OctoInformBeta

OpenAI preprint, September 2026 · §2. Words and strict drift · §2

§2.3 (setting): the first backward survivor: The natural filtration and increment sum on backward-revealed inventory words, characterizing the reduced stack and stopped order count once a burger survives

Reduced state and stopped defect at the first backward survivor.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Reveal burgers and orders backwards from a boundary, prepending each new letter to the front of the revealed word and reducing it with freshest-first burger stacks. One tracks the cumulative burger-minus-order sum along the natural filtration and waits for the first step at which the reduced word retains a surviving burger. The question is what reduced state emerges at that time and how the remaining pending orders relate to the stopped burger-minus-order count.

Inventory Words and Backward Reduction

Let 𝔹={false,true}\mathbb{B} = \{\mathrm{false}, \mathrm{true}\} index the two burger types. The set 𝒪\mathcal{O} of inventory orders consists of the flexible order ⋄\diamond and fixed orders fix(b)\mathrm{fix}(b) of boolean type b∈𝔹b \in \mathbb{B}.✦ The finite burger-order alphabet 𝒜\mathcal{A} is the set consisting of burger tokens B(b)\mathrm{B}(b) for b∈𝔹b \in \mathbb{B} and order tokens O(o)\mathrm{O}(o) for o∈𝒪o \in \mathcal{O}. The letter increment Δ:𝒜→ℤ\Delta \colon \mathcal{A} \to \mathbb{Z} is the map that assigns +1+1 to every burger token and −1-1 to every order token.✦

An inventory stack evaluation state e=(ord⁡(e),burg⁡(e))e = (\operatorname{ord}(e), \operatorname{burg}(e)) consists of a finite list ord⁡(e)\operatorname{ord}(e) of pending orders in 𝒪\mathcal{O} and a finite list burg⁡(e)\operatorname{burg}(e) of boolean burger types in 𝔹\mathbb{B} stored freshest-first. The word evaluation run⁡(w,e)\operatorname{run}(w, e) of a finite word ww of letters in 𝒜\mathcal{A} on an inventory stack evaluation state ee is the evaluation state obtained by processing ww on ee: on the empty word it returns ee, and on a word aw′aw' with leading letter aa and suffix w′w' it recursively evaluates w′w' from the state obtained by applying the single-step transition on aa to ee, in which an order consumes the freshest eligible burger. The word reduction red⁡(w)=run⁡(w,([],[]))\operatorname{red}(w) = \operatorname{run}(w, ([], [])) of ww is the evaluation state obtained from empty pending orders and an empty burger stack.✦

For an infinite sequence x=(x0,x1,…)∈𝒜ℕx = (x_0, x_1, \dots) \in \mathcal{A}^{\mathbb{N}} and an integer n≥0n \ge 0, the revealed word wn(x)=(xn−1,xn−2,…,x0)w_n(x) = (x_{n-1}, x_{n-2}, \dots, x_0) is the reversal of (x0,…,xn−1)(x_0, \dots, x_{n-1}), with w0(x)=[]w_0(x) = [].✦ On the product measurable space 𝒜ℕ\mathcal{A}^{\mathbb{N}}, the natural filtration (ℱn)n≥0(\mathcal{F}_n)_{n \ge 0} generated by the coordinate process n↦xnn \mapsto x_n is given by (1): ℱn=σ(xk:0≤k≤n).\mathcal{F}_n = \sigma(x_k : 0 \le k \le n).(1)✦ For an infinite sequence x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}} and an integer n≥0n \ge 0, the letter increment sum Sn(x)∈ℤS_n(x) \in \mathbb{Z} is given by (2): Sn(x)=∑i=0n−1Δ(xi).S_n(x) = \sum_{i=0}^{n-1} \Delta(x_i).(2)✦

The survivor indicator In(x)∈{0,1}I_n(x) \in \{0, 1\}, for an infinite sequence x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}} and an integer n≥0n \ge 0, is the indicator variable equal to 11 if burg⁡(red⁡(wn+1(x)))=[]\operatorname{burg}(\operatorname{red}(w_{n+1}(x))) = [] and 00 otherwise. Given an integer-valued process Z=(Zn)n≥0Z = (Z_n)_{n \ge 0} on a sample space Ω\Omega and an element ω∈Ω\omega \in \Omega, the first hitting time of zero T0(Z)(ω)∈ℕ∪{∞}T_0(Z)(\omega) \in \mathbb{N} \cup \{\infty\} is the least n≥0n \ge 0 at which Zn(ω)=0Z_n(\omega) = 0, and ∞\infty if no such nn exists.

The Claim

Definition 1. Let x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}}. The first backward survivor time τ(x)∈ℕ∪{∞}\tau(x) \in \mathbb{N} \cup \{\infty\} is the first hitting time of zero T0(I)(x)T_0(I)(x) of the survivor indicator process n↦In(x)n \mapsto I_n(x). For an infinite sequence x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}}, the survivor order count K(x)∈ℕK(x) \in \mathbb{N} is the non-negative integer equal to 00 when τ(x)=∞\tau(x) = \infty, and to the length of ord⁡(red⁡(wτ(x)+1(x)))\operatorname{ord}(\operatorname{red}(w_{\tau(x)+1}(x))) otherwise. For n≥0n \ge 0, the stopped defect process Dn(x)∈ℝD_n(x) \in \mathbb{R} is the process k↦1−Sk+1(x)k \mapsto 1 - S_{k+1}(x) stopped at τ(x)\tau(x). If n∈ℕn \in \mathbb{N} satisfies τ(x)=n\tau(x) = n, then there exists a burger type b∈𝔹b \in \mathbb{B} (with opposite type ¬b∈𝔹\neg b \in \mathbb{B}) such that (3) holds: red⁡(wn+1(x))=(ord⁡(red⁡(wn(x))),[b])ando=fix(¬b) for all o∈ord⁡(red⁡(wn(x))).\operatorname{red}(w_{n+1}(x)) = \bigl(\operatorname{ord}(\operatorname{red}(w_n(x))),\, [b]\bigr) \quad\text{and}\quad o = \mathrm{fix}(\neg b) \text{ for all } o \in \operatorname{ord}(\operatorname{red}(w_n(x))).(3)✦ Moreover, if n∈ℕn \in \mathbb{N} satisfies τ(x)≤n\tau(x) \le n, then (4) holds: Dn(x)=K(x).D_n(x) = K(x).(4)✦

In the paper, the result reads: letters are revealed backwards from a fixed boundary by prepending one letter at each step, SjS_j is the burger-minus-order sum with S0=0S_0 = 0, J=inf⁡{j≥1:the revealed suffix has a surviving burger}J = \inf\{j \ge 1 : \text{the revealed suffix has a surviving burger}\}, and at time JJ the reduction consists of one burger and KK fixed orders of the opposite type, with SJ=1−KS_J = 1 - K. The zero-based time τ(x)\tau(x) shifts JJ by one (τ(x)+1=J\tau(x) + 1 = J), so (3) and (4) capture the reduced shape and SJ=1−KS_J = 1 - K in stopped form.

What Is Checked

The claim is stated over infinite sequences x∈𝒜ℕx \in \mathcal{A}^{\mathbb{N}} in the finite burger-order alphabet 𝒜\mathcal{A} (with order type 𝒪\mathcal{O} and letter increment Δ\Delta), evaluated via inventory stack states and word reduction on backward-revealed prefixes wn(x)w_n(x), together with the natural filtration (1), the letter increment sum (2), the survivor indicator In(x)I_n(x), the first hitting time of zero T0T_0, the first backward survivor time τ(x)\tau(x), the survivor order count K(x)K(x), the stopped defect process Dn(x)D_n(x), and indices n∈ℕn \in \mathbb{N} satisfying τ(x)=n\tau(x) = n or τ(x)≤n\tau(x) \le n.