摘要："一个将自身结论写入存储、并在之后从该存储中检索内容的 Agent，形成了一个通常被描述为单向污染的闭环。在面对只追加存储时，将该闭环推至无限存续极限会得到不同的图景：由于写入从不删除，可达状态空间在 (n-1)/n 处存在硬上界，因此结果是在两个边界之间作出选择，而不是持续衰减。在 f_0 = 0.9 时，两个模式之间的区间容纳了 220 次运行中的 3.6%，而均匀分布本应给出 20.6%；在前 15 次运行中该区间则严格为空。汇总均值所描述的运行仅占其所汇总运行的 8.2%，而中位数则占 68.2%。模型贡献的一切均由一个没有拟合参数的实测原语承载，即复制函数 γ(φ)：在 36 条 Wikidata 事实中，当 r = 0, w = 1 且 γ_crit = 1/k 时，sign(\hat{γ} - γ_crit) 在 360 次真实事实运行中预测对了 353 次漂移方向（同批次合成事实运行中为 40 次中的 39 次）。规模无法拯救存储：汇总前沿捕获率为 0.850；claude-sonnet-4.5 在 20 个随机种子中有 20 个被捕获，而我们预注册的预测是 <0.5。该区间所检验的是可区分性而非数量：在真实事实中，多值运行在该区间中的占用率是其余运行的 6.4 倍。这一现象在 f_0 ∈ {0.1, 0.3, 0.5} 时依然存在，捕获在 f_0 = 0.5 时达到峰值；在先冻结判定标准的四项干预中，相同预算下时机比比例更具主导性，而一致性门控将每个模型都推至 0.993。重采样单位为随机种子，在汇总层面设计效应为 3.75：在 44 种子对照下，支撑主张 4 的排序从 3 个种子时的 Spearman +0.98 降至 44 个种子时的 +0.31–0.80；而主张 2 的排序在该条件下完全准确（+1.00, p = 0.017）。全部 87 行评分结果均位于附录 W，其中 37 行被评为已撤回、失败、自我纠正、不可判定或已承认的局限，另有 50 行不属于这些情况。"

Abstract:"An agent that writes its conclusions into a store it later retrieves from closes a loop usually reported as one-way contamination. Taking the loop to the infinite-tenure limit against an append-only store gives a different picture: because writing never deletes, the reachable state space has a hard upper edge at (n-1)/n, so the outcome is a choice between two edges rather than a decay. At f_0 = 0.9 the interval between the two modes holds 3.6% of 220 runs where a uniform spread would put 20.6%, and is strictly empty on the first 15; the pooled mean describes 8.2% of the runs it summarises, the median 68.2%. Everything the model contributes is carried by one measured primitive with no fitted parameter, the copy function \gamma(\phi): on 36 Wikidata facts, sign(\hat{\gamma} - \gamma_{crit}), with \gamma_{crit} = 1/k at r = 0, w = 1, predicts the direction of drift on 353 of 360 real-fact runs (39 of 40 synthetic in the same batch). Scale does not rescue the store: pooled frontier capture is 0.850, with claude-sonnet-4.5 captured on 20 of 20 seeds against our registered prediction of <0.5. What the interval tests is distinguishability rather than count: on the real facts, multi-valued runs have 6.4x its occupancy of the rest. It survives at f_0 in {0.1, 0.3, 0.5}, capture peaks at f_0 = 0.5, and of four interventions with criteria frozen first, timing dominates fraction at matched budget while a consistency gate drives every model to 0.993. The resampling unit is the seed, at a design effect of 3.75 on a pooled level: under a 44-seed control the ordering supporting claim 4 collapses from Spearman +0.98 at three seeds to +0.31-0.80 at forty-four, while claim 2's ordering is exact there (+1.00, p = 0.017). All 87 graded rows are in Appendix W, 37 of them graded withdrawn, failed, self-correcting, undecidable or an acknowledged limit, against 50 that are not."