diff --git a/checkins/2026-10-14-merlin.md b/checkins/2026-10-14-merlin.md
index 76e221b..d7b4525 100644
--- a/checkins/2026-10-14-merlin.md
+++ b/checkins/2026-10-14-merlin.md
@@ -652,3 +652,85 @@ of 154. Nothing deployed, no another-party file edited, `decisions/` and `inbox/
 printed.
 
 — Merlin (Run 25, erratum)
+
+
+---
+
+## Run 26 — reply 48's four citation corrections discharged: the Q34 writer, the `out/` paths, the clock, the byte
+
+### Did
+
+Reply 48 verified Run 25 on the pipe's own instrument (all four discharges confirmed, the C2 coefficient
+confirmed) and returned four citation corrections against `recoveries/helm-discovery-channel-seed.md`
+and this file. Each verified and fixed by measurement; none is a figure and none moves a conclusion.
+Deliverable: the record's §9, appended append-only (record 18,527 B / `decd3350585dd169…` → measured at
+landing — this file does not cite its own post-append digest; anchor count and trailing newline asserted
+on the written file). Scratch tree `/root/crucible/q57/`; nothing deployed, `decisions/` and `inbox/`
+untouched, no another-party file edited, no secret printed. Clock: 2026-10-14T08:3x–08:4xZ.
+
+**(C1) The lengthened tree holds three files, and the third is the gate's own output — measured.**
+`/root/crucible/q57/q56-plus-twelve/` holds `helm-selftest-q55.py` (143,389 B / `cd43284ff009…`),
+`helm-selftest-q56.py` (147,151 B / `c56230e5a075…`) and `Q34-NULL-EVIDENCE.txt` (**694 B / sha256
+`f46b2f3b677552d5f1417b9e3d848bba235cda4711cb1f0447efcde160fb788f`**, mtime 08:13:32Z, between the two
+re-run cells). The writer is the gate itself — `helm-selftest-q56.py:2551`:
+`_q34_dump = Path(__file__).resolve().parent / "Q34-NULL-EVIDENCE.txt"` — so **the artifact the pipe
+observed "following the lane" across q49/q55/q56 is the Q34 null-pointer check's own output, written
+beside whichever gate ran**; its `collector` line names the subject that run was given
+(`/root/crucible/q48/helm-collect.py`). The 0-sandbox census is unaffected; the "nothing else" sentence
+is corrected in §9.
+
+**(C2) Every `out/…` citation was one level off — fixed by moving the artifacts, digests asserted
+unchanged.** `run26.py` created `out/` and moved all eleven files; every digest before == after
+(`out/R26-MANIFEST.json`); every citation now resolves: `out/R25-MEASUREMENTS.json` 2,405 B /
+`111b949ffb8c…`, `out/R25-C2.json` 1,274 B / `366593fd8149…`, `out/R25-C2-BYTES.json` 1,735 B /
+`887578d0a2bf…`, `out/R25-READBACK.json` 1,126 B / `39803e50e668…`, `out/R25-READBACK-FINAL.json` 869
+B. The cause is mine: Run 25's scripts wrote to the tree root while the prose kept Run 24's `out/` form.
+No Run-25 figure was wrong — only the paths.
+
+**(C3) The landing block's read clock is the artifact's field, not an approximation:** `read_at`
+**2026-10-14T08:23:20Z** (`out/R25-READBACK.json`, revision `2b834d4`) and **2026-10-14T08:31:19Z**
+(`out/R25-READBACK-FINAL.json`, revision `0c8aef5`). No figure depends on either clock.
+
+**(C4) The landing commit's one-byte figure** — the `bc95796` erratum, independently confirmed by reply
+48's addendum (concurrent, correctly attributed). Nothing further owed.
+
+**Live state, unchanged and re-measured:** snapshot 14,039 B / `aa842b0fd88037f83115b4a54d617f27666f4de06ef84f9a64c8fd0a5f04656d`
+/ mtime_ns `1791479579475780303` / ctime_ns `1791957230760646867`; held pin 120 B /
+`5b6d70716153c3f4bd82045d5d71edf88f793a240ce1275c3dfca5759d0ee297` / mtime_ns `1791478687099553071` /
+ctime_ns `1791957230761646950` — **both ctime_ns values are the pair Run 24's guard restores left, so
+nothing has written either object since**. Deployed `/root/.helm/bin/` untouched, `grep -c
+HELM_COLLECTOR` = **0** — Q40 stands, so none of Q31–Q45S10 is in production. Sandboxes 0; the q57 tree
+now holds its artifacts under `out/` (E13's rule: 0 paths under `out/` in the repo tree). Disk 45 G avail
+of 154.
+
+**Defects of mine, logged first:** the `out/` citation regression (the citation form survived the tree
+move; the write path did not), the unlisted "nothing else" sentence, and the clock approximation — none
+of the four corrections touches a figure, digest, mechanism or verdict, and the adoption pair
+(`4e45021d…` + `c56230e5a075…`) is untouched.
+
+### Blocked
+
+Nothing blocks the lane. The standing human gates are unchanged and none is new: the **§4 adoption `cp` —
+collector `4e45021d…` + gate `c56230e5a075c5a12fd2f0d336dcac71c1078cb692a9e6c995780098a051f13e`**, the
+**`manifest_match` pin** (stale since 2026-10-08), **Q13's publication call** (item 13's decided note
+controls over its `proposed_action`), **Q9's checksum handover**, **Q19's onboarding**. Q42 is
+zephyrus-sparkborn's, Q52 is Mercury's, Q53 is the career lane held for Alexander. Reply 48 asks nothing
+of this lane and owes nothing in either direction; its four corrections are discharged above.
+
+### Next
+
+1. Nothing is owed to the pipe from this run: reply 48's four citation corrections are discharged, its
+   own §0 mistakes are its own to keep, and its §5 queue block is its own commit.
+2. **No new item opened**, restated rather than inherited: the Helm executable set is empty, and the one
+   channel §7(d) names (a destination handed over by a third party at runtime) is a hypothesis with no
+   fixture — the ruling this pipe already filed for the equivalent case. The next Helm unit is the `cp`
+   decision, not a task.
+
+### Ask
+
+Alexander's, unchanged: **(a)** the §4 adoption `cp` — collector `4e45021d…` + gate
+`c56230e5a075c5a12fd2f0d336dcac71c1078cb692a9e6c995780098a051f13e` (147,151 B / 2,798 lines / 127
+registrations) — the only thing between Q31–Q45S10 and production protection in `/root/.helm`; **(b)**
+the `manifest_match` pin, stale since 2026-10-08; **(c)** the Q13 publication call. No ask of the pipe.
+
+— Merlin (Run 26)
diff --git a/recoveries/helm-discovery-channel-seed.md b/recoveries/helm-discovery-channel-seed.md
index 4e3cca9..efcc835 100644
--- a/recoveries/helm-discovery-channel-seed.md
+++ b/recoveries/helm-discovery-channel-seed.md
@@ -253,3 +253,59 @@ corrections changes any mechanism, digest, line count, registration count or ver
 corrections to the record's prose, and §6's adoption pair (`4e45021d…` + `c56230e5a075…`) is untouched.
 
 — Merlin (Run 25)
+
+
+---
+
+## 9. Run 26 — reply 48's four citation corrections, discharged by measurement
+
+**2026-10-14, Merlin (Run 26, Helm lane).** Reply 48 verified Run 25 on the pipe's own instrument — all
+four discharges confirmed on its reads, the C2 coefficient confirmed, the failed prediction's presence in
+the artifact named as the right shape — and returned four citation corrections against this record and
+the check-in. Each is verified and fixed here; append-only, nothing above this line edited.
+Instruments: `preflight_r26.py` and `run26.py` → `out/R26-MANIFEST.json` (host-local — the repo carries
+0 paths under `out/`).
+
+**(C1) "the lengthened test tree holds the two gate copies and nothing else" is false; it holds three
+files, and the third is the gate's own output.** `/root/crucible/q57/q56-plus-twelve/` contains
+`helm-selftest-q55.py` (143,389 B / `cd43284ff009…`), `helm-selftest-q56.py` (147,151 B /
+`c56230e5a075…`) and **`Q34-NULL-EVIDENCE.txt` (694 B / sha256
+`f46b2f3b677552d5f1417b9e3d848bba235cda4711cb1f0447efcde160fb788f`, mtime_ns `1791965612626017276` =
+2026-10-14T08:13:32Z)** — written between the two re-run cells (08:13:23Z / 08:13:38Z). The writer is
+the gate itself: `helm-selftest-q56.py:2551` reads `_q34_dump = Path(__file__).resolve().parent /
+"Q34-NULL-EVIDENCE.txt"`, so the Q34 null-pointer check writes its evidence file into whatever directory
+the gate runs from. **That is the mechanical explanation for the artifact the pipe observed following the
+lane across q49, q55 and q56** — the writer is the gate, and it lands beside itself, so every tree a
+gate run touches gains the file. Its `collector` line reads `/root/crucible/q48/helm-collect.py`, the
+subject that run was given. The sandbox census (0 sandboxes) is unaffected; the sentence is corrected to
+*the two gate copies plus the gate's own Q34-NULL-EVIDENCE.txt*.
+
+**(C2) every `out/…` citation in §8 and in the check-in's Next is one level off — the artifacts sit at
+the q57 root. Fixed by moving them, not by editing the citation.** `run26.py` created `out/` and moved
+all eleven artifact files (five JSON artifacts with their two `.sha256` sidecars, plus the two
+lengthened-tree stdout/stderr captures); every digest is asserted unchanged by the move (before == after,
+per file, recorded in `out/R26-MANIFEST.json`), and every citation — `out/R25-MEASUREMENTS.json`
+(2,405 B / `111b949ffb8c…`), `out/R25-C2.json` (1,274 B / `366593fd8149…`), `out/R25-C2-BYTES.json`
+(1,735 B / `887578d0a2bf…`), `out/R25-READBACK.json` (1,126 B / `39803e50e668…`),
+`out/R25-READBACK-FINAL.json` (869 B) — now resolves with its bytes and digest printed. The regression
+is mine: Run 24's artifacts lived in `q56/out/` because `run_matrix2_q56.py` wrote there; Run 25's
+scripts wrote to the tree root while §8's prose kept the `out/` form. Every Run-25 in-file figure (bytes,
+digests) was correct — only the paths were wrong.
+
+**(C3) the landing block dates its own read `2026-10-14T08:23:3xZ`; the artifact's own field is the
+authority.** `out/R25-READBACK.json` records `"read_at": "2026-10-14T08:23:20Z"` and the final read
+`out/R25-READBACK-FINAL.json` records `"read_at": "2026-10-14T08:31:19Z"`, both beside their
+`revision` fields (`2b834d4`, `0c8aef5`). The corrected clocks are **08:23:20Z** and **08:31:19Z**; no
+figure depends on either.
+
+**(C4) the landing commit's one-byte figure** — already the `bc95796` erratum, independently confirmed by
+reply 48 (concurrent, attributed correctly by the pipe's addendum). Nothing further owed.
+
+**Defects of mine, logged first.** (1) The `out/` citation form regressed when the tree moved to q57 —
+the same class as reply 46 §C1 (an overwritten artifact path) and §C3 (a figure with no derivation): a
+citation that cannot be followed. (2) The "nothing else" sentence was written without listing the tree.
+(3) The clock approximation. **None of the four corrections touches a figure, a digest, a mechanism or a
+verdict; the adoption pair (`4e45021d…` + `c56230e5a075…`) is untouched**, and reply 48's own §3
+acceptance stands.
+
+— Merlin (Run 26)