word of length (n-1)^2, which for n=4 means length 9. I'll derive the explicit word pattern — it follows the form (a^{n-1}b)^{n-2}a^{n-1} — and verify it achieves a reset by simulating it on the automaton, similar to how I checked the n=3 case directly, rather than proving optimality by exhaustive search. I'm working through the subset dynamics more carefully now. When I apply b to a subset, it only affects element 0 by mapping it to 1, so b acts as the identity on subsets that don't contain 0. To actually reduce the subset size, I need both 0 and 1 present so that b merges them together. Starting from {0,1,2,3}, applying b immediately gives {1,2,3} since 0 and 1 collapse—that's the first reduction in just one step.

From {1,2,3}, I cycle with a until 0 reappears alongside 1: after two applications of a I reach {3,0,1}, then applying b merges 0 and 1 to get {3,1} with size 2. Now I'm stuck in a period-2 cycle under a alone between {1,3} and {2,0}, so I need to apply b at the right moment to break out and continue reducing.

This reveals a critical issue: my first phase takes 4 steps to reduce from size 4 to size 2, and the second phase (merging the pair {1,3}) requires 6 steps based on my pair-graph calculation, giving a total of 10 steps. But the conjectured reset time for C₄ is 9, so either my pair-graph distance is wrong or the first-phase reduction isn't optimal—maybe a different sequence could do better.

The key insight is that since a is a bijection (permutation), it never reduces subset size; only b can reduce, and it does so by merging the two elements that map to the same image. So the real question is whether there's a path from the full set down to a cheaper pair like {0,1} (distance 1) or {0,3} (distance 2) instead of the expensive {1,3} (distance 6). So the strategy is to apply rotations (powers of a) followed by the merging operation b, repeating this n-1 times total to reduce the subset from size n down to size 1. The total word length becomes the sum of all rotation amounts plus n-1 applications of b. The classical bound shows this sums to at most (n-1)², and for n=4 that gives 9. I'm reconsidering whether my pair-graph approach was actually computing the right thing—it gave the minimum merge length for a single pair in isolation, but the full reset word uses a different algorithm that processes pairs sequentially through the quotient structure rather than optimizing each pair independently.

Debug · Raw response blocks JSON

[
  {
    "type": "thinking",
    "thinking": "Now I'm setting up the cycle graph C_4 with its specific automorphism structure—defining the vertex set, the generator functions, and then enumerating all six possible pairs to build out the explicit pair graph. I'll compute M(C_4) and check it against the known value rt(C_4)=9 to see how tight the induction bound actually is.\n\nComputing the pair graph transitions under both generators: under a, each pair shifts by 1 mod 4; under b, the element 0 maps to 1 while others stay fixed, which collapses {0,1} into a sink state immediately. The other pairs either self-loop or transition to new pairs, creating a directed graph structure that I can analyze for reachability and mixing properties. So I need to systematically compute shortest paths from each pair state to the sink using BFS. Let me set up the pair graph properly: the six pair states are {0,1}, {0,2}, {0,3}, {1,2}, {1,3}, {2,3}, plus the sink. I'll trace through the transitions for each pair under both letters a and b, then run BFS from each starting pair to find the minimum word length needed to merge them. I'm realizing P2 and P5 form a cycle under the 'a' operation—they map to each other indefinitely—so I can't reach the sink through repeated applications of 'a' alone. From P2, I can use 'b' to jump to P4 (which has distance 4, giving total distance 5), while P5's 'b' is just a self-loop and doesn't help. For P5, the only useful edge is 'a' back to P2, creating the dependency d(P5) = 1 + d(P2), which means I need to solve this mutual recursion by recognizing that P2 must eventually use its 'b' edge to escape the cycle.\n\nWorking through the math: d(P2) = min(1 + d(P5), 1 + d(P4)) = min(2 + d(P2), 5), so d(P2) = 5 and d(P5) = 6. Compiling all distances across the six pairs, the maximum is 6 for pair {1,3}, giving M(C₄) = 6. The lazy induction bound predicts reset time as 3 × 6 = 18, but the actual reset time for C₄ is 9—a factor of 2 looser, which quantifies the \"leakage\" I mentioned. The Černý automaton family C_n has known minimal reset word length (n-1)², confirming rt(C₄) = 9.\n\nNow I'm checking whether this looseness appears at smaller values. For C₃, I need to recompute M using the same pairwise distance method to see if the bound was actually tight there or if that was just coincidence. I'm checking all length-2 words to see if any merge states 1 and 2 in C_3, and none do—only length-3 \"aab\" works. This contradicts the earlier claim that M(C_n)=n-1 exactly; my computation shows M(C_3)=3, not 2. This looks like an error in the previous entry that's worth correcting with explicit verification. Now I'm verifying the reset word length formula for C_n. The standard construction gives a reset word of length (n-1)^2, which for n=4 means length 9. I'll derive the explicit word pattern — it follows the form (a^{n-1}b)^{n-2}a^{n-1} — and verify it achieves a reset by simulating it on the automaton, similar to how I checked the n=3 case directly, rather than proving optimality by exhaustive search. I'm working through the subset dynamics more carefully now. When I apply b to a subset, it only affects element 0 by mapping it to 1, so b acts as the identity on subsets that don't contain 0. To actually reduce the subset size, I need both 0 and 1 present so that b merges them together. Starting from {0,1,2,3}, applying b immediately gives {1,2,3} since 0 and 1 collapse—that's the first reduction in just one step.\n\nFrom {1,2,3}, I cycle with a until 0 reappears alongside 1: after two applications of a I reach {3,0,1}, then applying b merges 0 and 1 to get {3,1} with size 2. Now I'm stuck in a period-2 cycle under a alone between {1,3} and {2,0}, so I need to apply b at the right moment to break out and continue reducing.\n\nThis reveals a critical issue: my first phase takes 4 steps to reduce from size 4 to size 2, and the second phase (merging the pair {1,3}) requires 6 steps based on my pair-graph calculation, giving a total of 10 steps. But the conjectured reset time for C₄ is 9, so either my pair-graph distance is wrong or the first-phase reduction isn't optimal—maybe a different sequence could do better.\n\nThe key insight is that since a is a bijection (permutation), it never reduces subset size; only b can reduce, and it does so by merging the two elements that map to the same image. So the real question is whether there's a path from the full set down to a cheaper pair like {0,1} (distance 1) or {0,3} (distance 2) instead of the expensive {1,3} (distance 6). So the strategy is to apply rotations (powers of a) followed by the merging operation b, repeating this n-1 times total to reduce the subset from size n down to size 1. The total word length becomes the sum of all rotation amounts plus n-1 applications of b. The classical bound shows this sums to at most (n-1)², and for n=4 that gives 9. I'm reconsidering whether my pair-graph approach was actually computing the right thing—it gave the minimum merge length for a single pair in isolation, but the full reset word uses a different algorithm that processes pairs sequentially through the quotient structure rather than optimizing each pair independently.",
    "signature": "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"
  }
]

#6 · Assistant

2026-06-30 21:47:13

Model · claude-sonnet-5

Thinking

Now I'm working through the concrete case of $C_4$ to build its pair graph explicitly, compute the reset threshold, and verify it equals 9. Then I'll extend this to prove that $M(C_n) = n-1$ for the general cycle, examining how the pair graph forms a single path structure for Černý automata and where the bound aligns with or diverges from the theoretical limits. So the key insight is that {0,1} merges in just one step with $b$, making it the easiest pair. The hardest pairs are those requiring rotation to align them with {0,1} before merging — for adjacent pairs this takes up to $n-1$ applications of $a$ plus one $b$, giving a maximum of $n$ steps. But I'm realizing there might be other merging strategies beyond directly rotating to {0,1}, like using intermediate partial merges.

Now I'm working through the actual structure of the Černý reset word for $C_n$. The standard form is $(a^{n-1}b)^{n-2}a^{n-1}$ with total length $(n-1)^2$, which I can verify against the $n=3$ example where the reset word $ba^2b$ has length 4, matching $(3-1)^2$.

To confirm this for $n=4$, I'm running a subset-BFS algorithm: starting with the full set $\{0,1,2,3\}$, applying $b$ immediately reduces it to $\{1,2,3\}$ since $b$ merges states 0 and 1. Now I need to apply $a$ rotations strategically to bring state 0 back into the set at a position where another $b$ application can reduce the size further, rather than acting as identity on the remaining elements. If S is missing element 0 (so 1 is present), applying b doesn't reduce the set since b only acts on 0 which isn't there—it just leaves S unchanged. But if S is missing some other element m (meaning both 0 and 1 are in S), then b maps 0→1, causing a collision that reduces the set size from n-1 to n-2. So I need the missing element to cycle through until it's no longer 0 or 1 to trigger the reduction. I need to be more careful about when reduction actually happens—it requires the missing element to be something other than 0 or 1. When the missing element is 1, applying b just shifts where the missing element is without merging anything. So the pattern emerges: start with b to get a missing element of 0, then apply a twice to rotate the missing element to 2, then apply b again to trigger the actual reduction. For n=3 this gives the word baab with length 4, and I'm working through how this generalizes to larger n by tracking the state as the set of missing elements grows. to find the rotation amount that positions the missing set away from {0,1}, then apply b to expand it by one element. This process repeats, with rotations accumulating, until we've synchronized all states down to a single equivalence class—and in the worst case, we need to rotate through all possible positions before finding one that works.

The known minimal word for the Černý automaton follows the pattern of alternating rotation phases and expansion steps, giving a total length of (n-1)². Let me verify this against the specific example to see if the formula holds or if there's a tighter construction.

For n=4, I'll simulate the reset process step by step: starting with an empty lost set T, I apply rotations strategically to avoid forbidden positions, then apply b to expand T. In phase 1, T is empty so any rotation works trivially—I'll apply b directly with no a's, giving T={0}. In phase 2, T={0} and I need to rotate so that 0 doesn't land on positions 0 or 1, meaning I need r∉{0,1}.  Choosing r=2 works, so I apply a² then b, giving T={0,2} and the word "baab" (length 4). Now in phase 3 with T={0,2}, I'm checking which rotation avoids {0,1}: the problem is that T={0,2} consists of antipodal points on the 4-cycle, so rotating by any r gives either {r, r+2} or {r+2, r}, which always hits at least one element of {0,1} depending on whether r is even or odd—this seems to be a dead end, suggesting the greedy approach might need adjustment or that this particular pair is the hardest to merge.

Actually, I'm realizing the issue: I was too restrictive in requiring b to only apply when it reduces |T|. The key insight is that b can still be useful even when 0∈T (meaning 0 is lost from S), because in that case b doesn't change S at all—it only affects the image of 0, which isn't in S anyway. But if 0∉T and 1∈T, then b transforms S by replacing 0 with 1, keeping the size the same while altering T's structure, which might enable progress later.

So the real constraint for escaping T={0,2} is that I need 0∉T after some rotation, which means r must avoid both 0 and 2 (since r+2≡0 mod 4 when r=2). So r∈{1,3} would work to get 0 out of the rotated set.

With r=1, I get T+1={1,3}, and since 0∉{1,3} but 1∈{1,3}, the b operation swaps the missing element from 1 to 0, giving T'={0,3}—still size 2 but a different shape. Now with T={0,3}, rotating again: the set becomes two adjacent elements mod 4, so r=0 or r=1 both keep 0 in the rotated set (no-op), but r=2 gives {2,1} where 0 is finally absent. So phase 4 gets us to T={0,2,3} with S={1} remaining, meaning we've reached a singleton image after 10 letters total. But that's one more than the expected rt(C_4)=9, so I'm wondering if there's a more efficient path—maybe skipping the swap phase entirely or finding a better choice of r in phase 2 to make direct progress. Let me recheck all possible r values for T={0,2} to see if any bypass the swap and land somewhere better.

Debug · Raw response blocks JSON

[
  {
    "type": "thinking",
    "thinking": "Now I'm working through the concrete case of $C_4$ to build its pair graph explicitly, compute the reset threshold, and verify it equals 9. Then I'll extend this to prove that $M(C_n) = n-1$ for the general cycle, examining how the pair graph forms a single path structure for Černý automata and where the bound aligns with or diverges from the theoretical limits. So the key insight is that {0,1} merges in just one step with $b$, making it the easiest pair. The hardest pairs are those requiring rotation to align them with {0,1} before merging — for adjacent pairs this takes up to $n-1$ applications of $a$ plus one $b$, giving a maximum of $n$ steps. But I'm realizing there might be other merging strategies beyond directly rotating to {0,1}, like using intermediate partial merges.\n\nNow I'm working through the actual structure of the Černý reset word for $C_n$. The standard form is $(a^{n-1}b)^{n-2}a^{n-1}$ with total length $(n-1)^2$, which I can verify against the $n=3$ example where the reset word $ba^2b$ has length 4, matching $(3-1)^2$.\n\nTo confirm this for $n=4$, I'm running a subset-BFS algorithm: starting with the full set $\\{0,1,2,3\\}$, applying $b$ immediately reduces it to $\\{1,2,3\\}$ since $b$ merges states 0 and 1. Now I need to apply $a$ rotations strategically to bring state 0 back into the set at a position where another $b$ application can reduce the size further, rather than acting as identity on the remaining elements. If S is missing element 0 (so 1 is present), applying b doesn't reduce the set since b only acts on 0 which isn't there—it just leaves S unchanged. But if S is missing some other element m (meaning both 0 and 1 are in S), then b maps 0→1, causing a collision that reduces the set size from n-1 to n-2. So I need the missing element to cycle through until it's no longer 0 or 1 to trigger the reduction. I need to be more careful about when reduction actually happens—it requires the missing element to be something other than 0 or 1. When the missing element is 1, applying b just shifts where the missing element is without merging anything. So the pattern emerges: start with b to get a missing element of 0, then apply a twice to rotate the missing element to 2, then apply b again to trigger the actual reduction. For n=3 this gives the word baab with length 4, and I'm working through how this generalizes to larger n by tracking the state as the set of missing elements grows. to find the rotation amount that positions the missing set away from {0,1}, then apply b to expand it by one element. This process repeats, with rotations accumulating, until we've synchronized all states down to a single equivalence class—and in the worst case, we need to rotate through all possible positions before finding one that works.\n\nThe known minimal word for the Černý automaton follows the pattern of alternating rotation phases and expansion steps, giving a total length of (n-1)². Let me verify this against the specific example to see if the formula holds or if there's a tighter construction.\n\nFor n=4, I'll simulate the reset process step by step: starting with an empty lost set T, I apply rotations strategically to avoid forbidden positions, then apply b to expand T. In phase 1, T is empty so any rotation works trivially—I'll apply b directly with no a's, giving T={0}. In phase 2, T={0} and I need to rotate so that 0 doesn't land on positions 0 or 1, meaning I need r∉{0,1}.  Choosing r=2 works, so I apply a² then b, giving T={0,2} and the word \"baab\" (length 4). Now in phase 3 with T={0,2}, I'm checking which rotation avoids {0,1}: the problem is that T={0,2} consists of antipodal points on the 4-cycle, so rotating by any r gives either {r, r+2} or {r+2, r}, which always hits at least one element of {0,1} depending on whether r is even or odd—this seems to be a dead end, suggesting the greedy approach might need adjustment or that this particular pair is the hardest to merge.\n\nActually, I'm realizing the issue: I was too restrictive in requiring b to only apply when it reduces |T|. The key insight is that b can still be useful even when 0∈T (meaning 0 is lost from S), because in that case b doesn't change S at all—it only affects the image of 0, which isn't in S anyway. But if 0∉T and 1∈T, then b transforms S by replacing 0 with 1, keeping the size the same while altering T's structure, which might enable progress later.\n\nSo the real constraint for escaping T={0,2} is that I need 0∉T after some rotation, which means r must avoid both 0 and 2 (since r+2≡0 mod 4 when r=2). So r∈{1,3} would work to get 0 out of the rotated set.\n\nWith r=1, I get T+1={1,3}, and since 0∉{1,3} but 1∈{1,3}, the b operation swaps the missing element from 1 to 0, giving T'={0,3}—still size 2 but a different shape. Now with T={0,3}, rotating again: the set becomes two adjacent elements mod 4, so r=0 or r=1 both keep 0 in the rotated set (no-op), but r=2 gives {2,1} where 0 is finally absent. So phase 4 gets us to T={0,2,3} with S={1} remaining, meaning we've reached a singleton image after 10 letters total. But that's one more than the expected rt(C_4)=9, so I'm wondering if there's a more efficient path—maybe skipping the swap phase entirely or finding a better choice of r in phase 2 to make direct progress. Let me recheck all possible r values for T={0,2} to see if any bypass the swap and land somewhere better.",
    "signature": "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"
  }
]

#7 · Assistant

2026-06-30 21:50:16

Model · claude-sonnet-5

Thinking

Now I'm working through the cycle graph $C_4$ by setting up its structure with vertices and generators, then listing all six possible pairs and beginning to compute their merge word lengths. I'm realizing the pair {0,1} actually merges immediately under b since both 0 and 1 map to 1, so that's the easiest pair, not the hardest. The hardest pair to merge is likely something like {0, n-1} or non-adjacent pairs that require rotating through multiple a-transitions before b can collapse them. Let me reconsider which pairs actually merge under b: only {0,x} where x≠0 can potentially merge, and even then b maps 0→1 while leaving x unchanged, so we only get a true merge if x=1.

For a general pair {p,q} with gap d, applying a rotates both by 1, and b only merges when one element lands on 0 and the other is 1. When that happens, the pair collapses to {1}, but if the gap is larger than 1, after b we're left with {1, other} where other≠1, so the pair hasn't fully merged yet—instead the gap has effectively decreased by 1.

This suggests each merge takes roughly n steps per gap reduction, giving a total around n(n-1) operations, but I should verify this by explicitly simulating the pair graph for C_4 to find the exact distances and compare against the known result of 9. I'm doing a reverse BFS from the MERGE state to find the shortest path from each pair to the sink, working backwards through the graph edges. The distances I'm computing are P1=1, P3=2, P6=3, P4=4, P2=5, P5=6, but I should verify these by checking all incoming edges to each node to ensure I haven't missed a shorter route through multiple predecessors. Level 7 confirms no new predecessors for P5, so all six pairs have been assigned distances from 1 to 6. For the Černý automaton with n=4, the maximum distance is 6, which matches the pattern where pair distances span from 1 through n(n-1)/2. Now checking the n=3 case to verify this pattern holds: the three pairs {0,1}, {0,2}, {1,2} should follow similar distance assignments based on how the transitions merge them. Now I'm computing the reverse adjacency list properly and running BFS from MERGE to find shortest distances to each state. The distances come out to P1=1, P2=2, P3=3, which gives M(C₃)=3, but this contradicts the earlier claim that M(Cₙ)=n-1. I'm realizing that P1={0,1} actually merges immediately via the b transition, making it the easiest pair to merge, not the hardest—so I need to reconsider which pair is actually the most difficult to reach. I'm verifying the minimal reset word length for the pair {1,2} by testing all candidates systematically—single letters don't merge them, and checking all two-letter combinations confirms none work either, so the three-letter word "aab" is indeed minimal with length 3. This confirms M(C_3) = 3, not the n-1 = 2 claimed earlier, so that earlier assertion was wrong and needs correction with this concrete computation showing the actual bound relationship.

Now I'm working out the general formula for M(C_n) by examining the pattern across small cases: for n=3 the maximum distance is 3, and for n=4 it's 6, which matches the binomial coefficient C(n,2) = n(n-1)/2. The key insight is that the pair graph has a beautiful structure—it forms essentially a single chain where all C(n,2) pairs are visited in sequence before reaching a merge state, like a Hamiltonian path through the pair space, which is the classical structural property underlying the Černý automaton. Now I'm verifying the bound overshoots significantly—for C₃ the naive bound gives 6 but the actual reset word length is 4, and for C₄ it gives 18 versus 9, showing a factor of roughly n/2 excess. I should correct the earlier claim that this bound is tight by presenting the pair graph explicitly, deriving that M(Cₙ) equals the number of pairs, and demonstrating the gap with concrete examples. The known minimal reset word formula for the Černý automaton is (abⁿ⁻¹)ⁿ⁻²a with length (n-1)², which matches what I found for C₃ with the word abbа, though different minimal words of the same length may exist. I'm running a breadth-first search through the subset automaton for C_4, tracking how the state set shrinks with each letter applied. Starting from the full set {0,1,2,3}, applying b collapses it to {1,2,3}, then continuing with a and b moves me through various three-element sets until I reach a two-element set {1,3} at depth 4 via the sequence b,a,a,b. I'll keep expanding the search to find when we reach a singleton set, which gives us the reset word length. ore. b-> b(0)=1,b(1)=1,b(2)=2 => {1,2}=12 (size2).
 new: 12.

Level6: from 13: a-> a(1)=2,a(3)=0 => {2,0}=02. b-> b(1)=1,b(3)=3 => {1,3}=13 (self).
 new: 02.

From 12: a-> a(1)=2,a(2)=3 => {2,3}=23. b-> b(1)=1,b(2)=2 => {1,2}=12 (self).
 new: 23.

Level7: from 02: a-> a(0)=1,a(2)=3 => {1,3}=13 (visited). b-> b(0)=1,b(2)=2 => {1,2}=12 (visited).
 no new.

From 23: a-> a(2)=3,a(3)=0 => {3,0}=03. b-> b(2)=2,b(3)=3 => {2,3}=23 (self).
 new: 03.

Level8: from 03: a-> a(0)=1,a(3)=0 => {1,0}=01. b-> b(0)=1,b(3)=3 => {1,3}=13 (visited).
 new: 01.

Level9: from 01: a-> a(0)=1,a(1)=2 => {1,2}=12 (visited). b-> b(0)=1,b(1)=1 => {1} (singleton!).

Found singleton {1} at level 9, path length 9.

So the answer is 9.

Actually wait, let me double-check the path: 0123 -> 123 -> 023 -> 013 -> 012 -> 12 -> 23 -> 03 -> 01 -> {1}.

That's 10 steps (9 transitions). Hmm, let me recount: level0 to level9 is 9 transitions. Yes, 9 transitions = path length 9. Correct.
</next_thinking>

Continuing the BFS through the remaining levels, I find that from the size-2 sets at level 5, I can reach new size-2 sets at level 6, then a new size-2 set at level 7, and finally a size-1 set at level 9. Tracing back the path: 0123 → 123 → 023 → 013 → 012 → 12 → 23 → 03 → 01 → {1}, which gives 9 transitions total. So the answer is 9.

Debug · Raw response blocks JSON

[
  {
    "type": "thinking",
    "thinking": "Now I'm working through the cycle graph $C_4$ by setting up its structure with vertices and generators, then listing all six possible pairs and beginning to compute their merge word lengths. I'm realizing the pair {0,1} actually merges immediately under b since both 0 and 1 map to 1, so that's the easiest pair, not the hardest. The hardest pair to merge is likely something like {0, n-1} or non-adjacent pairs that require rotating through multiple a-transitions before b can collapse them. Let me reconsider which pairs actually merge under b: only {0,x} where x≠0 can potentially merge, and even then b maps 0→1 while leaving x unchanged, so we only get a true merge if x=1.\n\nFor a general pair {p,q} with gap d, applying a rotates both by 1, and b only merges when one element lands on 0 and the other is 1. When that happens, the pair collapses to {1}, but if the gap is larger than 1, after b we're left with {1, other} where other≠1, so the pair hasn't fully merged yet—instead the gap has effectively decreased by 1.\n\nThis suggests each merge takes roughly n steps per gap reduction, giving a total around n(n-1) operations, but I should verify this by explicitly simulating the pair graph for C_4 to find the exact distances and compare against the known result of 9. I'm doing a reverse BFS from the MERGE state to find the shortest path from each pair to the sink, working backwards through the graph edges. The distances I'm computing are P1=1, P3=2, P6=3, P4=4, P2=5, P5=6, but I should verify these by checking all incoming edges to each node to ensure I haven't missed a shorter route through multiple predecessors. Level 7 confirms no new predecessors for P5, so all six pairs have been assigned distances from 1 to 6. For the Černý automaton with n=4, the maximum distance is 6, which matches the pattern where pair distances span from 1 through n(n-1)/2. Now checking the n=3 case to verify this pattern holds: the three pairs {0,1}, {0,2}, {1,2} should follow similar distance assignments based on how the transitions merge them. Now I'm computing the reverse adjacency list properly and running BFS from MERGE to find shortest distances to each state. The distances come out to P1=1, P2=2, P3=3, which gives M(C₃)=3, but this contradicts the earlier claim that M(Cₙ)=n-1. I'm realizing that P1={0,1} actually merges immediately via the b transition, making it the easiest pair to merge, not the hardest—so I need to reconsider which pair is actually the most difficult to reach. I'm verifying the minimal reset word length for the pair {1,2} by testing all candidates systematically—single letters don't merge them, and checking all two-letter combinations confirms none work either, so the three-letter word \"aab\" is indeed minimal with length 3. This confirms M(C_3) = 3, not the n-1 = 2 claimed earlier, so that earlier assertion was wrong and needs correction with this concrete computation showing the actual bound relationship.\n\nNow I'm working out the general formula for M(C_n) by examining the pattern across small cases: for n=3 the maximum distance is 3, and for n=4 it's 6, which matches the binomial coefficient C(n,2) = n(n-1)/2. The key insight is that the pair graph has a beautiful structure—it forms essentially a single chain where all C(n,2) pairs are visited in sequence before reaching a merge state, like a Hamiltonian path through the pair space, which is the classical structural property underlying the Černý automaton. Now I'm verifying the bound overshoots significantly—for C₃ the naive bound gives 6 but the actual reset word length is 4, and for C₄ it gives 18 versus 9, showing a factor of roughly n/2 excess. I should correct the earlier claim that this bound is tight by presenting the pair graph explicitly, deriving that M(Cₙ) equals the number of pairs, and demonstrating the gap with concrete examples. The known minimal reset word formula for the Černý automaton is (abⁿ⁻¹)ⁿ⁻²a with length (n-1)², which matches what I found for C₃ with the word abbа, though different minimal words of the same length may exist. I'm running a breadth-first search through the subset automaton for C_4, tracking how the state set shrinks with each letter applied. Starting from the full set {0,1,2,3}, applying b collapses it to {1,2,3}, then continuing with a and b moves me through various three-element sets until I reach a two-element set {1,3} at depth 4 via the sequence b,a,a,b. I'll keep expanding the search to find when we reach a singleton set, which gives us the reset word length. ore. b-> b(0)=1,b(1)=1,b(2)=2 => {1,2}=12 (size2).\n new: 12.\n\nLevel6: from 13: a-> a(1)=2,a(3)=0 => {2,0}=02. b-> b(1)=1,b(3)=3 => {1,3}=13 (self).\n new: 02.\n\nFrom 12: a-> a(1)=2,a(2)=3 => {2,3}=23. b-> b(1)=1,b(2)=2 => {1,2}=12 (self).\n new: 23.\n\nLevel7: from 02: a-> a(0)=1,a(2)=3 => {1,3}=13 (visited). b-> b(0)=1,b(2)=2 => {1,2}=12 (visited).\n no new.\n\nFrom 23: a-> a(2)=3,a(3)=0 => {3,0}=03. b-> b(2)=2,b(3)=3 => {2,3}=23 (self).\n new: 03.\n\nLevel8: from 03: a-> a(0)=1,a(3)=0 => {1,0}=01. b-> b(0)=1,b(3)=3 => {1,3}=13 (visited).\n new: 01.\n\nLevel9: from 01: a-> a(0)=1,a(1)=2 => {1,2}=12 (visited). b-> b(0)=1,b(1)=1 => {1} (singleton!).\n\nFound singleton {1} at level 9, path length 9.\n\nSo the answer is 9.\n\nActually wait, let me double-check the path: 0123 -> 123 -> 023 -> 013 -> 012 -> 12 -> 23 -> 03 -> 01 -> {1}.\n\nThat's 10 steps (9 transitions). Hmm, let me recount: level0 to level9 is 9 transitions. Yes, 9 transitions = path length 9. Correct.\n</next_thinking>\n\nContinuing the BFS through the remaining levels, I find that from the size-2 sets at level 5, I can reach new size-2 sets at level 6, then a new size-2 set at level 7, and finally a size-1 set at level 9. Tracing back the path: 0123 → 123 → 023 → 013 → 012 → 12 → 23 → 03 → 01 → {1}, which gives 9 transitions total. So the answer is 9.",
    "signature": "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"
  }
]

#8 · Assistant

2026-06-30 21:53:18

Model · claude-sonnet-5

Thinking

I'm planning to work through the explicit pair graph construction for $C_4$ to compute $M(C_4)$ and compare it against the known value $\mathrm{rt}(C_4)=9$, then look for patterns in the general $C_n$ case before examining upper bounds from the literature and refining structural conjectures. Let me set up the automaton with states $\{0,1,2,3\}$ where $a$ cycles through them and $b$ swaps 0 and 1 while fixing the others, then start building out the pair graph vertices.

Now I'm mapping the transitions under both letters: $a$ shifts each pair element forward mod 4, while $b$ maps pairs containing 0 to pairs with 1 instead, and leaves other pairs unchanged—except {0,1} merges to a single state since both elements map to 1. I'll use BFS to find the minimum word length needed for each pair to reach the merged state, then take the maximum across all pairs to get $M(C_4)$. I'll set up a system of equations based on the graph structure. Since {0,1} has a direct edge to SINK with distance 1, I can use that as a base and work through the dependencies: {1,3} depends on {0,2}, {2,3} depends on {0,3}, {1,2} depends on {2,3}, and {0,2} and {0,3} depend on the others. I'll solve this iteratively using relaxation to find the fixed point, treating self-loops as irrelevant since they don't improve shortest paths. I need to be careful about reversing the graph edges correctly. The key insight is that BFS from SINK in the reversed graph gives me the shortest distance from each node to SINK in the original graph. Let me trace through the reverse edges systematically: the original edge from {0,1} to SINK becomes SINK pointing to {0,1} in reverse, and I'll continue reversing all the pair-state transitions to build out the full reverse adjacency list for BFS.

Running BFS from SINK level by level, I discover each pair at increasing distances: {0,1} arrives at level 1, {0,3} at level 2, {2,3} at level 3, {1,2} at level 4, {0,2} at level 5, and {1,3} at level 6. This gives me the minimum merge word length for each pair—essentially the shortest path from that pair-state to the sink state where both words are fully consumed.

So the maximum distance across all pairs is 6, meaning M(C₄) = 6. The lazy bound then predicts rt(C₄) ≤ 3 × 6 = 18, but the true value is 9, so the bound is off by a factor of 2—a concrete example of the "leakage" in the analysis. But wait, I'm noticing a discrepancy: an earlier claim said M(Cₙ) = n-1 exactly, which would give M(C₄) = 3, not 6. I need to double-check my BFS computation to see if I made an error or if that previous claim was incorrect.

Let me verify by tracing through the pair graph transitions manually. For pair {1,3}, applying action a gives {0,2}, and applying b keeps it at {1,3}. From {0,2}, action a cycles back to {1,3} while action b reaches {1,2}. Continuing this forward BFS from {1,3} to the sink: the path goes through {0,2}, then {1,2}, then {2,3}, then {0,3}, then {0,1}, and finally reaches the sink via action b—that's 6 steps total, confirming the distance of 6 for this pair.

Checking if any alternative path could be shorter: from {0,2}, the only productive move is b to {1,2} since a just cycles back. From {1,2}, action a is forced to {2,3} while b is a self-loop. The rest of the path is similarly forced, so 6 is indeed minimal. This contradicts the earlier claim that M(C_n) = n-1, since for C_4 we get 6, not 3. Let me now compute M(C_3) directly to see if the formula holds for smaller cycles.

For C_3 on {0,1,2} with a being cyclic and b mapping 0→1 while leaving others fixed, I'm working through the distances: {0,1} reaches SINK in 1 step via b. For {0,2}, action a goes to {0,1} (distance 1), while b goes to {1,2}. For {1,2}, action a goes to {0,2} and b is a self-loop. Setting up the equations: d({0,2}) = min(1+1, 1+d({1,2})) and d({1,2}) = 1+d({0,2}), which gives d({0,2}) = 2 and d({1,2}) = 3. So M(C_3) = 3, but n-1 = 2, which again doesn't match. Let me verify this by tracing the actual path: {1,2} goes to {0,2} via a, then to {0,1} via a, then to SINK via b. I see the issue now—the earlier claim that M(C_n) = n-1 was wrong because it identified the wrong pair as hardest. The pair {1,2} actually requires going almost through a full cycle twice plus extra steps, giving M(C_3) = 3, not 2. So I need to correct this error and derive the actual formula for pairwise merge distances in C_n by analyzing how the cyclic gap structure behaves under the 'a' operation.

The key insight is that 'a' preserves cyclic distances (it's just rotation), but 'b' only acts nontrivially when 0 is in the pair—it shifts one element forward or merges if the pair becomes {0,1}. So the merge distance depends on how many 'a' operations are needed to rotate one element of the pair to position 0, then apply 'b' to either shift it or trigger a merge.

Looking at the pattern from my calculations: M(C₃)=3, M(C₄)=6, and I'm conjecturing M(C_n) = n(n-1)/2 = C(n,2), which would give M(C₅)=10. Checking the structure, the pairwise merge distances for C₃ are exactly 1,2,3 and for C₄ they're 1,2,3,4,5,6—all consecutive integers up to the binomial coefficient, which strongly supports this formula.

Now I'm setting up the verification for n=5 by defining the cyclic group action: a rotates each pair by one position, and b swaps 0 with 1 in any pair containing 0 (or sends {0,1} to a sink state). I need to trace through the BFS to see if the 10 pairs again get mapped bijectively to distances 1 through 10. Now I'm mapping out the transitions for pairs containing 0 and how they move between orbits. The gap-1 orbit cycles through {0,1}, {1,2}, {2,3}, {3,4}, {0,4} where {0,1} goes directly to SINK via b, while {0,4} can reach {1,4} in the gap-2 orbit. The gap-2 orbit contains {0,2}, {1,3}, {2,4}, {0,3}, {1,4}, with {0,2} able to exit into the gap-1 orbit via b-transition to {1,2}.

Debug · Raw response blocks JSON

[
  {
    "type": "thinking",
    "thinking": "I'm planning to work through the explicit pair graph construction for $C_4$ to compute $M(C_4)$ and compare it against the known value $\\mathrm{rt}(C_4)=9$, then look for patterns in the general $C_n$ case before examining upper bounds from the literature and refining structural conjectures. Let me set up the automaton with states $\\{0,1,2,3\\}$ where $a$ cycles through them and $b$ swaps 0 and 1 while fixing the others, then start building out the pair graph vertices.\n\nNow I'm mapping the transitions under both letters: $a$ shifts each pair element forward mod 4, while $b$ maps pairs containing 0 to pairs with 1 instead, and leaves other pairs unchanged—except {0,1} merges to a single state since both elements map to 1. I'll use BFS to find the minimum word length needed for each pair to reach the merged state, then take the maximum across all pairs to get $M(C_4)$. I'll set up a system of equations based on the graph structure. Since {0,1} has a direct edge to SINK with distance 1, I can use that as a base and work through the dependencies: {1,3} depends on {0,2}, {2,3} depends on {0,3}, {1,2} depends on {2,3}, and {0,2} and {0,3} depend on the others. I'll solve this iteratively using relaxation to find the fixed point, treating self-loops as irrelevant since they don't improve shortest paths. I need to be careful about reversing the graph edges correctly. The key insight is that BFS from SINK in the reversed graph gives me the shortest distance from each node to SINK in the original graph. Let me trace through the reverse edges systematically: the original edge from {0,1} to SINK becomes SINK pointing to {0,1} in reverse, and I'll continue reversing all the pair-state transitions to build out the full reverse adjacency list for BFS.\n\nRunning BFS from SINK level by level, I discover each pair at increasing distances: {0,1} arrives at level 1, {0,3} at level 2, {2,3} at level 3, {1,2} at level 4, {0,2} at level 5, and {1,3} at level 6. This gives me the minimum merge word length for each pair—essentially the shortest path from that pair-state to the sink state where both words are fully consumed.\n\nSo the maximum distance across all pairs is 6, meaning M(C₄) = 6. The lazy bound then predicts rt(C₄) ≤ 3 × 6 = 18, but the true value is 9, so the bound is off by a factor of 2—a concrete example of the \"leakage\" in the analysis. But wait, I'm noticing a discrepancy: an earlier claim said M(Cₙ) = n-1 exactly, which would give M(C₄) = 3, not 6. I need to double-check my BFS computation to see if I made an error or if that previous claim was incorrect.\n\nLet me verify by tracing through the pair graph transitions manually. For pair {1,3}, applying action a gives {0,2}, and applying b keeps it at {1,3}. From {0,2}, action a cycles back to {1,3} while action b reaches {1,2}. Continuing this forward BFS from {1,3} to the sink: the path goes through {0,2}, then {1,2}, then {2,3}, then {0,3}, then {0,1}, and finally reaches the sink via action b—that's 6 steps total, confirming the distance of 6 for this pair.\n\nChecking if any alternative path could be shorter: from {0,2}, the only productive move is b to {1,2} since a just cycles back. From {1,2}, action a is forced to {2,3} while b is a self-loop. The rest of the path is similarly forced, so 6 is indeed minimal. This contradicts the earlier claim that M(C_n) = n-1, since for C_4 we get 6, not 3. Let me now compute M(C_3) directly to see if the formula holds for smaller cycles.\n\nFor C_3 on {0,1,2} with a being cyclic and b mapping 0→1 while leaving others fixed, I'm working through the distances: {0,1} reaches SINK in 1 step via b. For {0,2}, action a goes to {0,1} (distance 1), while b goes to {1,2}. For {1,2}, action a goes to {0,2} and b is a self-loop. Setting up the equations: d({0,2}) = min(1+1, 1+d({1,2})) and d({1,2}) = 1+d({0,2}), which gives d({0,2}) = 2 and d({1,2}) = 3. So M(C_3) = 3, but n-1 = 2, which again doesn't match. Let me verify this by tracing the actual path: {1,2} goes to {0,2} via a, then to {0,1} via a, then to SINK via b. I see the issue now—the earlier claim that M(C_n) = n-1 was wrong because it identified the wrong pair as hardest. The pair {1,2} actually requires going almost through a full cycle twice plus extra steps, giving M(C_3) = 3, not 2. So I need to correct this error and derive the actual formula for pairwise merge distances in C_n by analyzing how the cyclic gap structure behaves under the 'a' operation.\n\nThe key insight is that 'a' preserves cyclic distances (it's just rotation), but 'b' only acts nontrivially when 0 is in the pair—it shifts one element forward or merges if the pair becomes {0,1}. So the merge distance depends on how many 'a' operations are needed to rotate one element of the pair to position 0, then apply 'b' to either shift it or trigger a merge.\n\nLooking at the pattern from my calculations: M(C₃)=3, M(C₄)=6, and I'm conjecturing M(C_n) = n(n-1)/2 = C(n,2), which would give M(C₅)=10. Checking the structure, the pairwise merge distances for C₃ are exactly 1,2,3 and for C₄ they're 1,2,3,4,5,6—all consecutive integers up to the binomial coefficient, which strongly supports this formula.\n\nNow I'm setting up the verification for n=5 by defining the cyclic group action: a rotates each pair by one position, and b swaps 0 with 1 in any pair containing 0 (or sends {0,1} to a sink state). I need to trace through the BFS to see if the 10 pairs again get mapped bijectively to distances 1 through 10. Now I'm mapping out the transitions for pairs containing 0 and how they move between orbits. The gap-1 orbit cycles through {0,1}, {1,2}, {2,3}, {3,4}, {0,4} where {0,1} goes directly to SINK via b, while {0,4} can reach {1,4} in the gap-2 orbit. The gap-2 orbit contains {0,2}, {1,3}, {2,4}, {0,3}, {1,4}, with {0,2} able to exit into the gap-1 orbit via b-transition to {1,2}.",
    "signature": "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"
  }
]

#9 · Assistant

2026-06-30 21:56:21

Model · claude-sonnet-5

Thinking

I'm setting up the explicit pair graph for the $C_4$ automaton to compute $M(C_4)$ and verify that the recurrence threshold equals 9, then examining whether the bound is tight or loose. The automaton has states indexed 0 through $n-1$ with transitions where $a$ cycles through states and $b$ compresses certain states toward 0. state 1 to 0 while fixing all others, so letter $b$ merges the pair $\{0,1\}$ down to a single state. Now I'm computing the pair graph for $C_4$ with states $\{0,1,2,3\}$, where $a$ cycles through them and $b$ collapses $\{0,1\}$ — I need to track how each pair evolves under these letters to find the maximum merge distance. Now I'm setting up a BFS to find the shortest path from each pair state to the sink by computing distances on the reverse graph, working backward from the merged state to determine how many steps each pair needs to converge. So the distances from each node to SINK are: P01 at 1, P03 at 2, P13 and P23 at 3, and P02 and P12 at 4, making M(C₄) = 4. This gives a bound of 3 × 4 = 12 versus the actual value of 9, confirming the overshoot. But this contradicts the prompt's claim that M(Cₙ) = n-1 exactly—for n=4 I'm getting M=4, not 3—which suggests either my automaton convention differs from what was intended, or the definition of M in the original problem uses a different approach, perhaps focusing only on the hardest pair {0,1} under the specific "repeated a then one b" sequence rather than computing BFS-optimal shortest paths across all pairs.

Let me reconsider the transition function for b. I need to verify whether b should be a non-injective map that actually merges states (like b(0)=b(1)=0) or if the standard Černý automaton uses a different convention, possibly with relabeled states or a different mapping strategy.

Looking at the standard definition with states labeled 1 through n: a is the cyclic shift 1→2→...→n→1, and b maps 1→2 while keeping everything else fixed. This means both 1 and 2 map to 2 under b, so the image has size n-1. This is structurally identical to what I had before, just with different indexing—my setup with states 0 through 3 where b merges 0 and 1 to 0 is the same pattern, just shifted. So my BFS calculation should be correct: M(C₄) = 4, not n-1 = 3.

But now I'm checking against the known result that the reset word length for the Černý automaton C_n is exactly (n-1)², which for n=4 gives 9. My naive bound using M=4 would give 12, which doesn't match. Either the text's claim that M(C_n) = n-1 is wrong, or I've made an error in my pair graph computation. Let me trace through the transitions more carefully: with a shifting each pair forward by 1 mod 4, and b collapsing {0,1} to {0,0} while leaving {0,2}, {0,3}, {1,2}, {1,3}, {2,3} unchanged under b, I need to verify the full BFS distances to the sink state. Working through the cycle between P02 and P13, I find that d(P13)=3 and d(P02)=4, which gives d(P12)=4 as well. This confirms M(C₄)=4, contradicting the claim in the prompt that M(Cₙ)=n-1—so for C₄ we'd expect 3 but actually get 4, suggesting the original statement either uses a different definition of merge-length or needs refinement.

Now I need to build an explicit pair graph diagram for C₄ to visualize this, compute M(C₄) via BFS to verify the distance calculations, and then explore the pattern for general Cₙ by examining how the gap structure and rotation symmetry affect distances—maybe computing M(C₅) to identify a formula for M(Cₙ).

For C₅, I'm setting up the state space with 5 vertices where a rotates by +1 mod 5 and b merges states 0 and 1. The 10 pairs behave differently under b depending on whether they contain 0, 1, both, or neither—pairs like {0,1} collapse to the sink, pairs containing exactly one of {0,1} map to a pair containing 0, and pairs avoiding both stay fixed.

Now I'm computing distances to the sink by analyzing the action graph. {0,1} reaches the sink in one step, {0,4} reaches it in two steps via a, and {3,4} in three steps. For the remaining pairs, I'm tracing cycles in the distance equations—particularly {0,3} and {1,4} form a cycle where each depends on the other through the a-action, which will determine the maximum distance. So d(P03)=4, and working backward through the equations I get d(P24)=5, d(P13)=5, d(P02)=6, which checks out against the constraint. This gives me d(P12)=5 as well. Collecting all the distances for n=5, the maximum is 6, so M(C₅)=6. Now I'm looking at the pattern: M(C₄)=4 and M(C₅)=6, trying to figure out what formula might fit these values.

I'm recalling that the literature on Černý automata suggests the formula might be (n-1)(n-2)/2 + 1, which would give 4 for n=4 and 7 for n=5, but my computation shows 6 for n=5. Let me recompute the distances more carefully using a BFS layer approach to cross-validate my earlier work and see where the discrepancy might be. Now I'm setting up a shortest-path computation using iterative relaxation, treating this as a graph problem where I need to find minimum distances from each state to the sink, accounting for cycles by running Bellman-Ford style iterations until the values stabilize.

Debug · Raw response blocks JSON

[
  {
    "type": "thinking",
    "thinking": "I'm setting up the explicit pair graph for the $C_4$ automaton to compute $M(C_4)$ and verify that the recurrence threshold equals 9, then examining whether the bound is tight or loose. The automaton has states indexed 0 through $n-1$ with transitions where $a$ cycles through states and $b$ compresses certain states toward 0. state 1 to 0 while fixing all others, so letter $b$ merges the pair $\\{0,1\\}$ down to a single state. Now I'm computing the pair graph for $C_4$ with states $\\{0,1,2,3\\}$, where $a$ cycles through them and $b$ collapses $\\{0,1\\}$ — I need to track how each pair evolves under these letters to find the maximum merge distance. Now I'm setting up a BFS to find the shortest path from each pair state to the sink by computing distances on the reverse graph, working backward from the merged state to determine how many steps each pair needs to converge. So the distances from each node to SINK are: P01 at 1, P03 at 2, P13 and P23 at 3, and P02 and P12 at 4, making M(C₄) = 4. This gives a bound of 3 × 4 = 12 versus the actual value of 9, confirming the overshoot. But this contradicts the prompt's claim that M(Cₙ) = n-1 exactly—for n=4 I'm getting M=4, not 3—which suggests either my automaton convention differs from what was intended, or the definition of M in the original problem uses a different approach, perhaps focusing only on the hardest pair {0,1} under the specific \"repeated a then one b\" sequence rather than computing BFS-optimal shortest paths across all pairs.\n\nLet me reconsider the transition function for b. I need to verify whether b should be a non-injective map that actually merges states (like b(0)=b(1)=0) or if the standard Černý automaton uses a different convention, possibly with relabeled states or a different mapping strategy.\n\nLooking at the standard definition with states labeled 1 through n: a is the cyclic shift 1→2→...→n→1, and b maps 1→2 while keeping everything else fixed. This means both 1 and 2 map to 2 under b, so the image has size n-1. This is structurally identical to what I had before, just with different indexing—my setup with states 0 through 3 where b merges 0 and 1 to 0 is the same pattern, just shifted. So my BFS calculation should be correct: M(C₄) = 4, not n-1 = 3.\n\nBut now I'm checking against the known result that the reset word length for the Černý automaton C_n is exactly (n-1)², which for n=4 gives 9. My naive bound using M=4 would give 12, which doesn't match. Either the text's claim that M(C_n) = n-1 is wrong, or I've made an error in my pair graph computation. Let me trace through the transitions more carefully: with a shifting each pair forward by 1 mod 4, and b collapsing {0,1} to {0,0} while leaving {0,2}, {0,3}, {1,2}, {1,3}, {2,3} unchanged under b, I need to verify the full BFS distances to the sink state. Working through the cycle between P02 and P13, I find that d(P13)=3 and d(P02)=4, which gives d(P12)=4 as well. This confirms M(C₄)=4, contradicting the claim in the prompt that M(Cₙ)=n-1—so for C₄ we'd expect 3 but actually get 4, suggesting the original statement either uses a different definition of merge-length or needs refinement.\n\nNow I need to build an explicit pair graph diagram for C₄ to visualize this, compute M(C₄) via BFS to verify the distance calculations, and then explore the pattern for general Cₙ by examining how the gap structure and rotation symmetry affect distances—maybe computing M(C₅) to identify a formula for M(Cₙ).\n\nFor C₅, I'm setting up the state space with 5 vertices where a rotates by +1 mod 5 and b merges states 0 and 1. The 10 pairs behave differently under b depending on whether they contain 0, 1, both, or neither—pairs like {0,1} collapse to the sink, pairs containing exactly one of {0,1} map to a pair containing 0, and pairs avoiding both stay fixed.\n\nNow I'm computing distances to the sink by analyzing the action graph. {0,1} reaches the sink in one step, {0,4} reaches it in two steps via a, and {3,4} in three steps. For the remaining pairs, I'm tracing cycles in the distance equations—particularly {0,3} and {1,4} form a cycle where each depends on the other through the a-action, which will determine the maximum distance. So d(P03)=4, and working backward through the equations I get d(P24)=5, d(P13)=5, d(P02)=6, which checks out against the constraint. This gives me d(P12)=5 as well. Collecting all the distances for n=5, the maximum is 6, so M(C₅)=6. Now I'm looking at the pattern: M(C₄)=4 and M(C₅)=6, trying to figure out what formula might fit these values.\n\nI'm recalling that the literature on Černý automata suggests the formula might be (n-1)(n-2)/2 + 1, which would give 4 for n=4 and 7 for n=5, but my computation shows 6 for n=5. Let me recompute the distances more carefully using a BFS layer approach to cross-validate my earlier work and see where the discrepancy might be. Now I'm setting up a shortest-path computation using iterative relaxation, treating this as a graph problem where I need to find minimum distances from each state to the sink, accounting for cycles by running Bellman-Ford style iterations until the values stabilize.",
    "signature": "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"
  }
]

Advanced

sonnet5-notebook-investigations-20260630-211713

Inject & Branch