组会（Meetings）：每个活跃 UTC 日一块，分成果 / 工作 / 问题 / 待 PI 决定四栏，每条结果都带信任状态（被推翻或被红队质疑的结论不写成普通结果），问题栏包含 agent 自己的错误与更正。最新一块 = 2026-10-18 上午（PI 答复 D105：重新抽谱长期授权 → P27/P28/P29 已派出；RT-58/RT-59/RT-60/RT-61/RT-62 已分诊 → R066/D107；Q-C 改标 Questioned（相机相关）、I069 计数订正、D84c 已解决）。
Group meetings: one block per active UTC day, split into results / work / problems / PI decisions pending, every result carrying its trust state (an overturned or red-teamed claim is never written as a plain result), and the problems column includes the agent's own errors and corrections. The newest block is 2026-10-18 morning (the PI's answer D105: re-extraction is standing-authorised → P27/P28/P29 dispatched; RT-58/RT-59/RT-60/RT-61/RT-62 triaged → R066/D107; Q-C re-labelled Questioned (camera-dependent), I069's counts corrected, D84c resolved).
XARTATOMS 研究 M104（草帽星系）的热气体晕：用 XMM-Newton 与 Chandra 的 X 射线光谱，测出晕内气体的温度、密度、熵随半径的分布，并检验“引力束缚的静力气体”与“反馈加热的风”两种图景。数据是 XMM 的三次观测（中心 0900170101 加两个偏置指向），拟合用 Sherpa 的 APEC 热等离子体模型加多种背景成分。当前科学基线是 v18 拟合管线。
XARTATOMS studies the hot-gas halo of M104 (the Sombrero Galaxy): using X-ray spectra from XMM-Newton and Chandra it measures how the temperature, density and entropy of the halo gas change with radius, and tests the two competing pictures — a gravitationally bound hydrostatic gas versus a feedback-heated wind. The data are three XMM observations (central 0900170101 plus two offset pointings); the fitting uses Sherpa with an APEC thermal-plasma model plus several background components. The current science baseline is the v18 fitting pipeline.
cognition-os/00_AI_BRIEFING.md（One-paragraph project summary）/ CLAUDE.md（repo root；cognition-os/CLAUDE.md 不存在，RT-45 的 OBJ-5 已订正）
项目骨架（最新）
Project skeleton (latest)
数据流：事件文件 → ESAS/SAS 提取 → v18 联合拟合 → 径向温度/熵轮廓 →（待做）2D 图。红色虚线为已搁置的 SZ 对比。 Data flow: event files → ESAS/SAS extraction → v18 joint fitting → radial kT/entropy profiles → (to do) 2D maps. The dashed red box is the parked SZ comparison.
四问目标（2026-10-14 PI 新增）
The four-question objectives (added by the PI on 2026-10-14)
| 目标Objective | 问题Question | 状态Status | 当前最佳数字（来源）Best current number (source) | 趋势Trend | ||
|---|---|---|---|---|---|---|
| Q-AQ-A | 哪种本地拟合策略给出的投影温度可靠？ | Which local fitting strategy gives a reliable projected temperature? | QuestionedQuestioned | 内区（2–4′）参考 = MOS2 单相机（S3）：四个背景错设臂全部通过（偏差 +0.008…+0.014 keV，覆盖率 0.71–0.77）；8–10′ 无策略合格（kT ≤ 1 keV 上界撞墙 4–11%，CXB×1.1 偏差 +0.035 keV）；视场边缘与 2D 图未合格；红队 RT-02 后 D48 收窄为 Questioned（MOS2 优于 MOS1 成立，但 MOS2 自身未被证明干净 — D50/T159）— cognition-os/insights/I067_*.md；cognition-os/04_DECISION_LOG.md D50。新增（2026-10-17 03:14Z，R036/D76）：“MOS2-only 是内区参考”只在 kT ∈ [0.45, 1.0] keV 成立——该区间外 MOS2 与 MOS1 的差超过 1σ，S3 的四个背景臂没有覆盖，因此 Q-A 的答案不能外推到区间外。新增（2026-10-16，R020/D64/B024）：PI 提出“v2 mask 把盘侧遮得太狠”，mask 深度阶梯阶段 A 检验后被红队 RT-17 推翻——v18 实际把源幂律（SrcPo）固定在 0，所谓“轻 mask 只偏 ≤ 0.012 keV”需要一个 v18 没有的自由幂律成分；按 v18 的真实合同，只遮延展源+恒星会把 2–4′ kT 推高 +0.18…+0.22 keV → 全套 v2 点源 mask 在所有 v18 合同谱、Fe-L 硬度与 2D 图上继续使用，窄阶段 B 取消。教训：写 brief/决策前必须读模型成分的最终状态（所有 set/freeze 之后），不能只读声明（B024）。新增（2026-10-18 11:03Z，R068/D109，红队 RT-66 分诊）：内区 2–4′ 的 MOS2-only kT（0.72–0.75 keV）对 PI 的表述改为"条件比较器"（conditional comparator）——它未被独立佐证，不是可靠的绝对值；RT-66 重放 felhr_vs_spec.csv：2–4′ 硬度 0.7857 ± 0.0164 对谱 0.7205 [0.6965, 0.7501] 差 +0.0652 keV（+2.24σ，区间不重叠），但换阶段 5b 的谱比较器（kT_in 0.7464 [0.6564, 0.8345]）则区间重叠——"方法在 2–4′ 不一致"取决于比较器；8–10′ 的分歧（硬度平 0.785 vs 谱 0.94–1.01）不依赖比较器。D105 给 PI 的可靠性条件清单被 R068/D109 整体替换为修正版：(1) 看背景占比/偏置，不只看计数；(2) 每一个预注册对照（含历史 mask 与 GTI 半段）移动 < 0.10 keV；(3) 方法一致性必须对着一个指名的谱比较器陈述；(4) 参数边界与多起点一致分开报告；(5) 三相机一致性在 pn0 产品存在之前不可检验；(6) "σ68 ≲ 0.05 keV"只是启发式（项目自己的图中位 σ68 = 0.0589 keV，R063），已被 D110 撤下 |
Inner (2–4′) reference = MOS2-only (S3): passes all four background-misspecification arms (bias +0.008…+0.014 keV, coverage 0.71–0.77); 8–10′ none qualified (kT ≤ 1 keV bound hits 4–11 %, CXB×1.1 bias +0.035 keV); edge and the 2D map unqualified; after red team RT-02, D48 is narrowed to Questioned (MOS2 preferred over MOS1 stands, but MOS2 itself is not shown clean — D50/T159) — cognition-os/insights/I067_*.md; cognition-os/04_DECISION_LOG.md D50. New (2026-10-17 03:14Z, R036/D76): "MOS2-only is the inner reference" holds only for kT in [0.45, 1.0] keV — outside that range the MOS2–MOS1 difference exceeds 1σ and S3's four background arms do not cover it, so Q-A's answer must not be extrapolated outside that range. New (2026-10-16, R020/D64/B024): the PI asked whether the v2 mask masks the disk side too hard; the mask-depth-ladder stage A test was overturned by red team RT-17 — v18 as run freezes the source power law (SrcPo) at 0, so the "lighter mask biases kT by only ≤ 0.012 keV" result needs a free power-law component that v18 does not have; under the literal v18 contract, masking only extended sources + stars pushes the 2–4′ kT up by +0.18…+0.22 keV → the full v2 point mask stays for every v18-contract spectrum, the Fe-L hardness and the 2D maps, and the narrow stage B is cancelled. Lesson: read a model component's final state (after all set/freeze lines), not its declaration, before writing a brief or a decision (B024). New (2026-10-18 11:03Z, R068/D109, red team RT-66 triaged): the inner 2–4′ MOS2-only kT (0.72–0.75 keV) is described to the PI as a "conditional comparator" — it is not independently corroborated and is not a reliable absolute value; RT-66 replayed felhr_vs_spec.csv: at 2–4′ the hardness 0.7857 ± 0.0164 vs the spectral 0.7205 [0.6965, 0.7501] differ by +0.0652 keV (+2.24σ, non-overlapping intervals), but against the stage-5b spectral comparator (kT_in 0.7464 [0.6564, 0.8345]) the intervals overlap — "the methods disagree at 2–4′" is comparator-dependent, while the 8–10′ disagreement (hardness flat at 0.785 vs spectral 0.94–1.01) is not. The D105 reliability-condition list given to the PI is replaced wholesale by R068/D109 with the corrected list: (1) background fraction/bias, not counts alone; (2) every pre-registered control (including the historical mask and the GTI half) shifts < 0.10 keV; (3) method agreement must be stated against a named spectral comparator; (4) bounds are reported separately from multistart agreement; (5) three-camera agreement is untestable until pn0 products exist; (6) the "σ68 ≲ 0.05 keV" condition was only a heuristic (the project's own map has a median σ68 of 0.0589 keV, R063) and is dropped by D110 |
|
| Q-BQ-B | Chandra+XMM 联合的 M104 温度 profile 是什么（10′ 内）？ | What is the Chandra+XMM kT profile of M104 (within 10′)? | AMBIGUOUS（功率不足）AMBIGUOUS (underpowered) | 径向 kT 问题（升 vs 平）开放，只由预注册的匹配合同测试（D55）判定：扇区谱外热（A2 +0.20 / A1 +0.25 keV 点对比（0.9018 − 0.7038 / 0.9669 − 0.7174，来自 v18_profile_analysis/v18_complete_parameters_v2.csv；只引点估计——该 CSV 的误差列无效，D84b；外区未合格）；单温 Fe-L 硬度平（0.77–0.79 keV，2–12′，I068/T160；两相混合可给出同样 H，RT-06）；Chandra-only 与单一温度 0.725 keV 一致（p = 0.54，D56；内环在 I3 瞄准芯片、外环在 FI 芯片，D61 更正 S3 说法；污染模型版本影响 ≤ 0.012 keV）——但红队 RT-23 已分诊（R028/D69）：这条不能用来反对 XMM 的外升，因为 Chandra 2–10′ 的“环”其实只有 35–62% 的部分环带（8–10′ 的两个 ObsID 各盖对面一半），而共同 kT 检验探不到 ΔkT ≲ 0.27 keV 的差（8–10′ 的 95% 上界 0.954 keV 包含 XMM 的 0.90）→ “Chandra 等温”既没有匹配的孔径、也没有功率去证实或否证 XMM 的升；“张力 = 合同不匹配”已撤回（R009）；测试暂停：Virgo 阶段 1 的口径谱缺 v2 mask（B023）→ 阶段 1b 重抽后已交付（D76c：9/9 口径、MANIFEST 506/506 OK、BACKSCAL 真实、REG 身份 PASS），但交付的 QPB 在全部 9 个口径里大了 1e6–1e7 倍（阶段 5 执行器发现，B026/R037/D77）→ 阶段 5 于 2026-10-17 03:14Z 重派、改用经验证的修复 back_pi（derived-QPB）。阶段 5 结果（2026-10-17 07:15Z，E049/T174/R040/D80；Provisional 暂定，derived-QPB）：预注册的 D55 判定 = AMBIGUOUS（含糊）——ΔkT = kT(8–10′) − kT(2–4′)（目标 0 = 不升）的 95% 区间：谱拟合 S4 天空 A +0.13（−0.21…+0.42）/ 天空 B +0.19（−0.004…+0.42，Δχ²(0) = 3.67 < 3.84）；Fe-L 硬度 天空 A +0.01（−0.32…+0.46）/ 天空 B +0.07（−0.13…+0.33）。最小可探测 ΔkT：0.43 keV 只是高斯近似（零假设 ΔkT 的 1.96·σ）；按预先声明的非参数定义（50% 功率、95% 区间），两种方法的最小可探测 ΔkT 都大于 0.50 keV（注入到 +0.50 仍未达到）。给大一学生的大白话：这些数据既没有显示、也没有排除约 0.5 keV 以内的上升——这个检验太弱（功率不足），它不是“没有上升”的证明。两个方法是同一个投影用了两遍（同一批 MOS2 谱），不是两次独立探测。注意事项：QPB 是推导出来的修复（交付的 back.pi 大 1e6–1e7 倍）；天空 B 是事后的，且拟合在同一个 8–10′ 谱上（把天空幅度作剖面扫描后，S 的区间变成 −0.054…≥+0.535，上端开放）；只用了一台相机。独立专家复核（Opus，D59a；来源 .agent/jobs/2026-10-17-stage5-expert-rederivation/REPORT.opus.md）确认 AMBIGUOUS，并给出上述两点修正（R042/D82）。P22 的 QPB 重查待定；红队 RT-15（0.7 keV 前景成分：v18 该槽位实为 0 且冻结 → D62 只是外区 kT 的附带假设）与 RT-16（固定 Z = 0.3：Z = 1.5 真值按 Z = 0.3 拟合在 8–10′ 只低 0.017 keV → 不是外升成因，D63）已分诊。组合压力对照的判据（D55/D71，RT-26 22:02Z 重放 D54/I069 算术后加的预先补充）：SP+CXB+QPB 的同号线性相加是 +0.142 keV，比 0.10 keV 的门还大，所以该对照按阶段 5 在阶段 1b 产品上实测的那一条组合臂来判，不按线性总和判；如果阶段 5 的组合臂本身把 ΔkT 移动 ≥ 0.10 keV，判定一律封顶为 AMBIGUOUS（实测组合臂的最大移动是 S 0.081 / H 0.006 keV，未触发该封顶）。 — cognition-os/experiments/E049_*.md；cognition-os/04_DECISION_LOG.md D80/D82；cognition-os/07_REVISION_LOG.md R040/R042 |
The radial kT question (rise vs flat) is open, decided only by the pre-registered matched-contract test (D55): sector spectra are hotter outside (A2 +0.20 / A1 +0.25 keV point contrasts (0.9018 − 0.7038 / 0.9669 − 0.7174, from v18_profile_analysis/v18_complete_parameters_v2.csv; point estimates only — that CSV's error columns are invalid, D84b; outer unqualified)); the one-temperature Fe-L hardness is flat (0.77–0.79 keV over 2–12′, I068/T160; a two-phase mixture can give the same H, RT-06); Chandra-only is consistent with one temperature 0.725 keV (p = 0.54, D56; the inner ring is on the I3 aim-point chip and the outer ring on FI chips — D61 corrects the "S3" wording; the contamination-model version moves kT by ≤ 0.012 keV) — but red team RT-23 is triaged (R028/D69): this cannot be used against the XMM rise, because the Chandra 2–10′ "rings" are only 35–62 % partial annuli (the two 8–10′ ObsIDs cover opposite halves) and the common-kT test cannot detect a ΔkT below ≈ 0.27 keV (the 8–10′ 95 % upper bound of 0.954 keV contains the XMM 0.90) → "Chandra isothermal" has neither a matched aperture nor the power to corroborate or refute the XMM rise; "tension = contract mismatch" is withdrawn (R009); the test was paused: the Virgo stage-1 aperture spectra lack the v2 mask (B023) → stage 1b has since been delivered (D76c: 9/9 apertures, MANIFEST 506/506 OK, real BACKSCAL, REG identity PASS), but the delivered QPB is 1e6–1e7× too large in all 9 apertures (found by the stage-5 executor, B026/R037/D77) → stage 5 was re-dispatched on 2026-10-17 03:14Z on the validated repair back.pi (derived-QPB). Stage 5's result (2026-10-17 07:15Z, E049/T174/R040/D80; Provisional, derived-QPB): the pre-registered D55 verdict is AMBIGUOUS — the 95 % intervals for ΔkT = kT(8–10′) − kT(2–4′) (target 0 = no rise) are: the spectral fit S4 sky A +0.13 (−0.21…+0.42) / sky B +0.19 (−0.004…+0.42, Δχ²(0) = 3.67 < 3.84); the Fe-L hardness sky A +0.01 (−0.32…+0.46) / sky B +0.07 (−0.13…+0.33). The minimum detectable ΔkT: 0.43 keV is only the Gaussian approximation (1.96·σ of the null ΔkT); under the pre-declared non-parametric definition (50 % power, 95 % interval) the minimum detectable ΔkT is larger than 0.50 keV for both methods (not reached up to +0.50). In plain words for a first-year student: these data neither show nor exclude a rise below about 0.5 keV — the test is too weak (underpowered), and it is not a "no rise". The two methods are one projection applied twice (the same MOS2 spectra), not two independent detections. Caveats: the QPB is a derived repair (the delivered back.pi is 1e6–1e7× too large); sky B is post hoc and was fitted on the same 8–10′ spectrum (with the sky amplitudes profiled, S's interval becomes −0.054…≥+0.535, its upper end open); one camera only. An independent expert re-derivation (Opus, D59a; source .agent/jobs/2026-10-17-stage5-expert-rederivation/REPORT.opus.md) confirms AMBIGUOUS and supplies those two corrections (R042/D82). The P22 QPB re-check is pending; red teams RT-15 (a 0.7 keV foreground component: that v18 slot is actually 0 and frozen → D62 is only a stated assumption for the outer kT) and RT-16 (fixed Z = 0.3: a Z = 1.5 truth fitted with Z = 0.3 reads only 0.017 keV low at 8–10′ → not the cause of the rise, D63) are triaged. The criterion for the combined-stress control (D55/D71; a pre-data addendum made by RT-26 at 22:02Z after replaying the arithmetic of D54/I069): the SP+CXB+QPB same-sign linear sum is +0.142 keV, larger than the 0.10 keV gate, so that control is judged on the measured stage-5 combined arm on the stage-1b products, not on the linear sum; if the stage-5 combined arm itself shifts ΔkT by ≥ 0.10 keV, the verdict is capped at AMBIGUOUS (the measured combined arm's largest shift is S 0.081 / H 0.006 keV, so that cap was not triggered). — cognition-os/experiments/E049_*.md; cognition-os/04_DECISION_LOG.md D80/D82; cognition-os/07_REVISION_LOG.md R040/R042 |
|
| Q-CQ-C | 为何 ~13′ 亮度反而比内侧和 15–30′ 更亮？ | Why is the intensity brighter at ~13′ than inside and at 15–30′? | Resolved: artefactResolved: artefact | 假象（Provisional，I065）：flux 已是每 arcmin²（B020），旧 flux/area 双重归一；凸起大小 = 仅 Angle2、1.0–2.0，取决于 bin 配对/世代（任何世代都没有保存流量误差；1.37 只是其中一个世代），冻结 kT 后 1.07；图像 1.041±0.068 无超出；WARP 星系团拟合为 0 — cognition-os/insights/I065_*.md；经受住红队 RT-01（4 条次要、0 条实质；图像检验 −2.7σ 支持无超出，I065 修订）。2026-10-16 23:24Z：红队 RT-28 已分诊（R033/D73）——“图/v18 比值 1.00 → 3.32 沿径向上升”不是物理上升：它的 Angle1 行不可重放（R7-9 没有 .fitting 会话）、不可复现（R9-11 有三代，CSV 比可重放会话高 58%）或不显著（R11-13 Δstat 1.36；R12-14 Δstat 0.0），而可复现的 Angle2 是平的 1.26 → 1.18；图内检验（E041：R11–13/R9–11 = 0.949 ± 0.029 / 1.029 ± 0.037）仍显示无 13′ 超出 → R030 的“无像过剩不成立”本身被撤回，图/v18 分支不可检验，直到每个区域钉住一个 v18 世代（提案 P21：CoPI，需 PI；在钉住的 v18 合同上重跑 Angle1 R7-9…R12-14，也补上缺失的会话）。开放：外环拟合方法；15–45′ profile 为 Questioned（E039：偏移拟合简并/不自洽，需 0501 ODF + Virgo 重处理） |
Artefact (Provisional, I065): the saved flux is already per arcmin² (B020), so the old flux/area double-normalised; the bump size is Angle2-only, 1.0–2.0 depending on bin pairing/generation (no flux errors were ever saved; 1.37 is one generation), 1.07 with kT frozen; images 1.041±0.068 show no excess; the WARP cluster fits to 0 — cognition-os/insights/I065_*.md; it survived red team RT-01 (4 minor, 0 material objections; the image test at −2.7σ supports no excess, I065 amendment). 2026-10-16 23:24Z: red team RT-28 is triaged (R033/D73) — the "image/v18 ratio 1.00 → 3.32 rising radially" is not a physical rise: its Angle1 rows are unreplayable (R7-9 has no .fitting session), unreproducible (R9-11 exists in three generations, the CSV 58 % above the replayable session) or non-significant (R11-13 Δstat 1.36; R12-14 Δstat 0.0), while the reproducible Angle2 is flat at 1.26 → 1.18; the image-internal test (E041: R11–13/R9–11 = 0.949 ± 0.029 / 1.029 ± 0.037) still shows no 13′ excess → R030's "no image excess is false" is itself withdrawn, and the image/v18 branch is not testable until one v18 generation per region is pinned (proposed P21: CoPI, needs the PI; re-run Angle1 R7-9…R12-14 on the pinned v18 contract, which also supplies the missing session). Open: the outer-ring fit method; the 15–45′ profile is Questioned (E039: offset fits degenerate/inconsistent; needs the 0501 ODF + Virgo re-reduction) |
|
| Q-DQ-D | 2D 强度图 + 2D 温度图（区分 feedback 与引力束缚气体） | 2D intensity map + 2D kT map (feedback vs gravitationally bound gas) | OpenOpen | 亮度图已完成；2D 代理图（阶段 2，E038）：Fe-L × 象限锚点；联合 MOS2+pn0 的小轴 vs 盘对比在 ≲ 0.08–0.11 keV 内（校准区间，随半径/几何变化，非对称：±45° 2–5′ [−0.056,+0.076] keV）——对比度响应 R≈0.67–1.0 决定灵敏度；N>S 只是提示（跨臂族 p≈1），不是探测（E038 + RT-07 修订）— cognition-os/experiments/E038_*.md（D49）。新增（2026-10-16）：2D 强度（E042/R016）— 2–5′ 的盘对齐不对称（小轴/盘 0.674 ± 0.100）与平滑 β 椭圆率一致（注入预测 0.752 [0.656, 0.846]），≥ +30% 小轴双锥在 2–5′ 被排除；2D kT（E043/R018）— 最佳估计器 = Fe-L 2 能带 + MOS1（自由归一）在 cone45 bin 上（σ ×1.12，注入无偏），但只适用于 3 相机共同足迹（面积的 56%），第二臂一致；生产图（P12/E045）：逐 bin 数值复现 阶段 3，读数 = 2–5′ 无小轴 vs 盘对比、内区 N−S +0.10 keV 提示（仅一种方法，待定；东箭头曾画成朝西，修正中，修好前不展示该图）。折叠 2D kT 图 v2（E046/T171，Provisional 暂定，见状态页 §2h）：四象限关于光学长/短轴折叠（每箱约 4 倍计数），主估计器 E1 = MOS2 + pn0，v2 mask 全程应用；2–12′ 的逐箱 kT 中位带为 0.70–0.80 keV（逐箱范围 0.53–0.89 keV，全局 0.736 keV [0.723, 0.748]，8/21 箱在带外——措辞按 RT-20/D68/R025 收窄），全角度斜率 +0.009 keV/角分（95 % [−0.003, +0.020]，+1.5σ）→ 对这种单温代理而言无 ≥ 2σ 的内冷外热；每 5′ 升 0.1 keV 的梯度在 E046 的单温代理注入里 91% 的次数能被找回（仅这一个温代理的自检；与真实逐楔 2–8′ 检验对应的统计量在 E047 端到端模拟里功率只有 27–36%，所以这 91% 不是探测功率）；盘楔外箱 0.85–0.88 vs 小轴楔 0.61–0.77 keV 只是提示（≈ 1.4σ，pn0 主导、近视场边缘）；MOS2 与 pn0 逐箱一致（p = 0.60）；pn0 QPB 按 M104 自己的 0.772 ± 0.016 缩放会把斜率推到 ≈ 2σ → 趋势对背景敏感（缩放仅测试用）；< 2′ 核被 mask 遮住，冷却流在此不可检验；E045 的内区 N−S +0.10 keV 提示经 RT-19 多重比较后降为 Questioned（R022）— cognition-os/experiments/E046_*.md |
Intensity map done; the 2D proxy map (stage 2, E038): Fe-L × quadrant anchors; the joint MOS2+pn0 minor-axis vs disk contrast lies within ≲ 0.08–0.11 keV (calibrated intervals, radius/geometry-dependent, asymmetric: ±45° 2–5′ [−0.056,+0.076] keV) — the contrast response R≈0.67–1.0 sets the sensitivity; N>S is only a hint (cross-arm familywise p≈1), not a detection (E038 + RT-07 amendment) — cognition-os/experiments/E038_*.md (D49). New (2026-10-16): 2D intensity (E042/R016) — the 2–5′ disk-aligned asymmetry (minor/disk 0.674 ± 0.100) is consistent with smooth β ellipticity (injection prediction 0.752 [0.656, 0.846]); a ≥ +30 % minor-axis bicone is excluded at 2–5′; 2D kT (E043/R018) — the best estimator is the Fe-L 2-band + MOS1 (free normalisation) on cone45 bins (σ ×1.12, unbiased in injections), but only on the 3-camera common footprint (56 % of the area); the second arm agrees; the production maps (P12/E045) replay stage 3 exactly and read: no minor-axis vs disk contrast at 2–5′, an inner N−S +0.10 keV hint (one method only, pending; the East arrow was once drawn toward West — the fix is in progress and the map is not shown until it is). The folded 2D kT map v2 (E046/T171, Provisional; see state §2h): the four quadrants folded about the optical major/minor axes (≈4× counts per bin), primary estimator E1 = MOS2 + pn0, the v2 mask applied throughout; the per-bin kT over 2–12′ has a median band of 0.70–0.80 keV (per-bin range 0.53–0.89 keV; global 0.736 keV [0.723, 0.748]; 8/21 bins outside the band — wording narrowed per RT-20/D68/R025) and the all-angle slope is +0.009 keV/arcmin (95 % [−0.003, +0.020], +1.5σ) → no inner-cool/outer-hot trend at ≥ 2σ for this one-temperature proxy; a gradient of +0.1 keV per 5′ is recovered 91 % of the time in E046's own one-temperature-proxy injection (that proxy's self-test only; the statistic matching the real per-wedge 2–8′ test has a power of only 27–36 % in the E047 end-to-end mock, so this 91 % is not a detection power); the disk wedge's outer cells 0.85–0.88 vs the minor wedge 0.61–0.77 keV are suggestive only (≈ 1.4σ, pn0-driven, near the field edge); MOS2 and pn0 agree bin by bin (p = 0.60); scaling pn0's QPB to M104's own 0.772 ± 0.016 pushes the slope to ≈ 2σ → the trend is background-sensitive (the scaling is test-only); the core inside 2′ is masked, so a cooling flow cannot be tested here; E045's inner N−S +0.10 keV hint is downgraded to Questioned after RT-19's multiplicity check (R022) — cognition-os/experiments/E046_*.md |
2D 温度图 v2（四象限折叠，E046/T171，暂定）
The 2D kT map v2 (four quadrants folded, E046/T171, Provisional)
总览图暂不展示（RT-32 的 OBJ-9，2026-10-17 04:06Z）：E046 的六联总览图含一个显示缺陷（东箭头被画成朝西），而红队 RT-32 用 sha256 证实本手册此前嵌入的副本与修复前的 08:41Z 文件逐字节相同，项目里不存在任何修复后的版本 → 按第 13 步自己的承诺，本页不再展示该图；无缺陷的分图（折叠温度图两档、径向剖面、对称模型展示、偏差图）与全部数字见状态页 §2h。数值本身不受影响：2–12′ 的逐箱 kT 中位带为 0.70–0.80 keV（逐箱范围 0.53–0.89 keV，全局 0.736 keV），全角度斜率 +0.009 keV/角分（+1.5σ）→ 对这种单温代理而言无 ≥ 2σ 的内冷外热。
The overview figure is not shown for now (RT-32's OBJ-9, 2026-10-17 04:06Z): E046's six-panel overview carries a display defect (the East arrow drawn toward West), and red team RT-32 verified by sha256 that the copy this handbook had embedded is byte-identical to the pre-fix 08:41Z file, with no fixed version existing anywhere in the project → per Step 13's own promise, this page no longer shows that figure; the defect-free component figures (the folded temperature maps on both colour scales, the radial profiles, the symmetric model display, the deviation map) and all the numbers are on the state page §2h. The values themselves are unaffected: the per-bin kT over 2–12′ has a median band of 0.70–0.80 keV (per-bin range 0.53–0.89 keV; global 0.736 keV) and the all-angle slope is +0.009 keV/arcmin (+1.5σ) → no inner-cool/outer-hot trend at ≥ 2σ for this one-temperature proxy.
更早的科学目标
Older science objectives
| 目标Objective | 状态Status | 当前最佳数字（来源）Best current number (source) | ||
|---|---|---|---|---|
| 密度/熵轮廓（S06 主线） | Density/entropy profile (S06 main line) | ProvisionalProvisional | K ≈ 67 → 407 keV cm²（10–43 kpc）；t_cool ≈ 10 → 137 Gyr — figure/radial_thermo_profiles_v18/entropy_cooling_time_profile_v18_v15beta.csv。kpc 用 D = 11.3 Mpc（1′ = 3.287 kpc，build_m104_radial_thermo_profiles.py:40；该距离未记录在 cognition-os）——与 2D kT 图的 8.9 Mpc 不同，见状态页 §1a |
K ≈ 67 → 407 keV cm² (10–43 kpc); t_cool ≈ 10 → 137 Gyr — figure/radial_thermo_profiles_v18/entropy_cooling_time_profile_v18_v15beta.csv. kpc uses D = 11.3 Mpc (1′ = 3.287 kpc, build_m104_radial_thermo_profiles.py:40; that distance is recorded nowhere in cognition-os) — a different distance from the 2D kT figures' 8.9 Mpc; see the state page §1a |
| SZ / X 射线张力（~2 倍差异） | SZ / X-ray tension (~2× discrepancy) | ParkedParked | X 射线导出 SZ 比参考高约 2 倍；数据版本问题 — cognition-os/00_AI_BRIEFING.md（S04） |
X-ray-derived SZ is ~2× the reference; data-version issue — cognition-os/00_AI_BRIEFING.md (S04) |
未解问题
Open problems
| 问题Problem | 为什么重要Why it matters | 最佳假设Best hypothesis | 下一个检验Next test | 负责人Owner | 状态Status | ||||
|---|---|---|---|---|---|---|---|---|---|
| Q-A：无策略被认证；全网格验证未跑，需加速 | Q-A: no strategy certified; the full validation grid has not run and needs acceleration | 没有“谁对谁错”的判据，任何温度图都建立在未验证的方法上 | Without a criterion, every temperature map rests on an unvalidated method | 内区参考已收窄（D50：MOS2 优于 MOS1 成立，MOS2 自身未证明干净；R036 再收窄：“MOS2-only 是内区参考”只在 kT ∈ [0.45, 1.0] keV 成立——在该区间外，MOS2 与 MOS1 的差超过 1σ，S3 的四个背景臂没有覆盖）；外区/边缘仍不合格——S4 宽 kT 多起点合同的测试背景偏差 ≤ 0.05 keV，但严格合格门失败 → 外区 kT 只作探索性引用；D54 修订（RT-09）：±0.05 keV 是逐臂中位偏差包络，逐拟合 68% 散布 ≈ ±0.15 keV 须由拟合区间给出，SP+CXB+QPB 组合压力未测（同号和 +0.14 keV @0.7 keV），撞顶 1–2/60（T162 修订后的口径；I069 明确禁止引用合并的“2–3%”，合并值 3/480 = 0.63% 低估了最差单元 3–5 倍；D51/D54/T162/I069） | The inner reference is narrowed (D50: MOS2 preferred over MOS1 stands; MOS2 itself not shown clean; R036 narrows it further: "MOS2-only is the inner reference" holds only for kT in [0.45, 1.0] keV — outside that range the MOS2–MOS1 difference exceeds 1σ and S3's four background arms do not cover it); outer/edge still not qualified — the S4 wide-kT multi-start contract sees tested background biases ≤ 0.05 keV but fails the strict gate → outer kT stays exploratory; D54 amendment (RT-09): the ±0.05 keV is the median-bias envelope of one-at-a-time stress arms, the per-fit 68 % scatter ≈ ±0.15 keV must come from the fit interval, the combined SP+CXB+QPB stress is untested (same-sign sum +0.14 keV at 0.7 keV), ceiling hits 1–2/60 (T162's amended wording; I069 explicitly forbids quoting the pooled "2–3 %" — the pooled 3/480 = 0.63 % understates the worst cell by 3–5×; D51/D54/T162/I069) | 阶段 3/4 已完成（S4 外区/边缘/2D 均未过严格门）；mask 深度阶梯阶段 A 的"内区谱可轻 mask"结论被 RT-17 推翻（R020/D64：v18 把 SrcPo 固定在 0，轻 mask 偏 2–4′ kT +0.18…+0.22 keV → 全套 v2 点源 mask 保留，窄阶段 B 取消）；阶段 5 匹配合同测试（D55）已于 2026-10-17 03:14Z 重派（D76c）并已于 07:15Z 完成：判定 = AMBIGUOUS（功率不足）（E049/T174/R040/D80；derived-QPB）——谱拟合 S4 的 ΔkT(8–10′−2–4′) = +0.13（天空 A，95% −0.21…+0.42）/ +0.19（天空 B，−0.004…+0.42），Fe-L 硬度 +0.01（−0.32…+0.46）/ +0.07（−0.13…+0.33）；最小可探测 ΔkT 的高斯近似是 0.43 keV，而按预先声明的非参数定义（50% 功率、95% 区间）两种方法都 > 0.50 keV（R042/D82）→ 这个检验太弱，不是“没有上升”；详见本表 Q-B 行与提案 P23 | Stages 3/4 are done (S4 fails the strict gate in the outer/edge/2D cells); mask-depth-ladder stage A's "lighter mask safe for inner spectra" conclusion was overturned by RT-17 (R020/D64: v18 freezes SrcPo at 0, a lighter mask biases the 2–4′ kT by +0.18…+0.22 keV → the full v2 point mask stays, the narrow stage B is cancelled); the stage-5 matched-contract test (D55) was re-dispatched on 2026-10-17 03:14Z (D76c) and finished at 07:15Z: the verdict is AMBIGUOUS (underpowered) (E049/T174/R040/D80; derived-QPB) — the spectral fit S4's ΔkT(8–10′−2–4′) = +0.13 (sky A, 95 % −0.21…+0.42) / +0.19 (sky B, −0.004…+0.42) and the Fe-L hardness +0.01 (−0.32…+0.46) / +0.07 (−0.13…+0.33); the minimum detectable ΔkT's Gaussian approximation is 0.43 keV, while under the pre-declared non-parametric definition (50 % power, 95 % interval) both methods exceed 0.50 keV (R042/D82) → this test is too weak; it is not a "no rise"; see this table's Q-B row and proposal P23 | I067/D48–D50；E044/D60；阶段 5 = D55/E049I067/D48–D50; E044/D60; stage 5 = D55/E049 | 待定 PendingPending |
| Q-B：预注册的匹配合同检验（阶段 5）已给出 AMBIGUOUS——功率不足；PI 已答复（D105），P27/P28/P29 已派出 | Q-B: the pre-registered matched-contract test (stage 5) returned AMBIGUOUS — underpowered; the PI has answered (D105) and P27/P28/P29 are dispatched | 10′ 内的 Chandra+XMM 温度 profile 是四问之一，也是温度图的输入 | The Chandra+XMM kT profile within 10′ is one of the four questions and an input to the temperature map | J-linked（ACIS CXB 自由）0.718/0.757/0.782/0.934 keV；Chandra-only 与 v18 每 ≤1σ 一致；但 c_ACIS 常压在 0.7 下界、c_MOS2 在 4′ 外压 1.3 上界 → 联合归一化 Questioned（T155） | J-linked (ACIS CXB free) 0.718/0.757/0.782/0.934 keV; Chandra-only agrees with v18 within 1σ per ring; but c_ACIS often sits at its 0.7 lower bound and c_MOS2 at 1.3 beyond 4′ → joint normalisation Questioned (T155) | 阶段 5 已完成（2026-10-17 07:15Z，E049/T174/R040/D80；Provisional 暂定，derived-QPB）：预注册的 D55 判定 = AMBIGUOUS（含糊）——功率不足，不是“没有上升”。ΔkT = kT(8–10′) − kT(2–4′)（目标 0）的 95% 区间：谱拟合 S4 天空 A +0.13（−0.21…+0.42）/ 天空 B +0.19（−0.004…+0.42，Δχ²(0) = 3.67 < 3.84）；Fe-L 硬度 天空 A +0.01（−0.32…+0.46）/ 天空 B +0.07（−0.13…+0.33）。最小可探测 ΔkT：0.43 keV 只是高斯近似（零假设 ΔkT 的 1.96·σ）；按预先声明的非参数定义（50% 功率、95% 区间）两种方法都大于 0.50 keV（注入到 +0.50 仍未达到）→ 用现有数据无法裁决 0.1 keV 量级的问题。两个方法是同一个投影用了两遍（同一批 MOS2 谱）。天空 B 是事后的、且拟合在同一个 8–10′ 谱上（剖面扫描后 S 的区间为 −0.054…≥+0.535，上端开放）；只用一台相机。独立专家复核（Opus，D59a）确认 AMBIGUOUS（.agent/jobs/2026-10-17-stage5-expert-rederivation/REPORT.opus.md，R042/D82）。PI 已于 2026-10-18 05:22Z 答复（D105）：“重新抽谱一直是可以的”——用任何 region 设计重新抽谱长期授权，只受 Virgo 资源门限制（内存 ≤60%，05:22Z 实测 ≈58%）；选项 (b) 不再"需要 PI 批准"，只是受内存门限制；据此派出 P27（MOS1+pn0 抽 R2-4…R8-10 加 R10-12，阶段 1b 合同，整条 profile 共享扰动参数联合拟合，先写预注册）、P28（PI 命令 D106：九能段面亮度图，codexnim 运行中）、P29（0.5–1.2 keV 次要 S4 臂 + R11-13_Angle2 通道，codexnim 运行中）。P22 的 QPB 重生成已于 11:04Z 被接受（9/9 口径过量级门，R 0.884–1.005），阶段 5b 在其上重跑已于 12:48Z 分诊：判定 AMBIGUOUS 不变，天空 B 的 ΔkT +0.1886 → +0.1982（−0.034…+0.478），QPB 构造系统误差 0.010 keV（R046/D86）。红队 RT-39（阶段 5 判定的措辞）已于 11:04Z 分诊（R045/D85），提了 12 条异议，但没有一条推翻 AMBIGUOUS：三条 D55 门（RISE 的下界、NO RESOLVED RISE 的上界、PROJECTION-COMPATIBLE 的触发条件）全部按其字面失败，所以 AMBIGUOUS 成立；被改变的是理由与措辞——(a) “0.43 keV 是 50% 功率下的最小可探测值”这个说法不成立（实测功率在 Δ = +0.50 处最高只有 19%，执行器自己的 JSON 里 50% 功率那一栏是空的）；(b) 天空 B 的 95% 下界在插值网格上会在 −0.0150…+0.0135 keV 之间摆动（Δχ²(0) 在 3.22–4.31 之间跨过 3.84 的门）→ “差 0.004 keV 没过门”这句话没有误差棒；(c) “主天空 B”其实是在 8–10′ 这个环上自己拟合出来的（不是独立确定），那 0.004 keV 的边缘是从这个循环性继承来的；(d) 外环硬度估计的 95% “区间”有 41.8% 的样本被裁剪在反演分支端（0.45/1.22 keV），σ_D 的约 1/3 是设计造成的、不是数据；(e) QPB 尺度扫描（×0.80–1.40，已执行）显示天空 A 的 ΔkT 在 ×0.85 处翻号（−0.006 keV）并在 ±10% QPB 变化里摆动 ±0.10 keV，而天空 B 保持 +0.129…+0.228 → “QPB 锚点系统误差 ≈ 0”说得太强；(f) 2T 那一支只死在参数边界上（kT2 顶在预先注册的 5 keV 帽），4 条接受判据里满足了 3 条；(g) 内环 R2-4 的单温拟合本身被正式拒绝（p = 0.010 天空 B / 0.019 天空 A）→ ΔkT 的区间可能偏窄；(h) E049 的控制 breached 清单少列了一条（天空 A 的天空归一 ×1.2，S：−0.114 keV）。→ 判定不变，这些都进了 E049 的措辞与限制清单（R045/D85，已分诊）。其他红队：RT-33/RT-34/RT-35 → R038/D78；RT-36（P19/P20 设计）无报告死亡；RT-37（Q-B 与文献对照，7 条实质）与 RT-38（P19/P21 设计，7 条）已于 10:14Z 分诊 → R044/D84；RT-40（P22 bundle，16 条 → R048/D88）、RT-41（折叠图统计量 → R047/D87）、RT-42（P23 功率投影 → R048/D88）均已分诊。截至 2026-10-18 04:43Z：RT-01…RT-60 共 55 支已分诊（RT-36 无报告死亡；RT-58/RT-59/RT-60 运行中），其中 RT-46 → R055/D95（偏置指向当天空已撤回）、RT-47 → R056/D96、RT-48 与 RT-50 → R059/D99（P19 终报）、RT-49 → R058/D98（Q-B 选项与天空归一）；RT-51（攻击 RT-48 O1b 的解释）→ R061/D100、RT-52（项目记录一致性审计，11 条）→ R060、RT-53（P26 设计审计：P26 已撤回）→ R061/R062/D100/D101、RT-54（选项 b 可行性）→ R062/D101、RT-55（Q-D 温度图主张）→ R063/D102、RT-56（本手册 vs cognition-os 审计）与 RT-57（D84c）→ R064/D103，均已分诊。见本页“研究进展”与状态页 §5 |
Stage 5 is done (2026-10-17 07:15Z, E049/T174/R040/D80; Provisional, derived-QPB): the pre-registered D55 verdict is AMBIGUOUS — underpowered, not a "no rise". The 95 % intervals for ΔkT = kT(8–10′) − kT(2–4′) (target 0) are: the spectral fit S4 sky A +0.13 (−0.21…+0.42) / sky B +0.19 (−0.004…+0.42, Δχ²(0) = 3.67 < 3.84); the Fe-L hardness sky A +0.01 (−0.32…+0.46) / sky B +0.07 (−0.13…+0.33). The minimum detectable ΔkT: 0.43 keV is only the Gaussian approximation (1.96·σ of the null ΔkT); under the pre-declared non-parametric definition (50 % power, 95 % interval) both methods exceed 0.50 keV (not reached up to +0.50) → a 0.1 keV decision is out of reach with the existing data. The two methods are one projection applied twice (the same MOS2 spectra). Sky B is post hoc and was fitted on the same 8–10′ spectrum (profiled, S's interval is −0.054…≥+0.535, its upper end open); one camera only. An independent expert re-derivation (Opus, D59a) confirms AMBIGUOUS (.agent/jobs/2026-10-17-stage5-expert-rederivation/REPORT.opus.md, R042/D82). The PI answered on 2026-10-18 05:22Z (D105): "re-extracting spectra has always been allowed" — re-extraction with ANY region design is standing-authorised, limited only by the Virgo resource gates (memory ≤ 60 %, measured ≈58 % at 05:22Z); option (b) no longer "needs PI approval", it is only limited by the memory gate; accordingly P27 (MOS1+pn0 for R2-4…R8-10 plus R10-12 on the stage-1b contract, whole-profile joint fit with shared nuisance parameters, pre-registration first), P28 (PI order D106: the nine-band surface-brightness images, running on codexnim) and P29 (the 0.5–1.2 keV secondary S4 arm plus the R11-13_Angle2 lane, running on codexnim) were dispatched. P22's QPB regeneration was accepted at 11:04Z (9/9 apertures pass the magnitude gate, R 0.884–1.005) and stage 5b's re-run on it was triaged at 12:48Z: the AMBIGUOUS verdict is unchanged, sky B's ΔkT +0.1886 → +0.1982 (−0.034…+0.478), QPB-construction systematic 0.010 keV (R046/D86). Red team RT-39 (the stage-5 verdict's wording) was triaged at 11:04Z (R045/D85); it filed 12 objections and not one overturns AMBIGUOUS: all three D55 gates (RISE's lower bound, NO RESOLVED RISE's upper bound, PROJECTION-COMPATIBLE's trigger) fail as literally written, so AMBIGUOUS stands; what changes is the reason and the wording — (a) "0.43 keV is the minimum detectable value at 50 % power" does not hold (the measured power tops out at 19 % at Δ = +0.50, and the executor's own JSON has an empty 50 %-power field); (b) sky B's 95 % lower bound swings between −0.0150…+0.0135 keV with the interpolation grid (Δχ²(0) straddles the 3.84 gate between 3.22 and 4.31) → "it missed the gate by 0.004 keV" has no error bar; (c) the "primary sky B" is itself a fit on the 8–10′ annulus (not an independent determination), so that 0.004 keV margin is inherited from that circularity; (d) 41.8 % of the outer hardness estimator's 95 % samples are clipped at the inversion branch ends (0.45/1.22 keV), so about 1/3 of σ_D is design, not data; (e) a QPB scale sweep (×0.80–1.40, executed) shows sky A's ΔkT flips sign at ×0.85 (−0.006 keV) and swings ±0.10 keV for a ±10 % QPB change, while sky B stays +0.129…+0.228 → "the QPB anchor systematic ≈ 0" is too strong; (f) the 2T arm died only on a parameter bound (kT2 pinned at the pre-registered 5 keV cap) with 3 of its 4 acceptance criteria met; (g) the inner R2-4 one-temperature fit is itself formally rejected (p = 0.010 sky B / 0.019 sky A) → ΔkT's interval may be too narrow; (h) E049's breached-control list omits one arm (sky A's sky-norm ×1.2, S: −0.114 keV). → the verdict is unchanged, and all of this is in E049's wording and limits list (R045/D85, triaged). Other red teams: RT-33/RT-34/RT-35 → R038/D78; RT-36 (the P19/P20 designs) died without a report; RT-37 (Q-B vs the literature, 7 material) and RT-38 (the P19/P21 designs, 7 objections) were triaged at 10:14Z → R044/D84; RT-40 (the P22 bundle, 16 objections → R048/D88), RT-41 (the folded-map statistic → R047/D87) and RT-42 (the P23 power projection → R048/D88) are all triaged. As of 2026-10-18 04:43Z: RT-01…RT-60, i.e. 55 teams triaged (RT-36 died without a report; RT-58/RT-59/RT-60 are running), of which RT-46 → R055/D95 (offsets-as-sky withdrawn), RT-47 → R056/D96, RT-48 and RT-50 → R059/D99 (P19's final report), RT-49 → R058/D98 (the Q-B options and the sky normalisation); RT-51 (attacking RT-48 O1b's interpretation) → R061/D100, RT-52 (the consistency audit of the project records, 11 objections) → R060, RT-53 (a design audit of P26, which led to P26 being withdrawn) → R061/R062/D100/D101, RT-54 (option-b feasibility) → R062/D101, RT-55 (the Q-D kT-map claims) → R063/D102, and RT-56 (this handbook vs cognition-os) and RT-57 (D84c) → R064/D103 are all triaged. See "Research progress" on this page and state §5 |
阶段 5 = D55（E049）；D80/D82/D95/D98/D104/D105；R040/R042/R055/R058/R065/R066；P26 已撤回；P27/P28/P29 已派出stage 5 = D55 (E049); D80/D82/D95/D98/D104/D105; R040/R042/R055/R058/R065/R066; P26 withdrawn; P27/P28/P29 dispatched | AMBIGUOUS（功率不足）AMBIGUOUS (underpowered) |
| v18 的 as-run 能段（0.375–9.0 keV）与损坏的 RASS 过滤器对已存 kT 的影响（B025） | The effect of v18's as-run band (0.375–9.0 keV) and the broken RASS filter on the saved kT (B025) | 所有引用 v18 温度的数字都建立在这个合同上；合同此前被误述为 0.4–3.2 keV，而 3.2–9 keV 的影响从未测量 | Every number quoting a v18 temperature rests on this contract; the contract was mis-stated as 0.4–3.2 keV and the effect of the 3.2–9 keV range was never measured | as run：源 0.375–9.0 keV、背景 0.375–6.23 keV（Sherpa notice/ignore 取并集，RT-18 从保存会话重放证实）；SrcPo 不在拟合表达式里；RASS 过滤器损坏（21.9% 的 noticed 信号落在原定 0.2 keV 以下，未重放）；16/65 个已存 kT 的 conf() 区间越出参数界 → 无效 | As run: source 0.375–9.0 keV, background 0.375–6.23 keV (Sherpa's notice/ignore take a union; RT-18 confirmed this by replaying the saved session); SrcPo is absent from the fitted expression; the RASS filter is corrupt (21.9 % of its noticed signal lies below the intended 0.2 keV, not yet replayed); 16/65 saved kT conf() intervals extend outside the parameter bounds → invalid | P16：恢复 7 个会话并引用其 noticed 能段；重拟合 (a) as restored 复现已存 kT（±0.005 keV 内）、(b) 0.4–3.2 keV、(c) RASS 修复、(d) 两者都改，给出 keV 与 σ 的 ΔkT 表；初步结果：R2-4 A1 从 0.4–3.2 keV 改回 as-run 使 kT 变化 −0.021 keV，且已存会话不在其最小值（重拟合 −0.018 keV，stat 72.8 → 42.4） | P16: restore 7 sessions with their noticed ranges quoted; refits (a) as restored reproduces the saved kT within 0.005 keV, (b) 0.4–3.2 keV, (c) RASS fixed, (d) both, with a ΔkT table in keV and σ; partial result: for R2-4 A1, going from 0.4–3.2 keV back to as-run changes kT by −0.021 keV, and the saved session is not at its minimum (the refit gives −0.018 keV, stat 72.8 → 42.4) | P16（luna）；B025；D65e；RT-18P16 (luna); B025; D65e; RT-18 | 待定 PendingPending |
| Chandra 匹配扇区重拟合（P19，已完成——2026-10-17 21:33Z 分诊，E053/T177/R057/R059/D97/D99，执行器跑了 18.6 h） | A Chandra matched-sector re-fit (P19, DONE — triaged 2026-10-17 21:33Z, E053/T177/R057/R059/D97/D99, the executor ran 18.6 h) | Q-B 的 Chandra 半边：P19 之前既无匹配孔径也无功率——2–10′ 的“环”只有 35–62% 的部分环带，共同 kT 检验探不到 ΔkT ≲ 0.27 keV，逐 ObsID 的空白天 α 相差 10% 未传播，拟合能段也与 XMM as-run 不同（R028/D69） | The Chandra half of Q-B: before P19 it had neither a matched aperture nor the power — the 2–10′ "rings" are only 35–62 % partial annuli, the common-kT test cannot detect a ΔkT below ≈ 0.27 keV, the two ObsIDs' blank-sky α differ by 10 % unpropagated, and the fit band differs from XMM's as-run band (R028/D69) | P19 终报（codexnim，18.6 h，2026-10-17 21:33Z 分诊；E053/T177/R057/R059/D97/D99）：Chandra-only kT = 0.690 / 0.755 / 0.747 / 0.858 keV（R2-4…R8-10，0.5–3.0 keV、逐 ObsID 自由 α）；kT(8–10′) − kT(2–4′) = +0.168 keV，只有 1.26σ → 判定 NO POWER（探不到），对“上升”这个问题既不证实也不否证（Provisional）。孔径是部分的：四个环只覆盖名义环带的 35–62%（P19 表 D1 重放：38.96 / 56.8–62.3 / 23.8–49.7 / 35.4–35.5%，8–10′ 两个 ObsID 合起来只有 4.1%）——报告里“覆盖 79–100%”是 XMM/Chandra 的面积比，不是覆盖率，所以 D69 的“匹配孔径”异议仍然成立。最小可探测 ΔkT = 0.30–0.33 keV（稳健 2×1.4826·MAD = 0.3025 keV；未截断重跑 0.3331 keV）——1.0762 keV 是形状网格 0.50–1.00 keV 截断造出来的假象，永不引用。P19 的 XMM-only kT（0.20–0.32 keV）已退役：它是背景构造假象——用 P19 自己的 PHA/ARF/RMF 与似然，只要把 LHB+MWhalo+CXB+SP 加进背景，R2-4 的 kT 就从 0.316 移到 0.710 keV（同一批数据，Δstat 411.95 → 366.07，6 个参数），而阶段 5 在同一批 MOS2 数据上给 0.75 / 0.94 keV → 那不是一个温度；其解释已被红队 RT-51 攻击并分诊（2026-10-18 01:02Z，R061/D100）：热的解确实被优先选择，但“模型 B = 0.71 keV”不是干净数字（α 比 9–12 keV 锚点高 5.07σ、MWhalo kT 贴在 0.4 的壁上），与阶段 5 的“一致”也不独立；Chandra-only 仍是 Provisional/Questioned。设计阶段的异议（已分诊 → R043/R044/D83/D84）：(a) “both”（9532 与 9533 足迹的交集）臂在 8–10′ 只占名义环带的 4.1%（4.687 / 113.10 arcmin²，12 个方位 bin 里只有 4 个有事件）、6–8′ 占 23.8%；(b) 执行器重建 QPB 用的 14 个 ff_ccd*.pi 全部计数为零（把 Chandra 探测器坐标的多边形 AND 进了 XMM 的 evselect 表达式）——执行者自己已发现并丢弃；(c) α 不能被同时设定为“自由参数 + 高斯先验”（继承的拟合器把 α 固定并折进 bg.exposure）。未交付：0.4–6 keV 能段、精确匹配扇区 ARF（arfgen 段错误，改用环平均，R8-10 低 10.2%）、pn0 产品、R4-6 宽标定 68% 区间。 |
P19's final report (codexnim, 18.6 h, triaged 2026-10-17 21:33Z; E053/T177/R057/R059/D97/D99): the Chandra-only kT = 0.690 / 0.755 / 0.747 / 0.858 keV (R2-4…R8-10, 0.5–3.0 keV, per-ObsID free α); kT(8–10′) − kT(2–4′) = +0.168 keV, only 1.26σ → the verdict is NO POWER, which neither corroborates nor refutes the rise question (Provisional). The apertures are partial: the four rings cover only 35–62 % of the nominal annuli (replayed from P19's Table D1: 38.96 / 56.8–62.3 / 23.8–49.7 / 35.4–35.5 %, and 8–10′ across both ObsIDs only 4.1 %) — the report's "covers 79–100 %" is the XMM/Chandra area ratio, not the coverage, so D69's "matched aperture" objection still stands. The minimum detectable ΔkT is 0.30–0.33 keV (robust 2×1.4826·MAD = 0.3025 keV; the unclipped re-run 0.3331 keV) — the 1.0762 keV is an artefact of a 0.50–1.00 keV shape-grid clip and is never quoted. P19's XMM-only kT (0.20–0.32 keV) is retired: it is a background-construction artefact — with P19's own PHA/ARF/RMF and likelihood, adding LHB+MWhalo+CXB+SP to the background moves R2-4's kT from 0.316 to 0.710 keV (same data, Δstat 411.95 → 366.07 for 6 parameters), while stage 5 gives 0.75 / 0.94 keV on the same MOS2 data → it is not a temperature; its interpretation was attacked by red team RT-51 and triaged (2026-10-18 01:02Z, R061/D100): the hot solution is indeed preferred, but "model B = 0.71 keV" is not a clean number (α is 5.07σ above the 9–12 keV anchor and MWhalo kT sits on its 0.4 wall) and its "agreement" with stage 5 is not independent; Chandra-only stays Provisional/Questioned. The design-stage objections (triaged → R043/R044/D83/D84): (a) the "both" (the intersection of the 9532 and 9533 footprints) arm covers only 4.1 % of the nominal 8–10′ annulus (4.687 of 113.10 arcmin², with events in only 4 of 12 position-angle bins) and 23.8 % at 6–8′; (b) all 14 ff_ccd*.pi spectra the executor rebuilt the QPB from have zero counts (the Chandra detector-frame polygons were ANDed into the XMM evselect expression) — the executor had already found and discarded them; (c) α cannot be both "a free parameter with a Gaussian prior" and what the inherited fitter does (it fixes α and folds it into bg.exposure). Not delivered: the 0.4–6 keV band, an exact matched-sector ARF (arfgen segfaulted; ring-averaged instead, 10.2 % low at R8-10), the pn0 products, and R4-6's wide-calibration 68 % interval. |
P19；E053/T177；R057/R059；D97/D99；D69；RT-23/R028；RT-38/R044；RT-48/RT-50/R059；RT-51/R061/D100P19; E053/T177; R057/R059; D97/D99; D69; RT-23/R028; RT-38/R044; RT-48/RT-50/R059; RT-51/R061/D100 | Provisional 暂定（NO POWER；孔径部分）Provisional (NO POWER; partial apertures) | ||
| as-run 能段的 mask 阶梯重拟合（P20，已完成 2026-10-17 06:45Z；E048/T173/R039/D79） | An as-run-band refit of the mask ladder (P20, done 2026-10-17 06:45Z; E048/T173/R039/D79) | D64“保留全套 v2 mask”的判定不变，但支撑它的“不用全套点源 mask 会偏 +0.2 keV”是在 0.4–3.2 keV 上量的，而 v18 as-run notice 0.375–9.0 keV；被 mask 挡掉的主要是 M104 自身的光与恒星，机制未被证明（R030/D70） | D64's verdict "keep the full v2 mask" stands, but the "+0.2 keV bias without the full point mask" that supports it was measured on 0.4–3.2 keV while v18 as run notices 0.375–9.0 keV; what the mask removes is mostly M104's own light and stars, so the mechanism is unproven (R030/D70) | 结果（Provisional 暂定，一个孔径）：在真实拟合能段 0.375–9.0 keV 上，只遮延展源+恒星仍把 R2-4_Angle1 的 kT 推高——ΔkT(M3−M0) = +0.287（AGN 注入）/ +0.279（纯热，源幂律按 v18 冻结在 0），而早期阶梯能段（0.4–3.2 keV）上是 +0.225 / +0.172 → D64（保留全套 v2 点源 mask）在 as-run 能段上被确认并加强。机制仍被质疑（Questioned）：自由源幂律那一臂把偏差降到 +0.031 / +0.022，但该臂 65.5% / 24.5% 的自助抽样压在幂律 norm 上限 1e-5（冻结臂则是 82% / 54% 压在 kT 上界 1.0 keV）→ 自由幂律的读数不可解读；R4-6_Angle2 的热臂在所有臂里都失败（−0.06）。尚无正常/专家对照臂（D79：不阻塞）。 — cognition-os/experiments/E048_*.md |
Result (Provisional, one aperture): on the real fitting band 0.375–9.0 keV, masking only extended sources + stars still biases R2-4_Angle1's kT upward — ΔkT(M3−M0) = +0.287 (AGN-injected) / +0.279 (thermal-only, with the source power law frozen at 0 as v18 does), against +0.225 / +0.172 on the earlier ladder band (0.4–3.2 keV) → D64 (keep the full v2 point mask) is CONFIRMED and strengthened on the as-run band. The mechanism stays Questioned: the free-source-power-law arm reduces the bias to +0.031 / +0.022, but in that arm 65.5 % / 24.5 % of the bootstrap draws sit at the power-law norm ceiling 1e-5 (the frozen arm has 82 % / 54 % at the kT upper bound 1.0 keV) → the free-power-law reading is not interpretable; R4-6_Angle2's thermal arm fails in every arm (−0.06). There is still no normal/expert control arm (D79: not blocking). — cognition-os/experiments/E048_*.md |
P20；E048/T173；R039；D79P20; E048/T173; R039; D79 | Provisional 暂定Provisional | ||
| 钉住 v18 世代后重跑 Angle1（P21，已完成 2026-10-17 14:34Z；E050/T175/R049/D89，Provisional） | Re-run Angle1 once one v18 generation is pinned (P21, done 2026-10-17 14:34Z; E050/T175/R049/D89, Provisional) | Q-C 的图/v18 比值分支：P21 之前不可检验（R7-9 Angle1 没有 .fitting 会话、R9-11 Angle1 有三代且 CSV 比可重放会话高 58%、R11-13/R12-14 不显著；可复现的 Angle2 是平的 1.26 → 1.18，R033/D73） | Q-C's image/v18 ratio branch: before P21 it was untestable (R7-9 Angle1 has no .fitting session, R9-11 Angle1 exists in three generations with the CSV 58 % above the replayable session, and R11-13/R12-14 are non-significant; the reproducible Angle2 is flat at 1.26 → 1.18, R033/D73) | 已完成（2026-10-17 14:34Z 分诊，R049/D89，Provisional）：先在每个区域钉住一个 v18 世代（sha 清单），再在钉住的合同上重跑 Angle1 R7-9…R12-14，并把 Angle2 与 Angle1 并列报告（D73）。image/v18 比值没有任何 ≥2σ 的径向上升——ratio_deabsorbed R7-9 0.929、R9-11 1.477、R11-13 1.369，计数空间闭合（目标 1）1.018 ± 1.185 / 1.639 ± 1.406 / 3.617 ± 5.198，R7-9→R9-11 的台阶只有 0.34σ（每一步都 <0.5σ）；R12-14 未定义（image_SB −1.09e-16，非探测）；归档的 3.32 来自一个不可重放的世代。分派方重放 MANIFEST 202/202 OK。这是"没有超出证据"，不是"比值 = 1"——误差棒和信号一样大。后续红队（RT-44 → R051/D91；P24 → R052/D92；P25 → R053/D93）：(i) 构造被质疑（Questioned）——R7-9 的两台相机逐相机闭合 1.311（mos1）vs 0.737（mos2），公布的 1.018 只是两个散差的简单平均，因此"闭合 1.018 证明天空扣除正确"这条论证已撤回；P24 重放显示该差别 0.574 ± 0.511 = 1.1σ，是泊松散差被净/总计数比（净只有总的 13–21%）放大，不是校准分歧（"相机不一致"的说法也已撤回）；(ii) 绝对 image/v18 比值仍未定——天空模型 vs 经验天空把 R7-9 的比值从 0.93 移到 1.77，因此不引任何绝对比值（D89a/D92a）；阶段 2 的 1.878 与本任务的 0.834 相差 ≈2.25× 未调和；(iii) 判定"没有 ≥2σ 径向上升"保持 Provisional（一台相机、构造被质疑；按逆方差加权的台阶 +0.9σ / +0.8σ）。遗留 caveat：Angle1 的 R2-4…R6-8 与 Angle2 对照没有拟合；R9-11 的 kT 对起点敏感且被截顶，R11-13 双峰（比值 1.369…3.43）；B029：P21 的脚本给 mos2 用了 mos1 的线集（0.7–1.0 keV 上影响可忽略，见 §5 P25 行）。历史记录（红队 RT-38 的设计异议，已分诊 → R043/R044/D83/D84；这些缺陷在 P21 的最终交付前已被执行者修正）：所谓“钉住的世代”一度没有 manifest 文件、4 个目标区域缺 _params.txt 等文本产品（FAST_MODE=1 提前 sys.exit(0)）；第一份比值表的 image_SB 全是负的（单位错误，B028）→ 该表已废弃，本手册不引它的数字。 — cognition-os/experiments/E050_p21_pinned_v18_angle1_image_ratio.md；07_REVISION_LOG.md R049/R051/R052/R053；04_DECISION_LOG.md D89/D91/D92/D93 |
Done (triaged 2026-10-17 14:34Z, R049/D89, Provisional): first pin one v18 generation per region (a sha manifest), then re-run Angle1 R7-9…R12-14 on the pinned contract and report Angle2 next to Angle1 (D73). The image/v18 ratio shows no ≥2σ rise with radius — ratio_deabsorbed R7-9 0.929, R9-11 1.477, R11-13 1.369; the count-space closure (target 1) is 1.018 ± 1.185 / 1.639 ± 1.406 / 3.617 ± 5.198, and the R7-9→R9-11 step is only 0.34σ (every step <0.5σ); R12-14 is undefined (image_SB −1.09e-16, a non-detection); the archived 3.32 came from an unreplayable generation. The dispatcher replayed MANIFEST 202/202 OK. This is "no evidence for an excess", not "the ratio = 1" — the error bars are as large as the signal. Follow-up red teams (RT-44 → R051/D91; P24 → R052/D92; P25 → R053/D93): (i) the construction is Questioned — at R7-9 the two cameras' per-camera closures are 1.311 (mos1) vs 0.737 (mos2), and the published 1.018 is only a plain mean of two scatters, so the argument "closure 1.018 confirms a single, correct sky subtraction" is withdrawn; P24's replay shows the difference is 0.574 ± 0.511 = 1.1σ, i.e. Poisson scatter amplified by the net/gross ratio (net is only 13–21 % of gross), not a calibration disagreement (the "cameras disagree" wording is also withdrawn); (ii) the absolute image/v18 ratio is still undetermined — sky model vs empirical sky moves the R7-9 ratio from 0.93 to 1.77, so no absolute ratio is quoted (D89a/D92a); stage 2's 1.878 vs this job's 0.834 differ by ≈2.25×, unreconciled; (iii) the verdict "no ≥2σ radial rise" stays Provisional (one method, construction Questioned; inverse-variance-weighted steps +0.9σ / +0.8σ). Open caveats: Angle1's R2-4…R6-8 and the Angle2 controls were not fitted; R9-11's kT is start-point sensitive and cap-truncated, R11-13 is bimodal (ratio 1.369…3.43); B029: P21's script used the mos1 line set for mos2 (negligible at 0.7–1.0 keV; see the P25 §5 row). Historical note (red team RT-38's design objections, triaged → R043/R044/D83/D84; these defects were fixed by the executor before P21's final delivery): the "pinned generation" at one point had no manifest and all 4 target regions lacked _params.txt-type text products (FAST_MODE=1 sys.exit(0)s early); the first ratio table had negative image_SB in all four regions (a unit error, B028) → that table is Deprecated and this handbook quotes none of its numbers. — cognition-os/experiments/E050_p21_pinned_v18_angle1_image_ratio.md; 07_REVISION_LOG.md R049/R051/R052/R053; 04_DECISION_LOG.md D89/D91/D92/D93 |
P21；E050/T175；R049/R051/R052/R053；D89/D91/D92/D93；RT-44；P24/P25P21; E050/T175; R049/R051/R052/R053; D89/D91/D92/D93; RT-44; P24/P25 | Provisional 暂定（构造 Questioned）Provisional (construction Questioned) | ||
| 给 Q-B 增加检验功率（提案 P23（P26 已撤回，R061/D100）；PI 已于 2026-10-18 05:22Z 答复（D105）：重新抽谱长期授权，选项 (b) 不再需要批准 → 已化为 P27/P28/P29 派出） | Add power to the Q-B test (proposed P23; P26 withdrawn, R061/D100; the PI answered on 2026-10-18 05:22Z (D105): re-extraction is standing-authorised and option (b) no longer needs approval → it became the dispatched P27/P28/P29) | 阶段 5 的判定是 AMBIGUOUS，而且功率不足：按预先声明的非参数定义，两种方法的最小可探测 ΔkT 都 > 0.50 keV，而 0.10 keV 的门低了一个量级；用现有数据无法裁决 0.1 keV 量级的问题（E049/T174/R040/D80，R042/D82） | Stage 5's verdict is AMBIGUOUS and underpowered: under the pre-declared non-parametric definition both methods' minimum detectable ΔkT exceeds 0.50 keV, an order of magnitude above the 0.10 keV gate; the existing data cannot decide a 0.1 keV-scale question (E049/T174/R040/D80, R042/D82) | P23 的形状（D95，2026-10-17 20:53Z 重排；Provisional 暂定，统计投影）：“先用偏置指向 0501/0601 固定天空”已撤回（R055/D95，RT-46 终审 6 条实质异议）——v18 把 BKG_OBSID = '0900170701' 硬编码在源码里，每个环的天空/背景参考本来就是 0701 空白天空场，不是偏置指向：MOS2 Fe-L 总计数（PI 140-174/175-209）在 0701 是 3634/3358（89.0 ks），而 0501 R15-20 只有 169/125（85.8 ks）——最好的偏置环比已在用的场还弱 ≥3×，R15-20 弱 21×（分派方重放）。偏置指向另有四个问题：距中心历元 187/193 天（SWCX）、mask 是另一种类型（2640 椭圆 vs 5616 圆）、8-10′ 无覆盖、35-40′/40-45′ 环带携带 +11σ/+3.9σ 的正晕面亮度（不是天空）；且天空项只占 S 区间宽度的 ~28%（点移 0.069 keV），而 Fe-L 泊松项是 0.22 keV、门对照散差 0.11–0.20 keV → “偏置优先”不是决定性的一步。PI 已于 2026-10-18 05:22Z 答复（D105）：重新抽谱长期授权，选项 (b) 不再需要批准，已化为 P27/P28/P29 派出（原四个选项 2026-10-17 22:26Z 交出，D98）：(a) 接受 Q-B 判 AMBIGUOUS——现有数据无法在 ≈0.3–0.5 keV 以下裁决；(b) 加 MOS1+pn0 或放宽到 8-12′——信息增益 1.44（总计数）/ 1.91（净计数）（RT-49 重放：1.198² = 1.436、0.0944/0.0683 的平方 = 1.910，即 D95 的 1.44/1.91；不是 Opus 的 2.66），最小可探测 ΔkT 0.31–0.39 keV（两种噪声基线：1.96·σ 投影 0.31–0.36 keV、RT-49 自己的直接噪声模型 0.366–0.393 keV；Provisional 暂定）——到不了 0.10 keV 的门，也过不了 0.11–0.20 keV 的门对照散差；RT-54 已分诊（2026-10-18 01:56Z，R062/D101）：选项 (b) 不便宜——v2 合同上没有任何 MOS1/pn0/8–12′ 产品（分派方重放 apertures 列表与 s5lib.py 的文件模式全部是 mos2U005-*，即阶段 5 执行器 MOS2-hardcoded），据估算约 40–45 个 Virgo wall-h（report-based，按估算引用），且需一份新的预注册；分派方建议默认选 (a)；
(c) 已撤回：P26 天空归一研究（R061/D100）——sky_background_scale 同时乘在源 APEC 上，天空归一的 ×0.8/×1.2 对照（D55 的 b: 族，只乘天空成分）早已跑过：天空 B 上 ΔkT 的移动 ≤0.052 keV，天空 A 上 S4 ×1.2 为 −0.114 keV，已包含在 0.11–0.20 keV 的门对照散差里；(d) 增加曝光：不可得。阶段 5 的判定保持 AMBIGUOUS。0701 自身是否含延展发射未检验（RT-49 的 OBJ-2：0701 距 M104 88.85 角分、在晕外，有自身 −11% 的内部梯度，未重放）。最便宜的决定性天空检验是对另一个孔径做留出天空拟合（RT-46 OBJ-4）。 — 07_REVISION_LOG.md R055、R058；04_DECISION_LOG.md D95、D98；cognition-os/experiments/E049_stage5_D55_matched_contract.md |
P23's shape (D95, re-ordered 2026-10-17 20:53Z; Provisional, a statistical projection): "first fix the sky from the offset pointings 0501/0601" is WITHDRAWN (R055/D95, 6 material objections in RT-46's final report) — v18 hard-codes BKG_OBSID = '0900170701' in its source, so the sky/background reference of every annulus is already the blank-sky field 0701, not the offset pointings: MOS2 Fe-L gross counts (PI 140-174/175-209) are 3634/3358 in 0701 (89.0 ks), against only 169/125 in 0501 R15-20 (85.8 ks) — the best offset annulus is ≥3× weaker than the field already in use and R15-20 is 21× weaker (dispatcher replay). The offsets have four further problems: they are 187/193 days from the central epoch (SWCX), their mask is another kind (2640 ellipses vs 5616 circles), they have no coverage at 8–10′, and the 35–40′/40–45′ annuli carry positive halo surface brightness of +11σ/+3.9σ (not sky); and the sky term is only ~28 % of the S-interval width (point shift 0.069 keV), while the Fe-L Poisson term is 0.22 keV and the gate-control spread is 0.11–0.20 keV → "offsets first" is not the decisive step. The PI answered on 2026-10-18 05:22Z (D105): re-extraction is standing-authorised and option (b) no longer needs approval → it became the dispatched P27/P28/P29 (the four options went to the PI at 2026-10-17 22:26Z, D98): (a) accept AMBIGUOUS for Q-B — the existing data cannot decide below ≈0.3–0.5 keV; (b) add MOS1+pn0 or widen to 8–12′ — information gain 1.44 (gross counts) / 1.91 (net counts) (RT-49's replay: 1.198² = 1.436 and the square of 0.0944/0.0683 = 1.910, i.e. D95's 1.44/1.91; not Opus's 2.66), a minimum detectable ΔkT of 0.31–0.39 keV (two noise baselines: the 1.96·σ projection 0.31–0.36 keV and RT-49's own direct noise model 0.366–0.393 keV; Provisional) — it cannot reach the 0.10 keV gate, nor pass it against the 0.11–0.20 keV gate-control spread; RT-54 was triaged (2026-10-18 01:56Z, R062/D101): option (b) is NOT cheap — no MOS1/pn0/8–12′ products exist on the v2 contract (the dispatcher replayed the apertures lists and s5lib.py's file patterns, all of which are mos2U005-*, i.e. the stage-5 executor is MOS2-hardcoded), it costs an estimated 40–45 Virgo wall-hours (report-based, quoted as an estimate), and it needs a new pre-registration; the dispatcher recommends (a) by default;
(c) WITHDRAWN: the P26 sky-normalisation study (R061/D100) — sky_background_scale also multiplies the source APEC, and the sky-normalisation ×0.8/×1.2 controls required by D55 (the b: family, which scale the sky components only) have already run: on sky B ΔkT moves by ≤0.052 keV, on sky A the S4 ×1.2 control moves it by −0.114 keV, all inside the 0.11–0.20 keV gate-control spread; (d) more exposure: not available. Stage 5's verdict stays AMBIGUOUS. Whether 0701 itself contains extended emission is untested (RT-49's OBJ-2: 0701 is 88.85 arcmin from M104, outside the halo, with its own −11 % internal gradient; not replayed). The cheapest decisive sky check is a hold-out sky fit on another aperture (RT-46 OBJ-4). — 07_REVISION_LOG.md R055, R058; 04_DECISION_LOG.md D95, D98; cognition-os/experiments/E049_stage5_D55_matched_contract.md |
P23（建议，未经评审；P26 已撤回）；D82d/D95/D98/D105/D106/D107；R055/R058/R066；E049/T174P23 (suggestion, not reviewed; P26 withdrawn); D82d/D95/D98/D105/D106/D107; R055/R058/R066; E049/T174 | PI 已答复（D105）→ P27/P28/P29 已派出The PI has answered (D105) → P27/P28/P29 dispatched | ||
| Q-C：外环谱的拟合方法；15–30′ 对比（R′）不可测；整体改标 Questioned（相机相关） | Q-C: the outer-ring spectral fit method; the 15–30′ comparison (R′) not testable; re-labelled Questioned as a whole (camera-dependent) | 13′ 假象已定性（I065），但外环（10–14′，FoV 边缘 ~50% 被裁）的谱点仍不可用：kT 上限撞 1 keV 参数顶、源/背景简并 | The 13′ artefact is characterised (I065), but the outer-ring (10–14′, ~50% clipped at the FoV edge) spectral points remain unusable: kT upper limits hit the 1 keV ceiling, source/background degenerate | 归一错误 + 弱信号简并（I065，Provisional）。2026-10-18 04:33Z 整体改标 Questioned（相机相关、未解决，R065/D104b）：双相机 IVW 均值无 ≥2σ 上升，但两相机在 R11-13 相差 2.35σ（mos1 闭合 8.53 ± 3.12 vs mos2 0.74 ± 1.13）、mos1 单独 +2.33σ → “13′ 超出就是 B020 假象”这句旧措辞按原样撤回。图内检验第三次确认无 13′ 超出（E041：R11–13/R9–11 = 0.949 ± 0.029 / 1.029 ± 0.037）；图/v18 比值分支不可检验（RT-28/R033/D73，见本表 P21 行） | Normalisation error + weak-signal degeneracy (I065, Provisional). 2026-10-18 04:33Z: re-labelled Questioned as a whole (camera-dependent, unresolved, R065/D104b): the two-camera IVW mean shows no ≥2σ rise, but the cameras disagree by 2.35σ at R11-13 (mos1 closure 8.53 ± 3.12 vs mos2 0.74 ± 1.13) and mos1 alone rises +2.33σ → the old wording "the 13′ excess is the B020 artefact" is withdrawn as worded. The image-internal test confirms no 13′ excess for the third time (E041: R11–13/R9–11 = 0.949 ± 0.029 / 1.029 ± 0.037); the image/v18 ratio branch is untestable (RT-28/R033/D73, see this table's P21 row) | 更宽 kT 范围 + profile-likelihood 误差，或绑定/正则化径向模型，或用偏置指向（0900170501/0601） | Wider kT range + profile-likelihood errors, or a tied/regularised radial model, or the offset pointings (0900170501/0601) | Q-C 后续（并入 Virgo 再抽谱 D47）；R065/D104bQ-C follow-up (folded into the Virgo re-reduction D47); R065/D104b | Questioned（相机相关）Questioned (camera-dependent) |
| Q-D：2D 温度图（等 Q-A） | Q-D: the 2D temperature map (waits for Q-A) | feedback 与引力束缚气体的区分需要 2D 温度 | Separating feedback from bound gas needs a 2D temperature field | 按 D49 探索：窄带比值代理（Fe-L 硬度等）+ 自适应网格（WVT/共享核）+ 象限锚定；验收 = 形态保真度（均匀真值不出假结构、注入结构原位恢复、≥2 相机/方案一致），整体可控偏差可接受；谱拟合 3–4 个关键 bin 作校验 | Explore per D49: narrow-band ratio proxies (e.g. Fe-L hardness) + adaptive grids (WVT/shared kernel) + quadrant anchors; acceptance = morphology fidelity (uniform truth → no spurious structure, injected structure recovered in place, consistent across ≥2 cameras/schemes), an overall controlled bias acceptable; spectral fits of 3–4 key bins as checks | 亮度图已完成（阶段 2：v2-mask MOS2、QPB+天空扣除、β 残差与扇区对比，E042）；2D kT 阶段 3 已完成（E043：估计器 E3 + cone45，双臂一致，共同足迹 56%）；生产图（P12/E045）首批已出（读数：2–5′ 无小轴 vs 盘对比、内区 N−S +0.10 keV 提示，仅一种方法；东箭头方向缺陷修正中，修好前不展示该图）→ 待定；区域设计建议 cone45 bin；全套 v2 mask 保留（R020/D64）。当前最佳数字（折叠图，E046/T171，Provisional 暂定；措辞按 R063/D102 收窄）：四象限折叠的 2D kT 图 v2 给出 2–12′ 的逐箱 kT 中位带 0.70–0.80 keV（逐箱范围 0.53–0.89 keV，逐箱中位 0.748 keV（0.736 是池化全局值），8/21 箱在带外，中位 σ68 0.059 keV 对 0.10 keV 带宽——即只在逐箱功率之内近等温；r≥8′ 的 6 个箱里有 4 个在 95% 上分不开 0.6 与 1.0 keV；措词按 RT-20/D68/R025 与 RT-55/R063/D102 收窄）、全角度斜率 +0.009 keV/角分（95 % [−0.003, +0.020]，+1.5σ）→ 对这种单温代理而言无 ≥ 2σ 的内冷外热；每 5′ 升 0.1 keV 的梯度在 E046 的单温代理注入里 91% 的次数能被找回（仅这一个温代理的自检；与真实逐楔 2–8′ 检验对应的统计量在 E047 端到端模拟里功率只有 27–36%，所以这 91% 不是探测功率）；盘楔外箱 0.85–0.88 vs 小轴楔 0.61–0.77 keV 只是提示（≈ 1.4σ）；MOS2 与 pn0 逐箱一致（p = 0.60）；pn0 QPB 缩放（测试用）会把斜率推到 ≈ 2σ → 趋势对背景敏感；< 2′ 核被 mask 遮住，冷却流不可检验（D66/P18 待定）；E045 的内区 N−S 提示降为 Questioned（R022） | The intensity map is done (stage 2: v2-masked MOS2, QPB + sky subtracted, β residuals and sector contrasts, E042); 2D kT stage 3 is done (E043: estimator E3 + cone45, both arms agree, common footprint 56 %); the first production maps (P12/E045) are out (readings: no minor-axis vs disk contrast at 2–5′, an inner N−S +0.10 keV hint, one method only; the East-arrow direction defect is being fixed and the map is not shown until it is) → pending; the region design recommends cone45 bins; the full v2 mask stays (R020/D64). Current best number (the folded map, E046/T171, Provisional; wording narrowed per R063/D102): the four-quadrant-folded 2D kT map v2 gives a median per-bin kT band of 0.70–0.80 keV over 2–12′ (per-bin range 0.53–0.89 keV; per-bin median 0.748 keV (0.736 is the pooled global value); 8/21 bins outside the band; median σ68 0.059 keV against the 0.10 keV band — i.e. near-isothermal only within per-bin power; 4 of the 6 bins at r≥8′ cannot separate 0.6 from 1.0 keV at 95 %; wording narrowed per RT-20/D68/R025 and RT-55/R063/D102) and an all-angle slope of +0.009 keV/arcmin (95 % [−0.003, +0.020], +1.5σ) → no inner-cool/outer-hot trend at ≥ 2σ for this one-temperature proxy; a gradient of +0.1 keV per 5′ is recovered 91 % of the time in E046's own one-temperature-proxy injection (that proxy's self-test only; the statistic matching the real per-wedge 2–8′ test has a power of only 27–36 % in the E047 end-to-end mock, so this 91 % is not a detection power); the disk wedge's outer cells 0.85–0.88 vs the minor wedge 0.61–0.77 keV are suggestive only (≈ 1.4σ); MOS2 and pn0 agree bin by bin (p = 0.60); the pn0 QPB scaling (test-only) pushes the slope to ≈ 2σ → the trend is background-sensitive; the core inside 2′ is masked, so a cooling flow is untestable here (D66/P18 pending); E045's inner N−S hint is downgraded to Questioned (R022) | E038/E042/E043/E046；P12 生产图；P15 折叠图 v2E038/E042/E043/E046; P12 production maps; P15 folded map v2 | 待定 PendingPending |
| MOS1 静态（非耀发）硬过剩：每个指向都偏高 | MOS1 quiescent (non-flare) hard excess: high in every pointing | MOS1 参与所有 v18 拟合；若其背景模型欠 5–20%，会偏置 kT | MOS1 enters every v18 fit; a 5–20% background underestimate would bias kT | 静态 SP 成分或 MOS1 参考系统误差（成因 Questioned） | A quiescent SP component or MOS1 reference systematics (cause Questioned) | Q-A 阶段 2：MOS1-free vs MOS1-with-extra-SP 拟合对比（I066） | Q-A stage 2: MOS1-free vs MOS1-with-extra-SP fit tests (I066) | I066/T151；Q-A 后续I066/T151; Q-A follow-up | ProvisionalProvisional |
| 点源/星系团遮蔽无法把残余压到 <5% | Point-source/cluster masking cannot push residuals below 5% | 弱晕区（R11-13 Angle2 残余/晕比当前 3.6）做任何温度测量前必须先处理残余 | Before any temperature measurement in the faint-halo region (residual/halo currently 3.6 in R11-13 Angle2), the residual must be handled | PSF 翼与星系团发射遮不掉（E035；mask 规则 v2 后残余/晕比仍 ≈0.37，相对 v18 局部晕面亮度——0.95 只是相对 v2 晕律，该晕律比 v18 暗 2.59×，RT-05 修订 R008/RT-32 的 OBJ-7） | PSF wings and cluster emission are not maskable (E035; after mask rule v2 the residual/halo ratio is still ≈0.37 relative to the v18 local halo surface brightness — 0.95 holds only relative to the v2 halo law, which is 2.59× fainter than v18; RT-05 amendment R008 / RT-32's OBJ-7) | 建模残余：星系团作为 APEC 成分 + 空间加权残余模型；固定 CXB 归一须与 mask 流量阈值匹配（E035） | Model the residual: the cluster as an APEC component + a spatially weighted residual model; the fixed CXB normalisation must match the mask flux cut (E035) | E035；Virgo 再抽谱 D47E035; the Virgo re-reduction D47 | ProvisionalProvisional |
| Virgo 再抽谱计划（D47） | The Virgo re-reduction plan (D47) | B020/I065/E035 的修正（正确 SB、mask v2、建模残余）都要落到新的提取产品上 | The corrections from B020/I065/E035 (correct SB, mask v2, modelled residual) must land in new extraction products | 以 v1.7.8 基线复现为对照，每阶段只改一个设计元素 | Reproduce the v1.7.8 baseline as the control, then change one design element per stage | S0 基线复现已 PASS（事件级复现，E040）；阶段 1 完成但口径谱只有自动 mask（B023）→ 阶段 1b 重抽 v2-mask 口径谱：已于 2026-10-17 02:45Z 交付、门 2 通过（D76c：9/9 口径、MANIFEST 506/506 OK、BACKSCAL 真实、REG 身份 PASS），但交付的 QPB 大 1e6–1e7 倍（B026/R037/D77）→ mask/口径/响应被接受、QPB 不被接受。P22 的 QPB 重生成已于 2026-10-17 11:04Z 被接受（R045/D85）：9/9 个口径通过量级门（R = 0.884–1.005，门 0.8–1.3），MANIFEST 776/776 项 OK；2 个口径与 Sonnet 的推导修复相差 >3%（R4-6_Angle1 +5.56%、R6-8_Angle2 +3.04% —— 锚点分散，Provisional）。阶段 5b（在重新生成的 QPB 上重跑）已于 12:48Z 分诊（R046/D86）：判定 AMBIGUOUS 不变，天空 B 的 ΔkT +0.1886 → +0.1982（−0.034…+0.478），QPB 构造系统误差 0.010 keV；derived-QPB 标签对天空 B 的 S4/H 已退役（D86），但 RT-39 O10 的 QPB 尺度扫描（×0.80–1.40）显示天空 A 在 ×0.85 翻号、天空 B 保持 +0.129…+0.228（±0.05 keV）→ 该项仍未量化。qpbfix bundle 按 D88d 已设为只读。Bundle 完整性教训（B027）：交付 bundle 里的一个 events 文件曾被发现被清空并已恢复（04:42:38Z）；阶段 5 不受影响（它用到的事件派生项在 03:59–04:01Z 就已完成）。按 D81，P22 重新交付后该 bundle 会被 chmod -R a-w |
The s0 baseline reproduction PASSed (event-level reproduction, E040); stage 1 finished but its aperture spectra carry only the automatic mask (B023) → stage 1b re-extracts the v2-masked aperture spectra: delivered on 2026-10-17 02:45Z with gate 2 PASSED (D76c: 9/9 apertures, MANIFEST 506/506 OK, real BACKSCAL, REG identity PASS), but the delivered QPB is 1e6–1e7× too large (B026/R037/D77) → the masks/apertures/responses are accepted, the QPB is not. P22's QPB regeneration was accepted on 2026-10-17 11:04Z (R045/D85): 9 of 9 apertures pass the magnitude gate (R = 0.884–1.005, gate 0.8–1.3) and MANIFEST is 776/776 entries OK; 2 apertures differ from Sonnet's derived repair by more than 3 % (R4-6_Angle1 +5.56 %, R6-8_Angle2 +3.04 % — anchor spread, Provisional). Stage 5b (the re-run on the regenerated QPB) was triaged at 12:48Z (R046/D86): the AMBIGUOUS verdict is unchanged, sky B's ΔkT +0.1886 → +0.1982 (−0.034…+0.478), and the QPB-construction systematic is 0.010 keV; the derived-QPB label is retired for sky B's S4/H (D86), but RT-39 O10's QPB scale sweep (×0.80–1.40) shows sky A flipping sign at ×0.85 while sky B stays at +0.129…+0.228 (±0.05 keV) → that item remains unquantified. The qpbfix bundle is read-only per D88d. Bundle-integrity lesson (B027): one events file in the delivered bundle was found emptied and restored (04:42:38Z); stage 5 is unaffected (its events-derived items were done at 03:59–04:01Z). Per D81 the bundle gets chmod -R a-w once P22 re-delivers it |
job virgo-rereduction-s0；D47/D57/D77/D81/D85/D86/D88；B023/B026/B027；R045/R046/R048job virgo-rereduction-s0; D47/D57/D77/D81/D85/D86/D88; B023/B026/B027; R045/R046/R048 |
Provisional 暂定（QPB 尺度扫描仍开）Provisional (the QPB scale sweep stays open) |
未来 1–4 周计划
Plan for the next 1–4 weeks
计划完全来自 cognition-os/02_CURRENT_QUESTION.md 与各 job 的 BRIEF/REPORT，不是本手册的发明。2026-10-17 23:27Z 更新（阶段 25，刷新 6）：P19（Chandra 匹配扇区重拟合）已收官——Chandra-only kT 0.690/0.755/0.747/0.858 keV，kT(8–10′) − kT(2–4′) = +0.168 keV 只有 1.26σ → 判定 NO POWER（探不到），对上升问题既不证实也不否证；孔径只是名义环带的 35–62%（报告里的“79–100%”是面积比），最小可探测 ΔkT 0.30–0.33 keV（1.0762 keV 是网格截断假象，永不引用）；P19 的 XMM-only kT（0.20–0.32 keV）已退役为背景构造假象（0.316 → 0.710 keV，同一批数据）。Q-B 的决定性检验（D55 预注册的匹配合同测试）已完成，判定 = AMBIGUOUS（含糊）——功率不足，不是"没有上升"（2026-10-17 07:15Z，E049/T174/R040/D80；Provisional 暂定）。谱拟合 S4 的 ΔkT(8–10′−2–4′) = +0.13（天空 A，95% −0.21…+0.42）/ +0.19（天空 B，−0.004…+0.42），Fe-L 硬度 +0.01（−0.32…+0.46）/ +0.07（−0.13…+0.33）；最小可探测 ΔkT 的 0.43 keV 只是高斯近似，按预先声明的非参数定义（50% 功率、95% 区间）两种方法都 > 0.50 keV（R042/D82）。独立专家复核（Opus，D59a）确认 AMBIGUOUS。阶段 5b（在重新生成的 QPB 上重跑）已于 12:48Z 分诊：判定不变（R046/D86）——天空 B 的 ΔkT +0.1886 → +0.1982（−0.034…+0.478），QPB 构造系统误差 0.010 keV；P22 重新生成的 QPB 已于 11:04Z 被接受（9/9 口径过量级门，R 0.884–1.005，R045/D85）。2026-10-18 07:46Z 更新（阶段 27，刷新 7 收官）：PI 已答复（D105，05:22Z）——“重新抽谱一直是可以的”，用任何 region 设计重新抽谱长期授权，只受 Virgo 资源门限制（内存 ≤60%，05:22Z 实测 ≈58%）；选项 (b) 不再“需要 PI 批准”。据此派出三个新任务：P27（D105：MOS1+pn0 抽 R2-4、R4-6、R6-8、R8-10 加 R10-12，阶段 1b 合同，整条 profile 共享扰动参数联合拟合，先写预注册再拟合，排队等 Virgo 内存门）、P28（PI 命令 D106：M104 专用多能段面亮度图，九个能段 0.4–1.25、0.4–0.6、0.6–0.8、0.8–1.0、0.8–1.2、0.4–0.5、0.5–0.6、0.6–0.7、0.7–0.8 keV，扣除 QPB/SP/LHB/MWhalo/CXB，点源适度遮蔽或 PSF 相减，单象限折叠版本，codexnim 运行中）、P29（D55 点名但从未执行的 0.5–1.2 keV 次要 S4 臂 + R11-13_Angle2 通道，codexnim 运行中）。2026-10-18 09:48Z 更新（阶段 28）：P29 已完成（R067/D108，Provisional）——次要臂 ΔkT(R8-10 − R2-4) = +0.077 keV（天空 A，95% [−0.307, +0.479]）/ +0.145 keV（天空 B，95% [−0.142, +0.481]），与主臂（+0.094 / +0.198）同号，但 95% 宽度比 1.00（A）/ 1.22（B）→ 不收窄区间，Q-B 判定保持 AMBIGUOUS；次要臂是否独立佐证由 RT-65 检验中（待定）。P27 不按 RT-54 形式启动（R067/D108）：外环信息预算占 98%、MOS1 通道买不到信息、R10-12/R12-14 撞 FoV/vignetting 墙 → 外环封顶 12′、无 MOS1 通道，改为 pn0 R2-4 + R8-10 预检（≈6–8 个 Virgo wall-h），先写短预注册（D55 逻辑）再拟合；最佳设计只能到 MDΔkT 0.22–0.23 keV，对 0.10 keV 的门（对照 0.12–0.20 keV 在约束）→ 能收窄区间，但不太可能翻转 AMBIGUOUS（RT-64 数字为 report-based）。P28 仍在运行（≈3.5 h），无结果，待定。（阶段 29 订正：P28 的产品已存在，但 make_figures.py 的 extent 上下颠倒（坐标轴 bug），图正在修复；P28 现为"待定 / Pending"，修好前不展示任何 P28 图，D112d。）阶段 5b 是记录版本（D104a）：kT(R8-10) 0.9446 keV、ΔkT +0.198 keV、95% [−0.034, +0.478]（S4）。红队普查（阶段 29，刷新 8）：RT-01…RT-69，65 支已分诊（RT-33 无队列行；RT-36 无报告死亡、由 RT-38 取代；RT-65 → R070/D112、RT-66 → R068/D109、RT-67 → D112 均已分诊；RT-68（P28 管线代码审计）与 RT-69（P30 设计审计）运行中）。此前记录（2026-10-17 23:27Z，阶段 25，刷新 6）的四个选项（a–d，D98）：选项 (b) 不再需要批准，已化为 P27/P29 执行（加 MOS1+pn0 的信息增益 1.44（总计数）/ 1.91（净计数），不是 Opus 的 2.66；最小可探测 0.31–0.39 keV，到不了 0.10 keV 的门，也过不了 0.11–0.20 keV 的门对照散差）；(a) 接受 AMBIGUOUS；(c) 已撤回：P26 天空归一研究（R061/D100）——sky_background_scale 同时乘在源 APEC 上，天空归一的 ×0.8/×1.2 对照（D55 的 b: 族，只乘天空成分）早已跑过：天空 B 上 ΔkT 的移动 ≤0.052 keV，天空 A 上 S4 ×1.2 为 −0.114 keV，已包含在 0.11–0.20 keV 的门对照散差里；(d) 增加曝光：不可得。阶段 5 的判定保持 AMBIGUOUS。“先用偏置指向 0501/0601 固定天空”已撤回（R055/D95：天空参考本来就是 0701，且偏置环弱 3–21×）。P20 已完成（E048/T173/R039/D79）：在真实能段 0.375–9.0 keV 上，只遮延展源+恒星仍把 R2-4_Angle1 的 kT 推高 +0.287（AGN 注入）/ +0.279（纯热）→ 全套 v2 点源 mask 保留（机制仍被质疑）。P21 已完成（E050/T175/R049/D89）：钉住一个 v18 世代后，image/v18 比值 没有任何 ≥2σ 的径向上升（0.929 / 1.477 / 1.369，每步 <0.5σ，R12-14 未定义），归档的 3.32 来自不可重放的世代；绝对量与阶段 2 差 ≈2.25× 未调和。Q-D：2D 强度阶段 2 完成（E042/R016：2–5′ 不对称与 β 椭圆率一致，≥+30% 双锥被排除），2D kT 阶段 3 完成（E043/R018：估计器 E3 + cone45，双臂一致，共同足迹 56%），折叠图 v2 已出（E046/T171），生产图的东箭头缺陷未修好、总览图不展示。红队：RT-01…RT-60 共 55 支已分诊（RT-46 → R055/D95、RT-47 → R056/D96、RT-48 与 RT-50 → R059/D99、RT-49 → R058/D98、RT-51 → R061/D100、RT-52 → R060、RT-53 → R061/R062/D100/D101、RT-54 → R062/D101、RT-55 → R063/D102、RT-56 与 RT-57 → R064/D103），RT-36 无报告死亡 （由 RT-38 取代）、RT-33 无异议死亡（由 RT-35 重试）、RT-46/RT-47 首次启动即死后已重启（F037/F039）；RT-51（攻击 RT-48 O1b 对 P19 XMM 腿的解释）→ R061/D100、RT-53（P26 设计审计：P26 已撤回）→ R061/R062/D100/D101、RT-54（选项 b 可行性）→ R062/D101、RT-55（Q-D 温度图主张）→ R063/D102、RT-56（本手册 vs cognition-os）与 RT-57（D84c）→ R064/D103，均已分诊；RT-52（记录一致性审计）已分诊 → R060；RT-58/RT-59/RT-60 运行中 → 待定。
The plan comes entirely from cognition-os/02_CURRENT_QUESTION.md and the job briefs/reports; it is not invented by this handbook. Update 2026-10-17 23:27Z (stage 25, refresh 6): P19 (the Chandra matched-sector re-fit) is closed — the Chandra-only kT is 0.690/0.755/0.747/0.858 keV and kT(8–10′) − kT(2–4′) = +0.168 keV at only 1.26σ → the verdict is NO POWER, which neither corroborates nor refutes the rise question; the apertures are only 35–62 % of the nominal annuli (the report's "79–100 %" is an area ratio) and the minimum detectable ΔkT is 0.30–0.33 keV (the 1.0762 keV is a grid-clip artefact and is never quoted); P19's XMM-only kT (0.20–0.32 keV) is retired as a background-construction artefact (0.316 → 0.710 keV on the same data). Q-B's decisive test (the D55 pre-registered matched-contract test) is done and its verdict is AMBIGUOUS — underpowered, not a "no rise" (2026-10-17 07:15Z, E049/T174/R040/D80; Provisional). The spectral fit S4's ΔkT(8–10′−2–4′) = +0.13 (sky A, 95 % −0.21…+0.42) / +0.19 (sky B, −0.004…+0.42) and the Fe-L hardness +0.01 (−0.32…+0.46) / +0.07 (−0.13…+0.33); the minimum detectable ΔkT's 0.43 keV is only the Gaussian approximation, and under the pre-declared non-parametric definition (50 % power, 95 % interval) both methods exceed 0.50 keV (R042/D82). An independent expert re-derivation (Opus, D59a) confirms AMBIGUOUS. Stage 5b (the re-run on the regenerated QPB) was triaged at 12:48Z: the verdict is unchanged (R046/D86) — sky B's ΔkT +0.1886 → +0.1982 (−0.034…+0.478) and the QPB-construction systematic is 0.010 keV; P22's regenerated QPB was accepted at 11:04Z (9/9 apertures pass the magnitude gate, R 0.884–1.005, R045/D85). Update 2026-10-18 07:46Z (stage 27, refresh 7 final): the PI answered (D105, 05:22Z) — "re-extracting spectra has always been allowed", so re-extraction with ANY region design is standing-authorised, limited only by the Virgo resource gates (memory ≤ 60 %, measured ≈58 % at 05:22Z); option (b) no longer "needs PI approval". Three new tasks were dispatched accordingly: P27 (D105: MOS1+pn0 for R2-4, R4-6, R6-8, R8-10 plus R10-12 on the stage-1b contract, whole-profile joint fit with shared nuisance parameters, pre-registration first, queued behind the Virgo memory gate), P28 (PI order D106: M104-only multi-band surface-brightness images in nine bands — 0.4–1.25, 0.4–0.6, 0.6–0.8, 0.8–1.0, 0.8–1.2, 0.4–0.5, 0.5–0.6, 0.6–0.7, 0.7–0.8 keV — with QPB/SP/LHB/MWhalo/CXB removed, point sources moderately masked or PSF-subtracted, plus one-quadrant folded versions; running on codexnim), P29 (the D55-named but never-executed 0.5–1.2 keV secondary S4 arm plus the R11-13_Angle2 lane; running on codexnim). Update 2026-10-18 09:48Z (stage 28): P29 is DONE (R067/D108, Provisional) — the secondary arm gives ΔkT(R8-10 − R2-4) = +0.077 keV (sky A, 95 % [−0.307, +0.479]) / +0.145 keV (sky B, 95 % [−0.142, +0.481]), the same sign as the primary (+0.094 / +0.198), but the 95 % width ratio is 1.00 (A) / 1.22 (B) → it does not narrow the interval, and the Q-B verdict stays AMBIGUOUS; whether the secondary is independent corroboration is under RT-65 (pending). P27 is NOT launched in the RT-54 form (R067/D108): the outer annulus holds 98 % of the information budget, MOS1 lanes buy no information, and R10-12/R12-14 hit the FoV/vignetting wall → the outer annulus is capped at 12′ with no MOS1 lanes, replaced by a pn0 R2-4 + R8-10 preflight (≈6–8 Virgo wall-hours) with a short pre-registration (D55 logic) written before any fit; the best design reaches only MDΔkT 0.22–0.23 keV against the 0.10 keV gate (the controls, 0.12–0.20 keV, bind) → it can narrow the interval but is unlikely to flip AMBIGUOUS (RT-64's numbers are report-based). P28 is still running (≈3.5 h), no result yet, pending. (Stage-29 correction: P28's products exist, but make_figures.py's extent has top/bottom swapped (an axis bug) and the figures are being corrected; P28 is now "Pending" and no P28 figure is shown until the fix, D112d.) Stage 5b is the bundle of record (D104a): kT(R8-10) 0.9446 keV, ΔkT +0.198 keV, 95 % [−0.034, +0.478] (S4). Red-team census (stage 29, refresh 8): RT-01…RT-69, 65 triaged (RT-33 has no queue row; RT-36 died without a report and was replaced by RT-38; RT-65 → R070/D112, RT-66 → R068/D109 and RT-67 → D112 are all triaged; RT-68 (the P28 pipeline code audit) and RT-69 (the P30 design audit) are running). The earlier record (2026-10-17 23:27Z, stage 25, refresh 6) had four options (a–d, D98): option (b) no longer needs approval and has become P27/P29 (adding MOS1+pn0 gives an information gain of 1.44 (gross-count) / 1.91 (net-count), not Opus's 2.66, and a minimum detectable of 0.31–0.39 keV, which cannot reach the 0.10 keV gate nor pass it against the 0.11–0.20 keV gate-control spread); (a) accept AMBIGUOUS; (c) WITHDRAWN: the P26 sky-normalisation study (R061/D100) — sky_background_scale also multiplies the source APEC, and the sky-normalisation ×0.8/×1.2 controls required by D55 (the b: family, which scale the sky components only) have already run: on sky B ΔkT moves by ≤0.052 keV, on sky A the S4 ×1.2 control moves it by −0.114 keV, all inside the 0.11–0.20 keV gate-control spread; (d) more exposure: not available. Stage 5's verdict stays AMBIGUOUS. "First fix the sky from the offset pointings 0501/0601" is withdrawn (R055/D95: the sky reference is already 0701, and the offset annuli are 3–21× weaker). P20 is done (E048/T173/R039/D79): on the real band 0.375–9.0 keV, masking only extended sources + stars still pushes R2-4_Angle1's kT up by +0.287 (AGN-injected) / +0.279 (thermal-only) → the full v2 point mask stays (the mechanism stays Questioned). P21 is done (E050/T175/R049/D89): with one v18 generation pinned, the image/v18 ratio shows no ≥2σ rise with radius (0.929 / 1.477 / 1.369, every step <0.5σ, R12-14 undefined), and the archived 3.32 came from an unreplayable generation; the absolute level differs from stage 2 by ≈2.25× and is unreconciled. Q-D: 2D intensity stage 2 done (E042/R016: the 2–5′ asymmetry is consistent with β ellipticity, a ≥ +30 % bicone is excluded), 2D kT stage 3 done (E043/R018: estimator E3 + cone45, both arms agree, common footprint 56 %), the folded v2 map is out (E046/T171), the production map's east-arrow defect is not yet fixed so the overview figure is not shown. Red teams: RT-01…RT-60, i.e. 55 triaged (RT-46 → R055/D95, RT-47 → R056/D96, RT-48 and RT-50 → R059/D99, RT-49 → R058/D98, RT-51 → R061/D100, RT-52 → R060, RT-53 → R061/R062/D100/D101, RT-54 → R062/D101, RT-55 → R063/D102, RT-56 and RT-57 → R064/D103); RT-36 died without a report (replaced by RT-38), RT-33 died without objections (retried as RT-35), and RT-46/RT-47 each died at their first launch and were relaunched (F037/F039); RT-51 (attacking RT-48 O1b's reading of the P19 XMM leg) → R061/D100, RT-53 (a design audit of P26, which led to P26 being withdrawn) → R061/R062/D100/D101, RT-54 (option-b feasibility) → R062/D101, RT-55 (the Q-D kT-map claims) → R063/D102, RT-56 (this handbook vs cognition-os) and RT-57 (D84c) → R064/D103 are all triaged; RT-52 (the consistency audit) is triaged → R060; RT-58/RT-59/RT-60 are running → pending.
| 周Week | 内容Content | |
|---|---|---|
| 11 | 已完成：Q-A 清单、试点、阶段 2 全网格（4,800 拟合）→ MOS2 内区参考（D48/I067） | Done: Q-A inventory, pilot, stage-2 full grid (4,800 fits) → MOS2 inner reference (D48/I067) |
| 11 | 已完成：Q-C 判为假象（I065）；红队 RT-01 攻击中；15–45′ 轮廓 job 运行中 | Done: Q-C judged an artefact (I065); red team RT-01 attacking; the 15–45′ profile job running |
| 1–21–2 | 已完成：Q-D 亮度图；2D 代理阶段 2（E038：联合 MOS2+pn0、校准对比区间、N>S 提示）；下一步 = Virgo 阶段 1 谱检验（P3 check_N/S_R2-4.9） | Done: the Q-D intensity map; 2D proxy stage 2 (E038: joint MOS2+pn0, calibrated contrast bounds, N>S a hint); next = the Virgo stage 1 spectral check (P3 check_N/S_R2-4.9) |
| 1–21–2 | 待匹配合同：谱 vs Fe-L 硬度分歧（I068/I070）不能在旧产品上裁决；只由 D55 预注册的匹配合同测试判定（阶段 1b 的 v2-mask 产品就绪后）；0900170501/0601 ODF 已下载校验（22:50Z）→ 阶段 1b 通过后 Virgo 重处理（外区 profile） | Awaiting the matched contract: the spectral vs Fe-L-hardness disagreement (I068/I070) cannot be settled on the old products; it is decided only by the D55 pre-registered matched-contract test (once the stage-1b v2-masked products exist); the 0900170501/0601 ODFs are downloaded and verified (22:50Z) → Virgo re-reduction for the outer profile after stage 1b passes |
| 2–42–4 | Q-A 阶段 3/4 完成（S4 外区/边缘未合格 → 探索性，D51/D54）；Virgo 阶段 1 完成但口径谱只有自动 mask（B023）→ 阶段 1b 重抽 v2-mask 口径谱：2026-10-17 02:45Z 已交付、门 2 通过（D76c：9/9 口径、MANIFEST 506/506 OK、BACKSCAL 真实、REG 身份 PASS），但交付的 QPB 大 1e6–1e7 倍（B026/R037/D77）→ mask/口径/响应被接受、QPB 不被接受；阶段 5 = 匹配合同测试（D55 预注册规则，已冻结）等阶段 1b；mask 阶梯"轻 mask"结论被 RT-17 推翻（R020/D64：全套 v2 mask 保留、阶段 B 取消）；红队 RT-15/RT-16/RT-17 已分诊（D62/D63/R020），RT-18/RT-19 已分诊（B025/R023、R022/D65），RT-20/RT-21 运行中；P16 测量 as-run 能段与 RASS 过滤器对已存 kT 的影响（待定）；P18 端到端模拟（含 S7 = PI 的冷却流情景）说明数据排除哪些冷却流版本（待定） | Q-A stages 3–4 done (S4 outer/edge not qualified → exploratory, D51/D54); Virgo stage 1 finished but its aperture spectra carry only the automatic mask (B023) → stage 1b re-extracts the v2-masked aperture spectra: delivered 2026-10-17 02:45Z with gate 2 PASSED (D76c: 9/9 apertures, MANIFEST 506/506 OK, real BACKSCAL, REG identity PASS), but the delivered QPB is 1e6–1e7× too large (B026/R037/D77) → the masks/apertures/responses are accepted, the QPB is not; stage 5 = the matched-contract test (D55 pre-registered rule, frozen) waiting for stage 1b; the mask ladder's "lighter mask" conclusion was overturned by RT-17 (R020/D64: the full v2 mask stays, stage B cancelled); red teams RT-15/RT-16/RT-17 triaged (D62/D63/R020), RT-18/RT-19 triaged (B025/R023, R022/D65), RT-20/RT-21 running; P16 measures the effect of the as-run band and the RASS filter on the saved kT (pending); P18, the end-to-end mock (including S7, the PI's cooling-flow scenario), will say which cooling-flow versions the data exclude (pending) |
本次更新
What's new
- 2026-10-18 13:03Z（阶段 29，刷新 8）：Q-B 的判据换成显著性（D110/D111，PI 原话）——任何外围 kT 比内区 kT 高 ≥3σ 即算（2σ 亦可接受），不再要求 σ(kT) 的绝对门槛（0.05/0.10 keV）；今天的最佳 2D 图箱 z = 1.72（r = 7.18′，0.880 ± 0.079 keV vs 内区 0.7403 ± 0.0186 keV）、环带 z = 1.69，无一达到 3σ（Provisional）。新任务 P30 立项（D111，运行中，Pending）：region × 相机 × 观测的预测优先设计搜索，中期结论 = 统计量本可到 z ≈ 5.7（ΔkT = 0.2 keV），但对照移动包络（0.12–0.20 keV）把预期 z 压在 1.0–1.7，瓶颈是对照移动、不是计数。可靠性条件被 R068/D109 整体替换（红队 RT-66 分诊）：背景占比而非计数、每个预注册对照 < 0.10 keV、方法一致性对着指名的谱比较器、边界与多起点分开、三相机一致在 pn0 之前不可检验、"0.05 keV"只是启发式（已撤下）；内区 2–4′ 的 MOS2-only kT 改称"条件比较器"。P29 的措辞被 R070 订正（红队 RT-65 分诊）：0.5–1.2 keV 是同一批谱的次能段重读，不是独立佐证；天空 A 的 +0.077 keV 来自 SP 钉在零的拟合。P28 改为"待定 / Pending"（D112d：图的坐标轴有 bug，修好前不给 PI 看图）。红队普查：RT-01…RT-69，65 支已分诊（RT-65/RT-66/RT-67 均已分诊；RT-68 与 RT-69 运行中）。
- 2026-10-18 13:03Z (stage 29, refresh 8): Q-B's criterion becomes significance (D110/D111, the PI's own words) — ANY outer kT above the inner kT by ≥3σ is enough (2σ may also be acceptable), with no absolute σ(kT) threshold (0.05 or 0.10 keV) required; today's best 2D-map bin is z = 1.72 (r = 7.18′, 0.880 ± 0.079 keV against the inner 0.7403 ± 0.0186 keV) and the annulus z = 1.69, none reaching 3σ (Provisional). The new task P30 is launched (D111, running, Pending): a forecast-first design search over region × camera × observation; the interim conclusion is that statistics alone could reach z ≈ 5.7 (at ΔkT = 0.2 keV), but the control-shift envelope (0.12–0.20 keV) caps the expected z at 1.0–1.7 — the bottleneck is the control shifts, not counts. The reliability conditions are replaced wholesale by R068/D109 (red team RT-66 triaged): background fraction rather than counts, every pre-registered control < 0.10 keV, method agreement against a named spectral comparator, bounds reported separately from multistart, three-camera agreement untestable until pn0, and the "0.05 keV" was only a heuristic (now dropped); the inner 2–4′ MOS2-only kT is relabelled a "conditional comparator". P29's wording is corrected by R070 (red team RT-65 triaged): the 0.5–1.2 keV result is a sub-band re-read of the same spectra, not independent corroboration; sky A's +0.077 keV comes from a fit with SP pinned at zero. P28 is changed to "Pending" (D112d: the figure frame has an axis bug, so no P28 figure is shown to the PI before the fix). Red-team census: RT-01…RT-69, 65 triaged (RT-65/RT-66/RT-67 are all triaged; RT-68 and RT-69 are running).
- 2026-10-18 09:48Z（阶段 28，短补丁）：红队普查更新——RT-01…RT-66，62 支已分诊（RT-36 无报告死亡；RT-65（P29 次要臂的独立性）与 RT-66（kT 可靠性条件）运行中；RT-63（P28 简报审计）与 RT-64（P27 设计审计）已于 09:17Z 分诊 → R067/D108）。（阶段 29 订正：RT-65 已于 12:26Z 分诊 → R070/D112，RT-66 已于 11:03Z 分诊 → R068/D109；现行普查见本页最新条目。）P29 已完成（Provisional）：0.5–1.2 keV 次要 S4 臂给出 ΔkT(R8-10 − R2-4) = +0.077 keV（天空 A，95% [−0.307, +0.479]）、+0.145 keV（天空 B，95% [−0.142, +0.481]）——与主臂（+0.094 / +0.198）同号，但 95% 宽度比 1.00（A）/ 1.22（B）：不收窄区间；Q-B 判定保持 AMBIGUOUS（Provisional）；次要臂是否独立佐证由 RT-65 检验中（待定）。P27 不按 RT-54 形式启动（R067/D108）：外环信息预算占 98%、MOS1 通道买不到信息、R10-12/R12-14 撞 FoV/vignetting 墙 → 外环封顶 12′、无 MOS1 通道；改为 pn0 R2-4 + R8-10 预检（≈6–8 个 Virgo wall-h），先写短预注册（D55 逻辑）再拟合；最佳设计只能到 MDΔkT 0.22–0.23 keV 对 0.10 keV 的门（对照 0.12–0.20 keV 在约束）→ 能收窄区间，但不太可能翻转 AMBIGUOUS（RT-64 数字为 report-based）。P28 仍在运行（≈3.5 h），无结果，待定。
- 2026-10-18 09:48Z (stage 28, short patch): red-team census updated — RT-01…RT-66, 62 triaged (RT-36 died without a report; RT-65 (whether P29's secondary arm is independent corroboration) and RT-66 (whether the kT-reliability conditions are supported by project evidence) are running; RT-63 (the P28 brief audit) and RT-64 (the P27 design audit) were triaged at 09:17Z → R067/D108). (Stage-29 correction: RT-65 was triaged at 12:26Z → R070/D112 and RT-66 at 11:03Z → R068/D109; see this page's newest entry for the current census.) P29 is DONE (Provisional): the 0.5–1.2 keV secondary S4 arm gives ΔkT(R8-10 − R2-4) = +0.077 keV (sky A, 95 % [−0.307, +0.479]) and +0.145 keV (sky B, 95 % [−0.142, +0.481]) — the same sign as the primary (+0.094 / +0.198), but the 95 % width ratio is 1.00 (A) / 1.22 (B): it does not narrow the interval; the Q-B verdict stays AMBIGUOUS (Provisional); whether the secondary is independent corroboration is under RT-65 (pending). P27 is NOT launched in the RT-54 form (R067/D108): the outer annulus holds 98 % of the information budget, MOS1 lanes buy no information, and R10-12/R12-14 hit the FoV/vignetting wall → the outer annulus is capped at 12′ with no MOS1 lanes; replaced by a pn0 R2-4 + R8-10 preflight (≈6–8 Virgo wall-hours) with a short pre-registration (D55 logic) written before any fit; the best design reaches only MDΔkT 0.22–0.23 keV against the 0.10 keV gate (the controls, 0.12–0.20 keV, bind) → it can narrow the interval but is unlikely to flip AMBIGUOUS (RT-64's numbers are report-based). P28 is still running (≈3.5 h), no result yet, pending.
- 2026-10-18 07:46Z（阶段 27，刷新 7 收官）：四页全部带到 R066/D107。红队普查：RT-01…RT-64，60 支已分诊（RT-36 无报告死亡；RT-63（P28 简报审计，hermes）与 RT-64（P27 设计审计，codexnim）运行中；RT-58/RT-59/RT-60/RT-61/RT-62 已于 06:40Z 分诊 → R066/D107）。Q-C 整体改标 Questioned（相机相关，R065/D104b）：双相机逆方差加权（IVW）均值看不到 ≥2σ 的上升，但两台相机在 R11-13 上相差 2.35σ（mos1 闭合 8.53 ± 3.12 vs mos2 0.74 ± 1.13），mos1 单独在 R11-13 上升 +2.33σ → 旧措辞"13′ profile 超出就是 B020 假象"按原样撤回（双重面积除法只会降低该比值，符号不对）。I069 的外/边缘失败单元数被记反（R065）：失败的是外区 S4 6/8（S2 3/8）、边缘 S4 5/8（S2 2/8）、proxy S4 1/4（S2 0/4）——I069 原文把通过数写成了失败数；判定不变（外区不合格）。"p = 0.12"是单边（二项 P(X≥17|n=27,p=0.5)=0.1239；双边 0.248）。D84c 已解决（R064/D103）："0.34 keV 冲突"是一个冻结的重绘点（0.6235 keV）对 CSV 值；v18 R8-10 扇区两半 0.9018/0.9669 keV 把阶段 5 的 0.9351 keV 夹在中间（相差 0.03 keV 以内）。阶段 5b 是记录版本（D104a）：kT(R8-10) 0.9446 keV、ΔkT +0.198 keV、95% [−0.034, +0.478]（S4）；天空 A 的 ×1.2 对照（−0.114 keV）是塌掉的拟合，不得当系统误差引用（D107c）。PI 已答复（D105，05:22Z）："重新抽谱一直是可以的"——用任何 region 设计重新抽谱长期授权，只受 Virgo 资源门限制（05:22Z 内存 ≈58%，限 60%）；据此派出三个新任务：P27（MOS1+pn0 抽 R2-4…R8-10 加 R10-12，阶段 1b 合同，整条 profile 共享扰动参数联合拟合，先预注册）、P28（PI 命令 D106：M104 专用九能段面亮度图，QPB/SP/LHB/MWhalo/CXB 扣除，点源适度遮蔽或 PSF 相减，单象限折叠版本，codexnim 运行中）、P29（D55 点名的 0.5–1.2 keV 次要 S4 臂 + R11-13_Angle2 通道，codexnim 运行中）。
- 2026-10-18 07:46Z (stage 27, refresh 7 final): all four pages brought up to R066/D107. Red-team census: RT-01…RT-64, 60 triaged (RT-36 died without a report; RT-63 (the P28 brief audit, hermes) and RT-64 (the P27 design audit, codexnim) are running; RT-58/RT-59/RT-60/RT-61/RT-62 were triaged at 06:40Z → R066/D107). Q-C is re-labelled Questioned as a whole (camera-dependent, R065/D104b): the two-camera inverse-variance-weighted (IVW) mean shows no ≥2σ rise, but the two cameras disagree by 2.35σ at R11-13 (mos1 closure 8.53 ± 3.12 vs mos2 0.74 ± 1.13) and mos1 alone rises +2.33σ at R11-13 → the old wording "the 13′ profile excess is the B020 artefact" is withdrawn as worded (the double-area division only lowers that ratio, so the sign is wrong). I069's outer/edge failure counts were inverted (R065): the failed cells are outer S4 6/8 (S2 3/8), edge S4 5/8 (S2 2/8), proxy S4 1/4 (S2 0/4) — I069 had written the PASSING counts as failures; the verdict is unchanged (not qualified outside). "p = 0.12" is one-sided (binomial P(X≥17|n=27,p=0.5)=0.1239; two-sided 0.248). D84c is RESOLVED (R064/D103): the "0.34 keV conflict" was a frozen replot point (0.6235 keV) versus the CSV value; v18's R8-10 sector halves 0.9018/0.9669 keV bracket stage 5's 0.9351 keV within 0.03 keV. Stage 5b is the bundle of record (D104a): kT(R8-10) 0.9446 keV, ΔkT +0.198 keV, 95 % [−0.034, +0.478] (S4); sky A's ×1.2 control (−0.114 keV) is a collapsed fit and must not be quoted as a systematic (D107c). The PI has answered (D105, 05:22Z): "re-extracting spectra has always been allowed" — re-extraction with ANY region design is standing-authorised, limited only by the Virgo resource gates (memory ≈58 % at 05:22Z against the 60 % limit); three new tasks were dispatched accordingly: P27 (MOS1+pn0 for R2-4…R8-10 plus R10-12 on the stage-1b contract, whole-profile joint fit with shared nuisance parameters, pre-registration first), P28 (PI order D106: M104-only nine-band surface-brightness images with QPB/SP/LHB/MWhalo/CXB removed, point sources moderately masked or PSF-subtracted, one-quadrant folded versions; running on codexnim), P29 (the D55-named 0.5–1.2 keV secondary S4 arm plus the R11-13_Angle2 lane; running on codexnim).
- 2026-10-18 04:43Z（阶段 26，刷新 7）：修订读者级红队 RT-56 对本手册审出的 11 条缺陷（R064/D103）。红队普查改为 RT-01…RT-60，55 支已分诊（RT-36 无报告死亡；RT-58/RT-59/RT-60 运行中）——此前本页把已分诊的 RT-51/RT-53/RT-54 仍写成“运行中”，且普查数写成 49/50 支，全部订正：RT-51 → R061/D100、RT-52 → R060、RT-53 → R061/R062/D100/D101、RT-54 → R062/D101、RT-55 → R063/D102、RT-56 与 RT-57 → R064/D103。P26（天空归一研究）确认已撤回（R061/D100）：
sky_background_scale同时乘在源 APEC 上，且 D55 的天空归一 ×0.8/×1.2 对照（b: 族，只乘天空成分）早已跑过——天空 B 上 ΔkT 移动 ≤0.052 keV，天空 A 上 S4 ×1.2 为 −0.114 keV。Q-B 选项 (b) 不便宜（RT-54 已分诊 → R062/D101）：v2 合同上没有任何 MOS1/pn0/8–12′ 产品，阶段 5 执行器 MOS2-hardcoded，据估算 ≈40–45 个 Virgo wall-h（report-based，按估算引用）且需一份新预注册；分派方建议默认选 (a)。PI 对 (a)/(b)/(d) 的选择仍未回答，阶段 5 判定保持 AMBIGUOUS。Q-D 措辞按 R063/D102 收窄：温度图只是在逐箱功率之内近等温 0.70–0.80 keV（中位 σ68 0.059 keV 对 0.10 keV 带宽；r≥8′ 的 6 个箱里 4 个分不开 0.6 与 1.0 keV）；逐箱中位 0.748 keV（0.736 是池化全局值）。状态页的“径向温度数值”一行从“已确立（Trusted）”移到“暂定（Provisional）”：数字本身来自 I061（validated-synthesis），但 I061 与信任表（T133/T134）都把该 profile 标为 Provisional/Questioned，而同一页的 §2 本来就这样写——两处不再互相矛盾。 - 2026-10-18 04:43Z (stage 26, refresh 7): the 11 defects that reader-level red team RT-56 filed against this handbook are fixed (R064/D103). The red-team census now reads RT-01…RT-60 with 55 triaged (RT-36 died without a report; RT-58/RT-59/RT-60 are running) — this page previously wrote the already-triaged RT-51/RT-53/RT-54 as "running" and gave the census as 49/50; all corrected: RT-51 → R061/D100, RT-52 → R060, RT-53 → R061/R062/D100/D101, RT-54 → R062/D101, RT-55 → R063/D102, RT-56 and RT-57 → R064/D103. P26 (the sky-normalisation study) is confirmed withdrawn (R061/D100):
sky_background_scalealso multiplies the source APEC, and D55's sky-norm ×0.8/×1.2 controls (the b: family, which scale the sky components only) already ran — on sky B ΔkT moves by ≤0.052 keV and on sky A the S4 ×1.2 control by −0.114 keV. Q-B option (b) is NOT cheap (RT-54 triaged → R062/D101): no MOS1/pn0/8–12′ products exist on the v2 contract, the stage-5 executor is MOS2-hardcoded, it costs an estimated ≈40–45 Virgo wall-hours (report-based, quoted as an estimate) plus a new pre-registration, and the dispatcher recommends (a) by default. The PI's choice among (a)/(b)/(d) is still unanswered and stage 5's verdict stays AMBIGUOUS. The Q-D wording is narrowed per R063/D102: the temperature map is near-isothermal 0.70–0.80 keV only within per-bin power (median σ68 0.059 keV against the 0.10 keV band; 4 of the 6 bins at r≥8′ cannot separate 0.6 from 1.0 keV); the per-bin median is 0.748 keV (0.736 is the pooled global value). The state page's "radial temperature values" row moved from "Established (Trusted)" to "Provisional": the numbers come from I061 (validated-synthesis), but I061 and the trust table (T133/T134) both label that profile Provisional/Questioned, as the same page's §2 already did — the two sections no longer contradict each other. - 2026-10-18 01:04Z（评审更正）：Q-B 选项 (c)=P26 已撤回（R061/D100）——红队 RT-53 发现 sky_background_scale 同时乘在源 APEC 上，而 D55 要求的天空归一 ×0.8/×1.2 对照（b: 族，只乘天空成分）早已跑过：天空 B 上 ΔkT 移动 ≤0.052 keV，天空 A 上 S4 ×1.2 为 −0.114 keV。RT-51 终审：P19 的 XMM-only kT 偏冷的方向成立，但“模型 B = 0.71 keV”不是干净数字（α 比 9–12 keV 锚点高 5.07σ，MWhalo kT 贴着 0.4 的壁），与阶段 5 的“一致”也不独立；Chandra-only 仍是 Provisional。现在给 PI 的真实选项只剩 (a) 接受 AMBIGUOUS、(b) 加 MOS1+pn0 / 放宽到 8–12′（可行性由 RT-54 检查）、(d) 增加曝光（不可得）。
- 2026-10-17 23:27Z（阶段 25，刷新 6）：P19（Chandra 匹配扇区重拟合）已收官（E053/T177/R057/R059/D97/D99，执行器 18.6 h）——Chandra-only kT 0.690 / 0.755 / 0.747 / 0.858 keV（R2-4…R8-10），kT(8–10′) − kT(2–4′) = +0.168 keV 只有 1.26σ → 判定 NO POWER，对“上升”既不证实也不否证；孔径只是名义环带的 35–62%（报告的“79–100%”是 XMM/Chandra 面积比，不是覆盖率 → D69 的匹配孔径异议仍成立）；最小可探测 ΔkT 0.30–0.33 keV（1.0762 keV 是形状网格 0.50–1.00 keV 截断的假象，永不引用）。P19 的 XMM-only kT（0.20–0.32 keV）已退役：它是背景构造假象——加上天空成分后 R2-4 从 0.316 移到 0.710 keV（同一批数据），与阶段 5 的 0.75/0.94 keV 冲突；其解释正被 RT-51 攻击。红队普查：RT-01…RT-52 共 49 支已分诊（此数为 2026-10-17 23:27Z 当时的快照，已过时；现行普查为 RT-01…RT-60、55 支已分诊，见本页最新一条）——RT-46 → R055/D95（“偏置指向当天空”撤回：天空参考本来就是 0701，偏置环弱 3–21×）、RT-47 → R056/D96（闭合步长 +0.94σ / +0.38σ，总 +0.78σ；“形态比值 <5%”只对 2–5′）、RT-49 → R058/D98（信息增益 1.44/1.91 是 σ 比的平方；v18 的
sky_background_scale在 1.0 处冻结、从未被拟合）、RT-48 与 RT-50 → R059/D99（P19 终报）；RT-51 与 RT-53 运行中（RT-52 已分诊 → R060）（这三支其后均已分诊：RT-51 → R061/D100、RT-53 → R061/R062/D100/D101，另 RT-54 → R062/D101、RT-55 → R063/D102、RT-56 与 RT-57 → R064/D103）。Q-B 的四个选项已交 PI（D98，待回复）：(a) 接受 AMBIGUOUS；(b) 加 MOS1+pn0 / 放宽到 8-12′，MDΔkT 0.31–0.39 keV，到不了 0.10 keV 的门；(c) 已撤回：P26 天空归一研究（R061/D100）——sky_background_scale 同时乘在源 APEC 上，天空归一的 ×0.8/×1.2 对照（D55 的 b: 族，只乘天空成分）早已跑过：天空 B 上 ΔkT 的移动 ≤0.052 keV，天空 A 上 S4 ×1.2 为 −0.114 keV，已包含在 0.11–0.20 keV 的门对照散差里；(d) 更多曝光：不可得。阶段 5 的判定保持 AMBIGUOUS。Q-C 不变：无 ≥2σ 径向上升、绝对比值未定、闭合步长 +0.94σ / +0.38σ（总 +0.78σ）、R11-13 mos2 是 0.54σ 非探测。 - 2026-10-17 20:58Z（阶段 23/24，由分派方补完）：RT-46、RT-47 已分诊（R055/D95、R056/D96），首页与状态页不再写“运行中”；P23 的“先用偏置指向 0501/0601 固定天空”已撤回，Q-B 的选项待 RT-49 检验后再问 PI；闭合步长改为 +0.94σ / +0.38σ（总计 +0.78σ），“形态比值对 QPB 变动 <5%”只对 2–5′ 成立。红队 RT-48（P19 的 XMM 腿）与 RT-49（Q-B 选项、天空场 0701 是否干净）运行中 → 待定。
- 2026-10-17 19:41Z（阶段 22：修订读者级红队 RT-45 的 6 条缺陷 + 写入 R050–R053 的新结果 + 新增组会页）：RT-45（把本手册当读者用一遍）已分诊 → R054/D94，6 条缺陷全部修订——(1) 首页计划表把已完成的 P21（14:34Z，E050/T175/R049/D89）仍写成"运行中；结果待定"，已改为已完成并补上 0.929/1.477/1.369 的结果；(2) 同表把已被接受的 P22 与已分诊的阶段 5b仍写成"运行中 → 待定"并引用被取代的 3/9 中期数字，已改为 11:04Z 接受（9/9 口径，R 0.884–1.005）与 12:48Z 分诊（判定不变，天空 B ΔkT +0.1982）；(3) "一支红队启动即死"不成立（队列与 08_RUNNING 都记录 RT-44 于 13:40Z 启动并活跃），红队台账改为 RT-01…RT-45 共 43 支已分诊、RT-46/RT-47 运行中；(4) 13′ 凸起不再写成裸 1.37，改为记录要求的 "仅 Angle2、1.0–2.0，取决于 bin 配对/世代，任何世代都没有保存流量误差"（I065 的 RT-01 修订）；(5) 引用的
cognition-os/CLAUDE.md不存在，改为仓库根目录的CLAUDE.md；(6) physics.html 教学算例的四个中间数曾经差 10×/10×/≈2000×，按页面自己的输入重算订正为 9.2×10⁻⁶⁷ cm⁻²、6.5×10⁶⁰ cm⁻³、2.35×10⁶⁸ cm³、n_e ≈ 1.7×10⁻⁴ cm⁻³（D = 9.55 Mpc = 2.95×10²⁵ cm，norm 6×10⁻⁶），并把"典型晕密度"改成对正确值成立的表述。新结果（全部来自 cognition-os，quote 数字不超出记录）：P21 的构造被质疑（Questioned）——R7-9 两台相机的逐相机闭合 1.311（mos1）vs 0.737（mos2），公布的 1.018 只是两个散差的简单平均 →"闭合 1.018 证明天空扣除正确"的论证撤回；P24 重放显示该差别 0.574 ± 0.511 = 1.1σ，是泊松散差被净/总计数比（净只有总的 13–21%）放大，不是校准分歧（"相机不一致"的说法也撤回）；绝对 image/v18 比值仍未定（天空模型 vs 经验天空把 R7-9 从 0.93 移到 1.77）；判定"没有 ≥2σ 径向上升"保持 Provisional。P25：阶段 2 的软能段 QPB 图像归一是正确的（等于 FWC 重建的 1.4–4%，含 Al-K/Si-K 仪器线；2.13× 是"线在内 vs 线在外"），E042 的形态学比值移动 <5% → 形态学结论成立，但阶段 2 的绝对面亮度仍是 Questioned（±50–85% 的净）。B029：P21 的脚本给 mos2 用了 mos1 的线集（0.7–1.0 keV 上影响可忽略）。P19 仍在运行；RT-43 已按 interim 分诊（R050/D90）：Chandra 的 MDΔkT 只引稳健的 ≈0.30 keV（1.08 keV 是尾部假象），Chandra-only kT 为 Provisional 且不得当作晕的 profile。Q-B：阶段 5b 已完成（R046），P23 仍是等待 PI 的提案，RT-46 正在检验"偏置指向当天空"的想法 → 待定。新增组会页 meetings.html（首页顶部加卡片，全部页面导航加入口）：2026-10-15/16/17 各一块，分成果 / 工作 / 问题 / 待 PI 决定，每条结果带信任状态，问题栏含 agent 当天的错误与更正（P21 的单位错误 B028、线集错用 B029、P24 的 erratum、RT-44/RT-46 启动即死、"相机不一致"的过度表述被撤回、手册自身的 6 条缺陷）。 - 2026-10-17 15:23Z（阶段 15/15e：阶段 5 判定 AMBIGUOUS、P20/P21/P22 与阶段 5b、红队 RT-37…RT-42 全线分诊）：Q-B 的预注册判定出来了：AMBIGUOUS（功率不足），不是“没有上升”——ΔkT = kT(R8-10) − kT(R2-4)（目标 0）：谱拟合 S4 天空 A +0.13（−0.21…+0.42）、天空 B +0.19（−0.004…+0.42），Fe-L 硬度 H +0.01（−0.32…+0.46） / +0.07（−0.13…+0.33）；最小可探测 ΔkT 的 0.43 keV 只是高斯近似，按预先声明的非参数定义（50% 功率）两方法都 > 0.50 keV（R042/D82）。QPB 换血：交付的 back.pi 大 1e6–1e7 倍（B026）→ P22 重新生成 11:04Z 被接受（9/9 口径过量级门，R 0.884–1.005）→ 阶段 5b 在其上重跑 12:48Z 分诊：判定不变，天空 B 的 ΔkT +0.1886 → +0.1982（−0.034…+0.478），QPB 构造系统误差 0.010 keV（R046/D86）。P20（as-run 能段 mask 阶梯）：只遮延展源+恒星仍把 R2-4_Angle1 的 kT 推高 +0.287 / +0.279 keV → D64 被确认并加强，机制仍被质疑。P21（钉住 v18 世代）：image/v18 比值 没有任何 ≥2σ 的径向上升（0.929 / 1.477 / 1.369），归档的 3.32 来自不可重放的世代；绝对量与阶段 2 差 ≈2.25× 未调和。RT-41：折叠图的 +2.19σ 是 13 个事后半径里的最大值，不显著；预先声明的统计量最高只 +1.78σ。RT-42：P23 的信息增益是 1.44/1.91 而非 2.66，且门对照自己就移动 0.11–0.20 keV → 加功率本身不给 RISE；P23 尚未经 PI 批准，形状重排为先用偏置指向 0501/0601 测天空。
- 2026-10-17 05:08Z（阶段 14：按读者级红队 RT-32 的 9 条缺陷修订本手册，R037/D77）：手册自身被当作读者用一遍，9 条缺陷全部修订——(1) §3.1 环境块改为实际运行的那一套（CIAO 4.18：
source /home/huangrui/software/ciao-4.18/bin/ciao.sh、HEADAS=/home/huangrui/software/ciao-4.18/spectral、XSPEC_MDATA_DIR=.../modelData/，Python 3.12.12 + sherpa 4.18.0），原写法（heasoft + conda 环境的零字节 ciao.sh）读者照抄会失败；(2) 删掉run_v18_fit.sh与本手册“内容一致”的说法，改为注明它的第 7 行 source 的是另一套 ciao.sh、并非同一套环境；(3) 状态页 R4–6 的误差 0.041 → 保守的 0.104 keV（0.730 ± 0.104，与 CSV 和本页正文一致，来源figure/radial_thermo_profiles_v18/radial_kT_profile_v18_sector_mean.csv）；(4) 说明 R8–10 Angle1 的已存拟合退化，且 R8–10/R10–12/R12–14 是单扇区（provisional_single_sector），外侧轮廓只靠一个扇区支撑；(5) 红队台账与阶段 1b 状态更新：RT-29 → R034、RT-30 → R035、RT-31 → R036、RT-32 → R037，RT-33 运行中；阶段 1b 已交付（9/9 口径、MANIFEST 506/506 OK、BACKSCAL 真实、REG 身份 PASS），但交付的mos2U005-back.pi（QPB）在全部 9 个口径里大了 1e6–1e7 倍（阶段 5 执行器发现；门 2 从未检查量级）→ 阶段 5 改用经验证的修复 back.pi（derived-QPB），重生成 P22 运行中 → 待定，本手册不引它的数字；(6) Q-A 的“MOS2-only 是内区参考”只在 kT ∈ [0.45, 1.0] keV 成立（R036/D76）；(7) design.html 的“v2 mask 后仍 0.95”改为 ≈0.37（相对 v18 局部晕面亮度；0.95 只相对 v2 晕律，后者比 v18 暗 2.59×，引 E035 的 RT-05 修订）；(8) 新增“本页各图用哪个 M104 距离”一节：逐图列出 11.3 / 8.9 / 9.55 Mpc 及其来源，并警告在距离钉死之前跨图数字不可比；(9) 撤下ktmap_v2_pi_summary.png（东箭头画成朝西的显示缺陷，且 sha256 证实本手册嵌的是修复前副本）——不造假文件。检查：check_bilingual.py与check_console.sh全部 PASS。 - 2026-10-16 23:4xZ（阶段 13：R028–R033 / D69–D73 收尾）：Q-C 的“图/v18 比值 1.00 → 3.32 径向上升”被撤回（RT-28 → R033/D73）——它不是物理上升：Angle1 行不可重放（R7-9 没有 .fitting 会话）、不可复现（R9-11 有三代，CSV 比可重放会话高 58%）或不显著（R11-13 Δstat 1.36；R12-14 Δstat 0.0），而可复现的 Angle2 是平的 1.26 → 1.18；图内检验（E041：R11–13/R9–11 = 0.949 ± 0.029 / 1.029 ± 0.037）仍显示无 13′ 超出 → R030 的“无像过剩不成立”本身被撤回，图/v18 分支不可检验，直到每个区域钉住一个 v18 世代（提案 P21：CoPI，需 PI）。v18 合同再收窄（RT-27 → R032/D72）：v18 加载了 RASS 但从未给它模型 → RASS 不约束任何东西，“RASS 过滤器损坏”只是更小的症状；P16 的“kT 能段效应 < 1σ”保持 Provisional 并附限制（事后误差侧 0.18–0.68σ → 最高 0.83σ；天空整族穿过 LocalBubble.kT = 0.08 下界钉住的背景拟合；22 自由参数 vs 17–43 dof；保存拟合远离极小值 Δstat ≤ 39），天空归一未被证明能段稳健。D55 的组合压力对照判据（RT-26 → R031/D71）：SP+CXB+QPB 的同号线性相加 +0.142 keV 比 0.10 keV 的门还大，所以该对照按阶段 5 在阶段 1b 产品上实测的那一条组合臂来判；如果它把 ΔkT 移动 ≥ 0.10 keV，判定一律封顶为 AMBIGUOUS。红队：RT-23/RT-24/RT-25/RT-26/RT-27/RT-28 已分诊（R028–R033，D69–D73），共 28 支已分诊；RT-29（阶段 1b 门）与 RT-30（Q-A 策略）运行中（待定）。开问题表新增三条建议（未经评审）：P19 Chandra 匹配扇区重拟合、P20 as-run 能段 mask 阶梯重拟合、P21 钉住 v18 世代后重跑 Angle1。过程：冷启动演练 4 运行中；阶段 1b 仍在出口径（8 条并行通道）。
- 2026-10-18 01:04Z (review correction): Q-B option (c) = P26 is withdrawn (R061/D100) — the red team RT-53 found that sky_background_scale also multiplies the source APEC, and the sky-normalisation ×0.8/×1.2 controls required by D55 (the b: family, which scale the sky components only) have already run: on sky B ΔkT moves by ≤0.052 keV, on sky A the S4 ×1.2 control moves it by −0.114 keV. RT-51 final: P19's XMM-only kT is biased cool in direction, but "model B = 0.71 keV" is not a clean number (α is 5.07σ above the 9–12 keV anchor, MWhalo kT sits on its 0.4 wall) and its "agreement" with stage 5 is not independent; Chandra-only stays Provisional. The real options for the PI are now (a) accept AMBIGUOUS, (b) add MOS1+pn0 / widen to 8–12′ (feasibility being checked by RT-54), (d) more exposure (not available).
- 2026-10-17 23:27Z (stage 25, refresh 6): P19 (the Chandra matched-sector re-fit) is closed (E053/T177/R057/R059/D97/D99, the executor ran 18.6 h) — the Chandra-only kT is 0.690 / 0.755 / 0.747 / 0.858 keV (R2-4…R8-10) and kT(8–10′) − kT(2–4′) = +0.168 keV at only 1.26σ → the verdict is NO POWER, neither corroborating nor refuting the rise; the apertures cover only 35–62 % of the nominal annuli (the report's "79–100 %" is an XMM/Chandra area ratio, not coverage → D69's matched-aperture objection still stands) and the minimum detectable ΔkT is 0.30–0.33 keV (the 1.0762 keV is an artefact of a 0.50–1.00 keV shape-grid clip and is never quoted). P19's XMM-only kT (0.20–0.32 keV) is retired: it is a background-construction artefact — adding the sky components moves R2-4 from 0.316 to 0.710 keV on the same data, in conflict with stage 5's 0.75/0.94 keV; its interpretation is being attacked by RT-51. Red-team census: RT-01…RT-52, 49 teams triaged (a snapshot as of 2026-10-17 23:27Z, now out of date; the current census is RT-01…RT-60 with 55 triaged, see the newest entry on this page) — RT-46 → R055/D95 (offsets-as-sky withdrawn: the sky reference is already 0701 and the offset annuli are 3–21× weaker), RT-47 → R056/D96 (closure steps +0.94σ / +0.38σ, total +0.78σ; "morphology ratios <5 %" holds only at 2–5′), RT-49 → R058/D98 (the information gain 1.44/1.91 is the square of the σ ratio; v18's
sky_background_scaleis frozen at 1.0 and has never been fitted), RT-48 and RT-50 → R059/D99 (P19's final report); RT-51 and RT-53 are running (RT-52 is triaged → R060) (all three were later triaged: RT-51 → R061/D100, RT-53 → R061/R062/D100/D101, and also RT-54 → R062/D101, RT-55 → R063/D102, RT-56 and RT-57 → R064/D103). Q-B's four options are with the PI (D98, awaiting the reply): (a) accept AMBIGUOUS; (b) add MOS1+pn0 / widen to 8–12′, MDΔkT 0.31–0.39 keV, unable to reach the 0.10 keV gate; (c) WITHDRAWN: the P26 sky-normalisation study (R061/D100) — sky_background_scale also multiplies the source APEC, and the sky-normalisation ×0.8/×1.2 controls required by D55 (the b: family, which scale the sky components only) have already run: on sky B ΔkT moves by ≤0.052 keV, on sky A the S4 ×1.2 control moves it by −0.114 keV, all inside the 0.11–0.20 keV gate-control spread; (d) more exposure: not available. Stage 5's verdict stays AMBIGUOUS. Q-C unchanged: no ≥2σ radial rise, the absolute ratio undetermined, closure steps +0.94σ / +0.38σ (total +0.78σ), and R11-13 mos2 is a 0.54σ non-detection. - 2026-10-17 20:58Z (stages 23/24, finished by the dispatcher): RT-46 and RT-47 are triaged (R055/D95, R056/D96) and the home and state pages no longer say "running"; P23's "first fix the sky from the offset pointings 0501/0601" is withdrawn, and the Q-B options go to the PI only after RT-49 has checked them; the closure steps now read +0.94σ / +0.38σ (total +0.78σ), and "morphology ratios move <5 % under a QPB change" holds only at 2–5′. Red teams RT-48 (the P19 XMM leg) and RT-49 (the Q-B options, and whether sky field 0701 is clean) are running → pending.
- 2026-10-17 19:41Z (stage 22: the reader-level red team RT-45's 6 defects fixed + the new R050–R053 results written in + a new group-meetings page): RT-45 (this handbook used as a reader would) is triaged → R054/D94, and all 6 defects are fixed — (1) the home page's plan table still showed the finished P21 (14:34Z, E050/T175/R049/D89) as "running; result pending"; it now reads done and carries the 0.929/1.477/1.369 table; (2) the same table still showed the accepted P22 and the triaged stage 5b as "running → pending" with superseded 3/9 interim numbers; it now reads accepted at 11:04Z (9/9 apertures, R 0.884–1.005) and triaged at 12:48Z (verdict unchanged, sky B ΔkT +0.1982); (3) "one red team died at startup" is unsupported (the queue and 08_RUNNING both record RT-44 as launched at 13:40Z and active), and the red-team ledger now reads RT-01…RT-45 with 43 triaged and RT-46/RT-47 running; (4) the 13′ bump is no longer written as a bare 1.37 but as the record mandates, "Angle2-only, 1.0–2.0 depending on bin pairing/generation, no flux errors in any generation" (I065's RT-01 amendment); (5) the cited
cognition-os/CLAUDE.mddoes not exist and is replaced by the repo-rootCLAUDE.md; (6) physics.html's teaching example was off by 10×/10×/≈2000× and is recomputed from the page's own inputs to 9.2×10⁻⁶⁷ cm⁻², 6.5×10⁶⁰ cm⁻³, 2.35×10⁶⁸ cm³, n_e ≈ 1.7×10⁻⁴ cm⁻³ (D = 9.55 Mpc = 2.95×10²⁵ cm, norm 6×10⁻⁶), with "a typical halo-gas density" reworded to hold for the correct value. New results (all from cognition-os; no number quoted beyond its record): P21's construction is Questioned — at R7-9 the two cameras' per-camera closures are 1.311 (mos1) vs 0.737 (mos2) and the published 1.018 is only a plain mean of two scatters → the argument "closure 1.018 confirms a single, correct sky subtraction" is withdrawn; P24's replay shows the difference is 0.574 ± 0.511 = 1.1σ, i.e. Poisson scatter amplified by the net/gross ratio (net is only 13–21 % of gross), not a calibration disagreement (the "cameras disagree" wording is withdrawn too); the absolute image/v18 ratio is still undetermined (sky model vs empirical sky moves R7-9 from 0.93 to 1.77); the verdict "no ≥2σ radial rise" stays Provisional. P25: the stage-2 soft-band QPB image is correctly normalised (equal to the FWC rebuild within 1.4–4 % and carrying the Al-K/Si-K instrument lines; the 2.13× is "lines in vs lines out"), E042's morphology ratios move <5 % → the morphology claims stand, but stage-2 absolute surface brightness stays Questioned (±50–85 % of its net). B029: P21's script used the mos1 line set for mos2 (negligible at 0.7–1.0 keV). P19 is still running; RT-43 is triaged as interim (R050/D90): the Chandra MDΔkT is quoted only as robust ≈ 0.30 keV (the 1.08 keV is a tail artefact) and the Chandra-only kT is Provisional, not quoted as the halo profile. Q-B: stage 5b is done (R046), P23 remains a proposal awaiting the PI, and RT-46 is testing the offsets-as-sky idea → pending. A new group-meetings page meetings.html (a card at the top of the home page, and a navigation entry on every page): one block each for 2026-10-15/16/17, split into results / work / problems / PI decisions pending, every result carrying its trust state, and the problems column includes that day's agent errors and corrections (P21's unit error B028, the line-set slip B029, P24's erratum, RT-44/RT-46 dying at startup, the withdrawn "cameras disagree" overstatement, and the handbook's own 6 defects). - 2026-10-17 15:23Z (stage 15/15e: stage 5's verdict AMBIGUOUS, P20/P21/P22 and stage 5b, red teams RT-37…RT-42 all triaged): Q-B's pre-registered verdict is in: AMBIGUOUS (underpowered), not "no rise" — ΔkT = kT(R8-10) − kT(R2-4) (target 0): the spectral fit S4 sky A +0.13 (−0.21…+0.42), sky B +0.19 (−0.004…+0.42), the Fe-L hardness H +0.01 (−0.32…+0.46) / +0.07 (−0.13…+0.33); the minimum detectable ΔkT's 0.43 keV is only the Gaussian approximation, and under the pre-declared non-parametric definition (50 % power) both methods exceed 0.50 keV (R042/D82). The QPB is rebuilt: the delivered back.pi was 1e6–1e7× too large (B026) → P22's regeneration was accepted at 11:04Z (9/9 apertures pass the magnitude gate, R 0.884–1.005) → stage 5b's re-run on it was triaged at 12:48Z: the verdict is unchanged, sky B's ΔkT +0.1886 → +0.1982 (−0.034…+0.478), and the QPB-construction systematic is 0.010 keV (R046/D86). P20 (the as-run-band mask ladder): masking only extended sources + stars still biases R2-4_Angle1's kT up by +0.287 / +0.279 keV → D64 confirmed and strengthened, mechanism still Questioned. P21 (one pinned v18 generation): the image/v18 ratio shows no ≥2σ rise with radius (0.929 / 1.477 / 1.369), and the archived 3.32 came from an unreplayable generation; the absolute level differs from stage 2 by ≈2.25× and is unreconciled. RT-41: the folded map's +2.19σ is the maximum over 13 post-hoc split radii and is not significant; the pre-declared statistics peak at +1.78σ. RT-42: P23's information gain is 1.44/1.91, not 2.66, and the gate controls themselves move ΔkT by 0.11–0.20 keV → more power alone does not give RISE; P23 is not yet approved by the PI and its shape was re-ordered to measure the sky from the offset pointings 0501/0601 first.
- 2026-10-17 05:08Z (stage 14: this handbook revised against the 9 defects found by reader-level red team RT-32, R037/D77): the handbook was used as a reader would use it, and all 9 defects are fixed — (1) the §3.1 environment block is replaced by the invocation that actually runs (CIAO 4.18:
source /home/huangrui/software/ciao-4.18/bin/ciao.sh,HEADAS=/home/huangrui/software/ciao-4.18/spectral,XSPEC_MDATA_DIR=.../modelData/, Python 3.12.12 + sherpa 4.18.0); the old block (heasoft + the conda env's zero-byte ciao.sh) would have failed for a reader who copied it; (2) the claim thatrun_v18_fit.shis "identical in content" is dropped, and the script is now described with the fact that its line 7 sources a different ciao.sh, i.e. not the same environment; (3) the state page's R4–6 error 0.041 → the conservative 0.104 keV (0.730 ± 0.104, matching the CSV and the page's own prose, sourcefigure/radial_thermo_profiles_v18/radial_kT_profile_v18_sector_mean.csv); (4) the page now says R8–10 Angle1's saved fit is degenerate and that R8–10/R10–12/R12–14 are single-sector (provisional_single_sector), so the outer profile rests on one sector; (5) the red-team ledger and stage-1b status are updated: RT-29 → R034, RT-30 → R035, RT-31 → R036, RT-32 → R037, RT-33 running; stage 1b is delivered (9/9 apertures, MANIFEST 506/506 OK, real BACKSCAL, REG identity PASS), but the deliveredmos2U005-back.pi(QPB) is 1e6–1e7× too large in all 9 apertures (found by the stage-5 executor; gate 2 never checked magnitudes) → stage 5 runs on the validated repair back.pi (derived-QPB) while regeneration P22 runs → pending, and this handbook quotes none of its numbers; (6) Q-A's "MOS2-only is the inner reference" holds only for kT in [0.45, 1.0] keV (R036/D76); (7) design.html's "still 0.95 after mask v2" becomes ≈0.37 (relative to the v18 local halo surface brightness; 0.95 holds only relative to the v2 halo law, which is 2.59× fainter than v18, citing E035's RT-05 amendment); (8) a new section "which M104 distance each figure on this page uses" lists 11.3 / 8.9 / 9.55 Mpc per figure with its source and warns that the numbers are not comparable across figures until the distance is pinned; (9)ktmap_v2_pi_summary.pngis taken down (a display defect drawing the East arrow toward West, with sha256 proving the embedded copy was the pre-fix one) — no fake files are created. Checks:check_bilingual.pyandcheck_console.shall PASS. - 2026-10-16 23:4xZ (stage 13, the R028–R033 / D69–D73 close-out): Q-C's "image/v18 ratio 1.00 → 3.32 rising radially" is withdrawn (RT-28 → R033/D73) — it is not a physical rise: its Angle1 rows are unreplayable (R7-9 has no .fitting session), unreproducible (R9-11 exists in three generations, the CSV 58 % above the replayable session) or non-significant (R11-13 Δstat 1.36; R12-14 Δstat 0.0), while the reproducible Angle2 is flat at 1.26 → 1.18; the image-internal test (E041: R11–13/R9–11 = 0.949 ± 0.029 / 1.029 ± 0.037) still shows no 13′ excess → R030's "no image excess is false" is itself withdrawn, and the image/v18 branch is not testable until one v18 generation per region is pinned (proposed P21: CoPI, needs the PI). The v18 contract narrowed again (RT-27 → R032/D72): v18 loads RASS but never gives it a model → RASS constrains nothing, so "the RASS filter is corrupt" is a smaller symptom; P16's "the band effect on kT is < 1σ" stays Provisional with limits (a post-hoc error side 0.18–0.68σ → up to 0.83σ; the whole frozen sky family moves through a bound-pinned background fit with LocalBubble.kT at its 0.08 floor; 22 free parameters vs 17–43 dof; the saved fits sit far from their minima, Δstat ≤ 39), and the sky norms are not shown band-robust. D55's combined-stress criterion (RT-26 → R031/D71): the SP+CXB+QPB same-sign linear sum +0.142 keV is larger than the 0.10 keV gate, so that control is judged on the measured stage-5 combined arm on the stage-1b products; if it shifts ΔkT by ≥ 0.10 keV, the verdict is capped at AMBIGUOUS. Red teams: RT-23/RT-24/RT-25/RT-26/RT-27/RT-28 are triaged (R028–R033, D69–D73) — 28 triaged in total; RT-29 (the stage-1b gate) and RT-30 (the Q-A strategy) are running (Pending). The open-problems table gains three suggestions (not reviewed): P19 a Chandra matched-sector re-fit, P20 an as-run-band refit of the mask ladder, and P21 an Angle1 re-run once one v18 generation is pinned. Process: cold-start drill 4 is running; stage 1b is still producing apertures (8 parallel lanes).
- 2026-10-16 20:38Z（阶段 12：E047 措辞按 RT-24 收窄 + RT-23/RT-25 分诊）：P18 端到端模拟的功率必须按统计量分开引（E047/T172/R026，Provisional 暂定；R029/D70）——那个 “0.70 → 0.90 keV 的向外升在 93.5% 的实现里以 ≥ 2σ 被探测到” 说的是全角度 2–12′ 斜率；与真实折叠图上逐楔 2–8′ 检验对应的统计量功率只有 27–36%（大轴/中轴/小轴楔 0.34/0.355/0.27）→ 真实图“没看到”这种上升，并不排除单个楔里存在它；93.5% 在修掉一次重复计算 vignetting 的归一化错误之前是 62%。模拟真值的折叠网格与拟合器逐位相同 → 响应误差未测。S7b 在逐点 99%（Holm 后 97%）上被否定，但真实数据的盘 − 小轴差（−0.030 keV）与 S7b 自己的（−0.036）几乎相同 → 被否定的只是 S7b 那条径向律，不是冷却流这个大类；S7a/S7c 不能被稳健排除（S7a 的 p = 0.045/0.065/0.075，随背景 −10/0/+10% 而变）。两相 S6 的单温 Fe-L 偏差按臂而变：+0.067 keV（sol）vs +0.172 keV（独立臂） → 引作范围。红队：RT-23（Q-B 的 Chandra 半边）已分诊 → R028/D69：Chandra 2–10′ 的“环”其实只有 35–62% 的部分环带（8–10′ 的两个 ObsID 各盖对面一半），共同 kT 检验探不到 ΔkT ≲ 0.27 keV 的差 → “Chandra 等温”不能用来反对 XMM 的外升；RT-24（E047 模拟）与 RT-25（mask/强度）已分诊 → R029/R030/D70；RT-26（OS 一致性）与 RT-27（拟合能段/P16）运行中（待定）。mask（R030）：D64“保留全套 v2 mask”的判定不变，但“不用全套点源 mask 会偏 +0.2 keV”这条是在 0.4–3.2 keV 上量的，而 as-run 拟合 notice 的是 0.375–9.0 keV；被 mask 挡掉的光主要是 M104 自身的光与恒星 → 机制待定，待 as-run 能段重拟合。
- 2026-10-16 20:38Z (stage 12: the E047 wording narrowed after RT-24 + the RT-23/RT-25 triage): the P18 mock's power must be quoted per statistic (E047/T172/R026, Provisional; R029/D70) — the "0.70 → 0.90 keV outward rise is detected at ≥ 2σ in 93.5 % of the draws" refers to the all-angle 2–12′ slope; the statistic matching the real folded map's per-wedge 2–8′ test has a power of only 27–36 % (major/middle/minor wedge 0.34/0.355/0.27) → the real map's failure to see such a rise does not exclude it inside a single wedge; the 93.5 % was 62 % before a double-counted vignetting normalisation was fixed. The mock truth's folding grid is bit-for-bit the fitter's → response error is untested. S7b is disfavoured at 99 % pointwise (97 % after Holm), but the real data's disk − minor difference (−0.030 keV) is essentially the same as S7b's own (−0.036) → what is rejected is the S7b radial law, not cooling flow as a class; S7a/S7c cannot be robustly excluded (S7a's p = 0.045/0.065/0.075 as the background moves −10/0/+10 %). The two-phase S6 one-temperature Fe-L bias is arm-dependent: +0.067 keV (sol) vs +0.172 keV (independent arm) → quote it as a range. Red teams: RT-23 (the Chandra half of Q-B) is triaged → R028/D69: the Chandra 2–10′ "rings" are only 35–62 % partial annuli (the two 8–10′ ObsIDs cover opposite halves) and the common-kT test cannot detect a ΔkT below ≈ 0.27 keV → "Chandra isothermal" cannot be used against the XMM rise; RT-24 (the E047 mock) and RT-25 (mask/intensity) are triaged → R029/R030/D70; RT-26 (OS consistency) and RT-27 (fit band / P16) are running (Pending). Mask (R030): the D64 verdict "keep the full v2 mask" stands, but the "+0.2 keV bias without the full point mask" was measured on 0.4–3.2 keV while the as-run fit notices 0.375–9.0 keV; the light the mask removes is mostly M104's own light and stars → the mechanism is open (Pending), awaiting a refit on the as-run band.
- 2026-10-16 14:2xZ（阶段 10：折叠 2D kT 图 v2 + v18 合同更正）：PI 12:2xZ 要求的四象限折叠 2D 温度图 v2已出（E046/T171，Provisional 暂定，首页与状态页 §2h 六图）——四象限关于光学长/短轴折叠（每箱约 4 倍计数，对称性经 N−S/E−W 两半检验，1–2 个 5–7′ 30–60° 箱不安全 → 天空面板只叫"对称模型展示"）；2–12′ 的逐箱 kT 中位带为 0.70–0.80 keV（逐箱范围 0.53–0.89 keV，全局 0.736 keV，8/21 箱在带外——措辞按 RT-20/D68/R025 收窄），全角度斜率 +0.009 keV/角分（95 % [−0.003, +0.020]，+1.5σ）→ 对这种单温代理而言无 ≥ 2σ 的内冷外热；每 5′ 升 0.1 keV 的梯度在 E046 的单温代理注入里 91% 的次数能被找回（仅这一个温代理的自检；与真实逐楔 2–8′ 检验对应的统计量在 E047 端到端模拟里功率只有 27–36%，所以这 91% 不是探测功率）；盘楔外箱 0.85–0.88 vs 小轴楔 0.61–0.77 keV 只是提示（≈ 1.4σ，pn0 主导）；MOS2 与 pn0 逐箱一致（p = 0.60）；pn0 QPB 按 M104 自己的 0.772 ± 0.016 缩放（测试用）把斜率推到 ≈ 2σ → 趋势对背景敏感；< 2′ 核被 v2 mask 遮住，冷却流在此不可检验（D66 定义 + P18 模拟，待定）；E045 的内区 N−S +0.10 keV 提示经 RT-19 多重比较后降为 Questioned（R022）。v18 合同更正（B025/R023，RT-18 重放）：v18 实际拟合源 0.375–9.0 keV、背景 0.375–6.23 keV（脚本注释的 0.4–3.2 keV 是错的，Sherpa notice/ignore 取并集）；RASS 过滤器损坏；P16 测量其对已存 kT 的影响（待定，新增开问题行）；mask 阶梯 +0.18…+0.22 keV 数字按 0.4–3.2 keV 测量重新定界（R021），但"保留全套 v2 mask"的判定（D64）不变。红队 19 支已分诊，RT-20（攻击折叠图）/RT-21（B025 后的阶段 5 规格书）运行中。
- 2026-10-16 14:2xZ (stage 10: the folded 2D kT map v2 + the v18-contract correction): the four-quadrant-folded 2D temperature map v2 the PI asked for at 12:2xZ is out (E046/T171, Provisional; home page and state §2h, six figures) — the four quadrants folded about the optical major/minor axes (≈4× counts per bin; the symmetry was tested with N−S/E−W halves and 1–2 cells at 5–7′ 30–60° are unsafe, so the sky panel is called only a "symmetric model display"); the per-bin kT over 2–12′ has a median band of 0.70–0.80 keV (per-bin range 0.53–0.89 keV; global 0.736 keV; 8/21 bins outside the band — wording narrowed per RT-20/D68/R025) and the all-angle slope is +0.009 keV/arcmin (95 % [−0.003, +0.020], +1.5σ) → no inner-cool/outer-hot trend at ≥ 2σ for this one-temperature proxy; a gradient of +0.1 keV per 5′ is recovered 91 % of the time in E046's own one-temperature-proxy injection (that proxy's self-test only; the statistic matching the real per-wedge 2–8′ test has a power of only 27–36 % in the E047 end-to-end mock, so this 91 % is not a detection power); the disk wedge's outer cells 0.85–0.88 vs the minor wedge 0.61–0.77 keV are suggestive only (≈ 1.4σ, pn0-driven); MOS2 and pn0 agree bin by bin (p = 0.60); scaling pn0's QPB to M104's own 0.772 ± 0.016 (test-only) pushes the slope to ≈ 2σ → the trend is background-sensitive; the core inside 2′ is masked by v2, so a cooling flow cannot be tested here (the D66 definition + the P18 mock, pending); E045's inner N−S +0.10 keV hint is downgraded to Questioned after RT-19's multiplicity check (R022). The v18-contract correction (B025/R023, replayed by RT-18): v18 as run fits source 0.375–9.0 keV and background 0.375–6.23 keV (the script comment's "0.4–3.2 keV" is wrong — Sherpa's notice/ignore take a union); the RASS filter is corrupt; P16 measures its effect on the saved kT (pending, a new open-problem row); the mask ladder's +0.18…+0.22 keV number was measured on 0.4–3.2 keV and is re-scoped (R021), but the "keep the full v2 mask" verdict (D64) stands. 19 red teams triaged; RT-20 (attacking the folded map) and RT-21 (the stage-5 specification after B025) are running.
- 2026-10-16 11:34Z（第三次刷新 + 修正轮 1）：Q-B 只由 D55 预注册的匹配合同测试判定，但暂停——阶段 1 口径谱缺 v2 mask（B023，执行器在任何拟合前抓住）→ 阶段 1b 重抽（内容门已过，引擎恢复合同不允许复用 B3 → 全新整跑）；Chandra-only 与 0.725 keV 等温一致（D56；D61 更正：ACIS-I 全前照芯片，非 S3；污染模型 ≤ 0.012 keV）；E040 事件级复现（ARF +4%）；2D 强度阶段 2 + 注入检验（E042/R016：2–5′ 与 β 椭圆率一致，≥+30% 双锥排除）；2D kT 阶段 3（E043/R018：E3 + cone45，双臂一致，共同足迹 56%）；mask 阶梯"轻 mask"结论被 RT-17 推翻（R020/D64：v18 无自由源幂律 → 全套 v2 mask 保留、阶段 B 取消；B024：两次从声明而非最终状态写前提）；Z = 0.3 不是外升成因（D63：Z = 1.5 真值只低 0.017 keV）；文献表（Li+2007/2011、2026 SZ、Igarashi+2014：平 kT 无法区分束缚气体与慢风）；红队 17 支已分诊，RT-18（v18 合同审计）/RT-19（生产图）运行中。
- 2026-10-16 11:34Z (third refresh + fix round 1): Q-B is decided only by the D55 pre-registered matched-contract test, but it is paused — the stage-1 aperture spectra lack the v2 mask (B023, caught by an executor before any fit) → stage 1b re-extracts them (content gate passed; the engine's resume contract admits no lane reusing B3 → a fresh full run); Chandra-only is consistent with isothermal 0.725 keV (D56; D61 corrects: ACIS-I, all front-illuminated chips, not S3; contamination model ≤ 0.012 keV); E040 event-level reproduction (ARF +4 %); 2D intensity stage 2 + injection test (E042/R016: 2–5′ consistent with β ellipticity, ≥ +30 % bicone excluded); 2D kT stage 3 (E043/R018: E3 + cone45, both arms agree, common footprint 56 %); the mask ladder's "lighter mask" conclusion was overturned by RT-17 (R020/D64: v18 has no free source power law → the full v2 mask stays, stage B cancelled; B024: two premises written from declarations rather than final states); Z = 0.3 is not the cause of the rise (D63: a Z = 1.5 truth reads only 0.017 keV low); the literature table (Li+2007/2011, the 2026 SZ paper, Igarashi+2014: a flat kT profile cannot separate bound gas from a slow wind); 17 red teams triaged, RT-18 (the v18-contract audit) / RT-19 (the production maps) running.
- 2026-10-15 13:09Z（首次 8 小时刷新）：Q-A 阶段 2 裁定内区参考 = MOS2 单相机（D48/I067），外区/边缘/2D 未合格，阶段 3 运行中；Chandra-only profile（E037）与 J-linked profile（T155）每 ≤1σ 一致；亮度图完成；2D 温度图按 D49 转入代理探索；红队 RT-01/RT-02 启动。
- 2026-10-15 21:49Z（第二次刷新）：Q-A 阶段 3（S4 外区/边缘未合格，测试偏差 ≤ 0.05 keV → D51 探索性引用）；阶段 4（8–10′ 张力 = 合同不匹配，I070/D52）；Fe-L 硬度第二方法平温（I068，RT-06 修订：单温反演平，等温性未确立）；2D 代理阶段 2（E038 + RT-07：校准区间 ≲0.08–0.11 keV、N>S 仅提示）；红队 6 支已分诊（RT-01 存活、RT-03 J-linked 非独立、RT-05 mask 61% → D53）；外区 15–45′ profile Questioned（E039）；Virgo 阶段 1 运行中。刷新期间新增（D54，21:55Z）：RT-08 撤回 I070 主表述（→ Questioned）、RT-09 修订 ±0.05 keV 口径（逐拟合 68% ≈ ±0.15 keV、组合压力未测）；RT-04/RT-10 启动；0501/0601 ODF 已下载校验。
- 2026-10-15 13:09Z (first 8-hour refresh): Q-A stage 2 sets the inner reference to MOS2-only (D48/I067), outer/edge/2D unqualified, stage 3 running; the Chandra-only profile (E037) and the J-linked profile (T155) agree within 1σ per ring; the intensity map is done; the 2D kT map moves to proxy exploration per D49; red teams RT-01/RT-02 launched.
- 2026-10-15 21:49Z (second refresh): Q-A stage 3 (S4 outer/edge not qualified, tested biases ≤ 0.05 keV → D51 exploratory quoting); stage 4 (the 8–10′ tension = contract mismatch, I070/D52); the Fe-L hardness second method is flat (I068, RT-06 amendment: the one-T inversion is flat, isothermality not established); 2D proxy stage 2 (E038 + RT-07: calibrated bounds ≲0.08–0.11 keV, N>S a hint only); 6 red teams triaged (RT-01 survives, RT-03 J-linked not independent, RT-05 mask 61 % → D53); the outer 15–45′ profile Questioned (E039); Virgo stage 1 running. During the refresh (D54, 21:55Z): RT-08 withdrew I070's headline (→ Questioned), RT-09 amended the ±0.05 keV wording (per-fit 68 % ≈ ±0.15 keV, the combined stress untested); RT-04/RT-10 launched; the 0501/0601 ODFs downloaded and verified.
- 2026-10-15（阶段 2 修订）：Q-C 13′ 超出改判为假象（I065/B020）——v18 保存的 flux 本就是每 arcmin²，旧 flux/area 双重归一；图与文字全部更正。Q-A 试点结果（无策略认证，T153）、Q-B 搁置与逐仪器路线（D46）、GTI 保留 2.filter（I066）、mask 规则 v2 与"残余必须建模"（E035）、Virgo 再抽谱计划（D47）入表。新增开问题：MOS1 静态硬过剩。
- 2026-10-15 (stage-2 revision): the Q-C 13′ excess is reclassified as an artefact (I065/B020) — the v18 saved flux was already per arcmin², so the old flux/area double-normalised; figure and text corrected throughout. The Q-A pilot result (no strategy certified, T153), the Q-B parking and per-instrument route (D46), keeping GTI 2.filter (I066), mask rule v2 with "the residual must be modelled" (E035), and the Virgo re-reduction plan (D47) entered the tables. New open problem: the MOS1 quiescent hard excess.
- 2026-10-15：手册首次构建（H-XART-001）；PI 新增四问（Q-A…Q-D），四个研究 job 派出。
- 2026-10-15: first handbook build (H-XART-001); the PI added four questions (Q-A…Q-D) and four research jobs were dispatched.
- 2026-07-15：D45 — P0 反馈检验 NO-GO（full-nuisance 区间过零），P1 停止。
- 2026-07-15: D45 — P0 feedback test NO-GO (full-nuisance interval crosses zero); P1 stopped.
- 2026-10-05/06：Chandra+XMM 联合管线停在 8/20 PASS。
- 2026-10-05/06: the Chandra+XMM joint pipeline stalled at 8/20 PASS.
数据截至
Data as of
2026-10-18 13:03Z（cognition-os/sessions/2026-10-14.md 从 22:50Z 起加 sessions/2026-10-16.md 到 23:25Z 与 sessions/2026-10-17.md 与 sessions/2026-10-18.md 的全部条目；07_REVISION_LOG.md R001–R070；04_DECISION_LOG.md D46–D112；03_TRUST_TABLE.md T148–T177；08_RUNNING.md 06:41Z；REDTEAM_QUEUE.md RT-01…RT-69；E040–E053、B022–B029、references/literature_m104_hot_halo_2026-10-16.md；本手册读者级红队 RT-32、RT-45、RT-56 与 RT-67 的 REPORT.hermes.md / REPORT.codexnim.md）。运行中的 job（红队 RT-68、RT-69；任务 P28 与 P30）结果待定，本手册不引它们的数字；已完成并写入本页的：RT-46/R055/D95、RT-47/R056/D96、RT-49/R058/D98、RT-48 与 RT-50/R059/D99、P19（E053/T177/R057/R059/D97/D99）、RT-20/R025/D68、RT-21/D67、RT-22/R027、RT-23/R028/D69、RT-24/R029/D70、RT-25/R030/D70、RT-26/R031/D71、RT-27/R032/D72、RT-28/R033/D73、RT-29/R034/D74、RT-30/R035/D74、RT-31/R036/D76、RT-32/R037/D77、RT-37/R044/D84、RT-38/R041/R043、RT-39/R045/D85、RT-40 与 RT-42/R048/D88、RT-41/R047/D87、RT-43（interim）/R050/D90、RT-44/R051/D91、RT-45/R054/D94、RT-51/R061/D100、RT-52/R060、RT-53/R061/R062/D100/D101、RT-54/R062/D101、RT-55/R063/D102、RT-56 与 RT-57/R064/D103、RT-58 与 RT-59/R065/D104、RT-60 与 RT-61 与 RT-62/R066/D107、RT-63 与 RT-64/R067/D108、P29（R067/D108，措辞按 R070 订正）、RT-65/R070/D112、RT-66/R068/D109、RT-67/D112、P16、P18 sol 臂（E047/T172/R026）、P20（E048/T173/R039/D79）、P21（E050/T175/R049/D89，后续 R051/R052/R053）、P22 与阶段 5b（R045/D85、R046/D86）、P24（E051/R052/D92）、P25（E052/R053/D93）、阶段 5（E049/T174/R040/D80）、冷启动演练 3–8。
2026-10-18 13:03Z (all cognition-os/sessions/2026-10-14.md entries from 22:50Z plus sessions/2026-10-16.md to 23:25Z and all of sessions/2026-10-17.md and sessions/2026-10-18.md; 07_REVISION_LOG.md R001–R070; 04_DECISION_LOG.md D46–D112; 03_TRUST_TABLE.md T148–T177; 08_RUNNING.md 06:41Z; REDTEAM_QUEUE.md RT-01…RT-69; E040–E053, B022–B029, references/literature_m104_hot_halo_2026-10-16.md; this handbook's own reader-level red teams RT-32, RT-45, RT-56 and RT-67, REPORT.hermes.md / REPORT.codexnim.md). The running jobs (red teams RT-68 and RT-69; tasks P28 and P30) are pending, and this handbook quotes none of their numbers; done and written into this page: RT-46/R055/D95, RT-47/R056/D96, RT-49/R058/D98, RT-48 and RT-50/R059/D99, P19 (E053/T177/R057/R059/D97/D99), RT-20/R025/D68, RT-21/D67, RT-22/R027, RT-23/R028/D69, RT-24/R029/D70, RT-25/R030/D70, RT-26/R031/D71, RT-27/R032/D72, RT-28/R033/D73, RT-29/R034/D74, RT-30/R035/D74, RT-31/R036/D76, RT-32/R037/D77, RT-37/R044/D84, RT-38/R041/R043, RT-39/R045/D85, RT-40 and RT-42/R048/D88, RT-41/R047/D87, RT-43 (interim)/R050/D90, RT-44/R051/D91, RT-45/R054/D94, RT-51/R061/D100, RT-52/R060, RT-53/R061/R062/D100/D101, RT-54/R062/D101, RT-55/R063/D102, RT-56 and RT-57/R064/D103, RT-58 and RT-59/R065/D104, RT-60 and RT-61 and RT-62/R066/D107, RT-63 and RT-64/R067/D108, P29 (R067/D108, wording corrected per R070), RT-65/R070/D112, RT-66/R068/D109, RT-67/D112, P16, P18's sol arm (E047/T172/R026), P20 (E048/T173/R039/D79), P21 (E050/T175/R049/D89, followed by R051/R052/R053), P22 and stage 5b (R045/D85, R046/D86), P24 (E051/R052/D92), P25 (E052/R053/D93), stage 5 (E049/T174/R040/D80) and cold-start drills 3–8.