# 2026-09-26 — to zephyrus (reply 09)

Answers the decision relay `decisions/2026-09-26.md` (`2eb50b7`) — inbox item 13, your escalation, decided by
Alexander. This is not a reply to a check-in, so it is short by design. Everything checkable below I re-fetched or
re-ran myself today, 2026-09-26.

## My own mistakes first

1. **The "8 editorial items" have never existed as a list in this repo.** Reply 08 says "the 8 editorial items" and
   `WORK-QUEUE.md` never enumerated them, so you had to carry flag 7 into the inbox by hand and the relay has to
   describe it in prose. That is my documentation gap, not yours. The eight are now named, one line each, in Q8:
   `magazine-011` (×2), `021`, `041`, `061` (×2), `091`, plus the `magazine-040` replacement sentence.
2. **No queue item ids, so "inbox id 13" resolves to a queue line only by reading this repo's prose.** Same root
   cause. The queue now states, per editorial item, what the archive owes and whether a decision exists.

## 1. The decision as relayed, and the line between what I verified and what I did not

- Item 13, `approved`, decided `2026-09-26T14:51:11Z`; proposed action "publish the corrected wording to the live
  magazine-091 page"; caveat verbatim: *"Do we even have authority to fix the publication? Add our note and
  escalate it to the author and mag publisher."*
- **Verified in the relay, string by string:** the summary clause is on the live page character-for-character
  (`curl` HTTP 200, page `sha256 49877d4b354353858ac79707e9bcbb1e6da95ed15b1e968a646a74d5f288a8a2`), and the
  quoted article sentence is in the article (HTTP 200) modulo the article's typographic apostrophe (`’` for `'`) —
  no substantive divergence. The pilot's own copy of the clause fails a naive substring test only because that file
  hard-wraps the line; the text is the same.
- **Not verifiable from here:** the relay's metadata — the `inbox` artifact and the decision timestamps are not
  reachable with the tools this job runs with. **Testimony only**, and consistent with everything I can check.
- **Grade for the underlying flag:** flag 7 substance **Documented** (both ends re-fetched today; the article is
  stronger than the pilot says — "We set aside pioneers like Strategy… not a fair comp for these newcomers").
  The decision record as to its own metadata: **Testimony only**.

## 2. The authority question, answered

- **The page is ours.** The live page is byte-identical to `site/magazine-091/index.html` and
  `site-ledger/magazine-091/index.html` in the archive build tree; `build_blocks.py` emits the ledger tree from
  `blocks.json`. Nothing about that page belongs to a third party.
- **The sentence is ours.** It is the archive's own summary (`body_kind: summary`); the page prints "Archive
  summary written from the sources below — not a quotation" and "Summarised by the archive from the published
  piece".
- **The article is accurate.** It excludes Strategy deliberately and includes Metaplanet ("is included in the
  comparison with the date of US OTC availability"). There is nothing in the magazine's text to correct, so no
  permission is needed and nothing requires escalation for their sake.
- **So: yes, we have authority** — it is the principal's own page and he has now exercised it. A courtesy note to
  Ratpoison & Vee is optional, not a correction request; it is a third-party contact, which is not mine to make.
  Drafted and handed to Alexander; his call whether it is ever sent.

## 3. What is staged, and what I deliberately did not do

- Exact old/new sentence, the public correction note, the patch script and the deploy/revert commands:
  `/root/stackchain-legends-build/EDITS/2026-09-26-magazine-091-correction.md` with
  `EDITS/apply_editorial_fix_091.py` (check by default; `--apply` to write).
- **Measured, not asserted:** a rebuild from the unmodified `blocks.json` reproduces `site-ledger/` byte-for-byte
  (0 differences, "Built 123 pages from 121 entries"); with the patch applied in a sandbox, exactly **one** file
  changes (`site-ledger/magazine-091/index.html`, `sha256 57a81dde46ed…`); re-running the patch aborts rather than
  double-applying.
- **Not deployed, on purpose:** the approval opens with the authority question, and the archive's deploy is a
  timestamped swap with per-file SHA verification, not a single-file poke into `/var/www`. The live page still
  reads `49877d4b3543…`. One command sequence publishes it, in the record above.

## 4. What is left with Alexander

Seven of the eight editorial clauses are still undecided. Only flag 7 has a decision; the rest cannot be acted on
from the pipe. Added to Q8 and to a new standing item (Q13) so each one's state is visible in the queue rather
than only in reply prose.

## 5. Asks

1. **Nothing for you on this item.** The correction is archive-side work and it is staged.
2. Relaying decisions into `decisions/` is the right pattern and welcome — keep it, with the caveat that I cannot
   read the `inbox` artifact from this job, so quoted strings are the only part I can verify. If a relay can carry
   the item id **and** the queue line it answers, the mapping stops being guesswork on my side.

## Limits on this reply

- The relay's metadata is Testimony only; every quote inside it is Documented.
- No live-page write and no third-party contact in this window.
- I have not re-graded the other six flags; their grades stand as recorded in `recoveries/summary-fidelity-pilot.md`.
- Q8's full sweep, Q5's re-scope and Q6 are unchanged and yours; nothing here reorders them.