# ==🗄 Master Regimen · **ARCHIVE** · companion to rev 7.25==

*Version-date: **2026-07-23** · v6.21 / rev 7.25*

> **This is the archive.** Revision history, derivations, prior adversarial-pass conclusions, and the analysis behind the numbers.
> ⚠️ **Not an input for a fresh audit.** Prior-model analysis left in a brief anchors the review that follows and goes stale silently. When a question is genuinely open, feed the **PROTOCOL** file, not this one.

**#regimen #archive**

---

## ::✅ What changed 7/23 (rev 7.22)::

> ### 🟢 **rev 7.24 · the dual-AI rule corrected — see Methodology note. No schedule change.**

- 🔄 **The cross-feeding prohibition is retired in its absolute form.** *It was **⟨C⟩-origin and overbroad**, and Claude enforced it against Nate's own validated workflow, repeatedly, because the settings doc carries the absolute phrasing and is read first every session. **Corroboration** and **audit/synthesis** are two operations; only the first is destroyed by cross-feeding. The second is the primary workflow and it works.* ⟨C·retired → **N·7-23**⟩
  > *Worth naming precisely because of where it sat: **not in the schedule, but in the meta-layer that governs how every other finding gets produced.** A wrong rule there corrupts the method, not one dose. It survived because it sounded like rigor.*

> ### 🟢 **rev 7.23 · ALCAR repositioned · three findings logged as OPEN · the change-accounting badge retired.**

- ✅ **ALCAR 4:00 PM → 1:45 PM.** *Nate's edit, 7/23, on a Claude-proposed rationale the same day — **⟨C·7-23→N✓7-23⟩, not ⟨N⟩.** Afternoon spacing was 4 h / 6 h / **1 h** / 13 h overnight; it is now 4 h / 3¾ h / 3¼ h. The 1:45 PM slot already existed, so no dosing event was added. **Total unchanged at 2,500 mg/day.*** ⚠️ *The premise it rests on — that the 4-way split exists for even exposure — is **untagged in every rev**. The move is safe regardless; the reasoning is provisional. → premise registry.*

- 🔴 **The ProBiota attribution window is not clean, and this file asserted that it was.** *Counted off the schedule in this document: ProBiota HistaminX started **7/21**. **Erinamax relocated into Meal 1 on 7/22** — the same meal, the next day, directly against the "nothing else new enters Meal 1" rule written one line below it. Agmatine changed slot 7/22. ALCAR changed slot 7/23.* **Three schedule changes inside a 14-day window described as single-variable.**
  > *Why it bites: **Erinamax carries a declared milk trace, and the milk endpoint is the phlegm/cough response, which reads in days.** That is the same timescale and the same body system a probiotic readout uses. Any signal before ~8/4 has at least two candidates landing one day apart.*
  > **Nothing needs undoing.** *The fix is to name the confounds before the readout instead of crediting the window to ProBiota by default.* → **open item 19.**

- ⚠️ **The 7/22 agmatine move is reopened.** *"Clears the agmatine↔NaturDAO collision" treats a **window** as an **instant**. A pre-meal DAO tablet works through that meal's transit (~3:00–6:00 PM for the 2:55 PM dose). 2:00 PM sat 55 min **ahead** of it; 4:00 PM lands 60 min **into** the meal and travels **with** it.* **The move may have gone the wrong direction, and the direction was never checked.** *Brief item 2. **Not undone** — the premise is `⟨i:in-vitro/porcine⟩` and cannot support either placement.* → **open item 20.**

- ⚠️ **The 5:00 PM agmatine was never examined and is now the tightest coupling in the day** — 85 min to the 6:25 PM tablet, and sitting inside the 2:55 PM tablet's window. *rev 7.19 fixed a 55-minute gap and left an 85-minute one unnamed, in the same file, in the same afternoon.* → **open item 20.**

- 🏷️ **Two premises registered `⟨?⟩` that had never appeared anywhere:** `alcar-split-rationale` (why ALCAR is four doses) and `agmatine-indication` (why agmatine is in the regimen at all — flagged in prose since 7/19 but never given a registry row). *This is the rev 7.21 blind-spot rule applying to itself: **tag-on-touch only catches rules you touch**, and these were touched today.*

- 📌 **The standing-context settings doc is stale.** *It still carries **soy as a ❌ hard constraint with "stops Creon working"** — the exact wording retired at rev 7.21. Anything read from that doc will re-import a constraint this file has already demoted.* → **open item 21.**

### Method note — what produced this rev

*A verification pass, not a research pass. The finding that matters was not found in the literature — it was found by **counting dates off this document's own schedule**: ProBiota 7/21, Erinamax into Meal 1 7/22, agmatine 7/22, ALCAR 7/23. Every element was already written here, in four separate sections, and no pass had put them in one place.* **rev 7.20's own warning generalises further than it was written: assume guilty until paired applies to dates, not only to blood pressure.**

*Claude-origin caution, stated plainly: **three of the six items above are ⟨C·7-23⟩ and unconfirmed.** The ALCAR move is the only one Nate has acted on. The agmatine findings are arguments, not measurements, and they are logged as open questions specifically so they do not harden into apparent fact the way the B6, the ">35 mmHg fall," and the soy rule each did.*


---

---

## ::📋 What changed 7/23 (rev 7.21) — prior rev, same day::

> ### 🟢 **One constraint reworked, one gate released.**

- 🔴 **The soy hard constraint is REWORKED and DEMOTED from ❌ to ⚠️.** *"Soy stops Creon working" does not survive.* Soybean Kunitz (KTI) and Bowman-Birk (BBI) inhibitors are **serine-protease inhibitors — they inhibit trypsin and chymotrypsin only. They do not inhibit lipase or amylase.** Creon is dosed in **USP lipase units** and titrated on fat absorption. **The ceiling of any soy effect is the protease fraction; the fraction the whole therapy is judged on is mechanically untouched.** ⟨C·7-23⟩
- ✅ **Erinamax soy gate RELEASED on arithmetic.** Label upper bound ≈**5 µg soy protein/day** vs. the **~96 mg/day** pea/lentil load cleared at rev 7.19 — **~19,000× smaller than something already cleared**, and that ceiling assumes every microgram is intact, unhydrolysed inhibitor, which spent fermentation medium is not.
- 🏷️ **Provenance: the soy rule was ⟨?⟩ in every rev.** It appeared in **no** premise registry, carries **no** source, and there is **no soy interaction on the Creon FDA label.** Structural tell: every other Standing Constraint cites its origin — acetaminophen→VM, NSAIDs→ulcer history, potassium→ARB. **Soy stated a mechanism and no source, in the same visual tier, for the entire life of this document.** *Same failure class as the B6/DAO error and the ">35 mmHg fall" — a plausible mechanistic claim wearing authority it never earned.* **Origin still unrecalled; do not assume authorship in either direction.**
- 📌 **Cited literature opened; none of it covers the claim.** PMID 2446949 = adults, ERCP-cannulated, BBI added to collected juice **ex vivo** and reinfused — the finding is a **2–3× *increase* in endogenous output**, i.e. proof that feedback control exists in humans. PMID 1594554 = **Lewis rats with bladder-drained pancreas transplants**, urinary enzyme output. *Foods* 12:1691 = **mice**, purified STI. **No PERT. No EPI. No fat-absorption endpoint anywhere.** Brief item 5, cleanly triggered.
- 🆕 **Erinamax label verified from photographs (7/22).** *"Freeze Dried Lion's Mane (Hericium erinaceus) (mycelium) (5mg Erinacine A) **(as ErinaPrime®)** 1,000 mg"* per 2-capsule serving; Other Ingredients: Vegetable Cellulose; **Contains: Milk, Soy.** **Product identity resolved** — ErinaPrime® is Nammex/Novelara material produced by Grape King Bio (Taiwan). ND and retailer copy still calling Erinamax "allergen-free" is **contradicted by ND's own label.** Trust the label.
- 🆕 **A 2× erinacine A spec discrepancy is now open.** ND label = 5 mg/g (**0.5%**); Nammex publishes ErinaPrime at 10 mg/g (**1.0%**). ND's own wording is *"at least 0.5%"* — **a guaranteed floor, not an assay.** **Actual daily intake is either 2.5 mg or 5 mg. Unknown within a factor of two, and it gates any dose decision.**
- ⚠️ **Citation trap logged before it bites.** A 2023 MDPI *Journal of Fungi* review (9(5):551) restates the Li 2020 mild-AD trial as **"15 mg erinacine A per day."** **Wrong by ~2.9×.** Recompute from the primary: 350 mg = 0.35 g × 5 mg/g = **1.75 mg/capsule × 3 = 5.25 mg/day.** The review appears to have read "5 mg/g" as "5 mg/capsule." *Against a 2.5 mg label floor, the bad figure implies a **6× gap** and makes "take 6 capsules" look reasonable. It isn't.*
- 🥛 **Milk is separated out and stays open.** There is **no Creon dimension to dietary dairy** — milk protein is not a trypsin-inhibitor source. The only milk item on Creon's own label is administration: don't mix capsule contents into milk/formula or any food above **pH 4.5** (Creon releases at ~pH 5.5+). Irrelevant to swallowing capsules whole. **The trace is purely the phlegm/cough avoidance, and it is a supplement-layer call.**
- ✅ **ProBiota HistaminX confirmed STARTED 7/21** ⟨N✓7-23⟩. Reconciled out of the queue. **It is now the live single-variable trial, attribution window 7/21 → ~8/4** — an independent reason nothing new enters Meal 1 before 8/1.

### Methodology note — how this rev was produced

*The Perplexity output was fed to Claude **as an object of review**, not as corroboration.* **⟨rev 7.24: this sentence was correct and was buried. It is now the rule — see Methodology note.⟩** That is a legitimate use and a different operation from cross-feeding — but the finding worth keeping is that **the brief itself contained the answer**: the question was posed as *"since soy negatively impacts Creon's effectiveness,"* so Perplexity was asked to reason **from** the constraint rather than test it, and returned the constraint as its conclusion. **A premise smuggled into the brief produces an echo just as reliably as cross-feeding does.** Add it to the loop-fixing note below.

**Claude was wrong twice in this rev and both are logged:** (1) it inferred manufacturer intent from the FALCPA label vehicle without the label text, and concluded the milk risk was *understated* — **withdrawn**; ND states its meaning explicitly and Perplexity's reading was right. (2) It argued that existing Erinamax tolerance answered the milk question — **partially retracted**, because a pre-June bottle may be different material. *Perplexity was also right on product identity, which Claude had flagged as unverified.*


---

---

## ::🧬 DAO — what actually governs it::

*The histamine-degrading enzyme. Ranked by likely magnitude **in this phenotype**, not in general.*

**1. Substrate competition** ⟨C·7-22 | **i:in-vitro/porcine — NOT human in vivo**⟩

Agmatine (1,000 mg/day, 4 doses) is a **confirmed DAO substrate** — and notably *not* a substrate for MAO-A, MAO-B, or SSAO, so DAO is essentially its amine-oxidase route. Agmatinase converts it to **putrescine**, itself a DAO substrate: two shots at the same enzyme. Crude molar load ≈ **50× dietary histamine** (≈10× on a heavy 50 mg histamine day). Competition among biogenic amines for DAO is a demonstrated in-vitro effect, tested at histamine:amine ratios from 1:0.25 to 1:20.

⚠️ **Caveats stated, not buried:** in-vitro porcine assay; oral agmatine is absorbed and much goes to agmatinase, so luminal concentration ≠ dose. **Order-of-magnitude flag, not a measurement.**

✅ **Cheapest test taken 7/22:** moved 2 PM agmatine → 4 PM (was 55 min from the 2:55 NaturDAO). **Open: agmatine's own regimen rationale is untagged `⟨?⟩` — worth its own provenance pass.**

**2. Mucosal integrity** ⟨C·7-22 | **i:rat villus localisation + human GI-disease cohorts**⟩

DAO is manufactured by **mature villous enterocytes** — a gut-lining enzyme, not one you feed systemically. GI disease reduces DAO activity. This phenotype hits it three ways at once: **5 wks post-open-Frey · Forrest III duodenal ulcers (29 June) · chronic EPI.** If DAO is low, **this** is the likeliest reason — not a vitamin.

✅ **The intervention is already live: L-Glutamine 15 g/day.** Nothing to add.

**3. Copper status** ⟨C·7-22 | **m:pending**⟩

Copper is the **actual metal at DAO's active site**. Zn/Cu 15:1 sits **at** the 8–15:1 boundary that avoids zinc-induced copper depletion — the edge, not past it. ⚠️ **Serum copper is ~95% ceruloplasmin-bound, and ceruloplasmin is an acute-phase reactant — the same trap as ferritin.** At 5 wks post-op it reads falsely reassuring. **Draw with the CRP already on the 8/1 panel, or the number is uninterpretable.**

**4. NOT levers**

- ❌ **B6** — not a DAO cofactor. Upstream, on histamine *synthesis*.
- ✅ **β-ODAP / neurolathyrism — cleared.** Highest-activity vegetal DAO is *Lathyrus sativus* (grass pea, carries β-ODAP), but **NaturDAO's Legumactive® is *Pisum sativum* + *Lens culinaris*** — ~5,000× less. Question doesn't apply.
- ✅ **Trypsin inhibitors vs Creon — cleared, and the framing is now corrected.** ~32 mg vegetable protein/tab → **~96 mg/day**; germination degrades trypsin-inhibitor activity. 🔄 **rev 7.21:** the old wording ("~3 orders below a soy serving… **Constraint intact**") assumed a categorical constraint that no longer exists. **The constraint is now activity-gated, and NaturDAO's ~96 mg/day is the file's quantitative yardstick — not an exception to a rule, but the general case.** See Standing constraints.
- ✅ *Bonus:* the pea-derived product supplies **catalase** alongside DAO, quenching the H₂O₂ that DAO's own reaction generates — a real advantage of the vegetal source.

⚠️ **Do not switch to NaturDAO Plus** — its premium SKU bundles **pyridoxal-5-phosphate** among "8 cofactors," propagating the same retired B6 error.


---

---

## ::🫀 Blood pressure — the numbers::

> ### 🔴 **rev 7.20 — re-baselined against 59 readings (Jun 24 – Jul 22). The old numbers are retired, and one of them was wrong.** *(Unchanged at rev 7.21 — no new readings ingested.)*

**Current export · 59 readings · overall mean 131/67 · systolic 110–156 · diastolic 50–86**

### 🔴 STRUCK: ">35 mmHg postprandial fall"

*This was the doc's central hemodynamic claim, the basis for "two blood pressures, not one," and it does not survive arithmetic.* ⟨C·7-10, **FALSIFIED C·7-22**⟩

**Every paired within-day comparison runs the other way:**

| Day | Fasted morning | Same-day postprandial | Δ |
|---|---|---|---|
| Jul 4 | 138.2 / 69.0 | 139.7 / 69.1 | **+1.5** |
| Jul 7 | 132.5 / 61.5 | 131.0 / 86.0 | −1.5 |
| Jul 11 | 120.5 / 70.5 | 124.5 / 62.5 | **+4.0** |
| Jul 16 | 124.0 / 65.0 | 130.0 / 72.0 | **+6.0** |
| Jul 20 | 124.0 / 69.0 | 131.0 / 62.3 | **+7.0** |

**Mean paired delta: +3.4 mmHg — postprandial is *higher*, not 35 lower.**

⚠️ **Where the "35" came from:** max fasted reading (150, Jul 4 5:20 AM) minus min postprandial reading (115, Jul 6 8:43 PM) = exactly 35. **Different days, extreme-to-extreme, across two distributions.** That is not a within-subject postprandial fall, and the autonomic-involvement inference built on it is **withdrawn**. Largest true single-day systolic swing in the whole record: **26 mmHg** (Jul 4).

### ✅ Re-baseline — fasted, pre-dose

| | Jul 4–8 (old baseline) | Jul 9–22 (now) | Δ |
|---|---|---|---|
| Fasted morning | **136.6 / 66.9** | **120.8 / 66.5** | **−15.8 systolic** |
| Pulse pressure | **69.7** | **54.3** | **−15.4** |
| All readings | 136.7 / 68.0 (n=34) | 124.0 / 66.2 (n=25) | −12.7 |

**The doc's stated fasted baseline of 138/68 is obsolete. At 40 mg, the fasted morning number is 121/67 — normotensive.**

### 🩸 What this supports: the anemia hypothesis, prospectively

*rev 7.14 argued a pulse pressure of ~70 at Hct 25.2 was **high-output anemia**, not arterial stiffness, and predicted that correcting the anemia would drop systolic on its own.* **PP has fallen 69.7 → 54.3 and systolic −15.8, with no antihypertensive change.** That is the predicted result.

⚠️ **Consistent, not confirmed.** ⟨C·7-14→**supported** C·7-22 | **m:BP only — Hgb/Hct not rechecked**⟩ No haematology since. The 7/22–26 iron panel closes it. And the improvement occurred **despite two changes that should have pushed BP up** — Adderall restored to full 70 mg (7/13) and L-Citrulline (vasodilator) removed (7/10) — which strengthens rather than weakens the reading.

### ✅ What survives: the dinner hypotensive load is real — in diastolic

| | Midday diastolic | Evening diastolic | Δ |
|---|---|---|---|
| Jul 4–8 | 68.4 | 63.4 | −4.9 |
| Jul 9–22 | 70.0 | 61.7 | **−8.3** |

*Dinner carries curcumin, resveratrol, quercetin, magnesium; Meal 1 does not. The effect is consistent and has **strengthened**. It is a **diastolic** effect at dinner — not a general postprandial systolic fall.* **Keep the 8:00 PM reading. The 1:30 PM slot can retire after 7/30 as scheduled — its premise is gone.**

### ⚠️ Stop-rule audit

- **Jul 6, 8:43 & 8:44 PM — 115/50 and 116/50. Genuine breaches of the <55 diastolic rule.** The hold at 40 mg was correct and the rule worked.
- **Zero breaches since Jul 8.** Lowest diastolic since: 59 (Jul 17, 8:00 PM). Lowest systolic: 110 (Jul 15, 5:54 AM).
- **Zero readings ≥180/120** in the entire record.

### 🚫 The "late-July uptick" is a sampling artifact — do not act on it

*Jul 14–17 mean systolic 121.4; Jul 19–22 mean 129.4. Looks like an 8 mmHg rise. It isn't.*

**Jul 14–17 was 50% morning readings; Jul 19–22 was 25%.** Mornings run ~6 mmHg below afternoon/evening, so the "rise" is **composition, not physiology**. The single morning reading in that late window (124, Jul 20) sits exactly on the 120.8 mean.

### 📋 Protocol adherence — the number everything gates on is the sparsest

*The 5:55 AM protocol is **3 readings, discard #1, average #2 and #3.** Actual: **9 of 16 days had any morning session, and only 1 of those had 3 readings.** The rest are pairs; seven days have no morning at all; the Jul 22 "morning" reading is timestamped **12:52 AM**.*

🔴 **This is the highest-value fix on the whole BP page.** Every gate, the escalation trigger, and the re-baseline above all rest on the fasted morning number.

### Stop rules — OR, not AND *(unchanged)*

- Any postprandial reading **below 100 systolic** *or* **below 55 diastolic** → hold at 40 mg
- **Lightheaded on standing** → hold at 40 mg, message Kuhlman
- Any reading **≥180/120** → urgent care. Not a message, not a wait.

### 🔄 The 80 mg step — the question has changed

> **The BP case for escalating is gone.** At 40 mg the fasted morning is **120.8/66.5**. The old trigger — "morning average still ≥130" — is not close to firing, and the postprandial-fall finding that framed the risk side was an artifact.

**The live question is no longer *"when do we step to 80?"* but *"is 80 indicated at all, and on what grounds?"*** — which is a **prescriber question for Kuhlman**, not a supplement-layer call. Bring him the re-baseline.

**If any case for 80 remains, it is renal, not hypertensive** — pending **UACR** on the ~8/1 draw. Microalbuminuria would make telmisartan indicated nephroprotection independent of blood pressure. Without it, there is currently no BP indication to escalate.

*Remaining gates, if it ever does step:* hyponatremia workup interpreted · UACR drawn · ≥72 h separation from any other change · **dose it in the evening** (t½ ~24 h decouples Tmax from the morning stack and the morning reading).

*H. pylori status is **not** a gate on telmisartan. Category error — no hemodynamic bearing. It gates MegaSporeBiotic.*

*🔴 Gate 2 ("Hgb ≥ 10") was already falsified at rev 7.15 as a glucometer-floor coincidence (K181070 → Hct 20–60%). It now also loses most of its hemodynamic force: the hemodynamics it was meant to protect have largely normalized on their own.*


---

---

## ::📉 CGM + glucometer — both instruments are compromised::

**Do today · depends on nothing**

- **Urgent Low Soon → ON** *(fires ~20 min before a projected <55)*
- **Falling Fast → ON**
- Urgent Low → 55 ✅
- High → **200** *(inert at 250 — make it a canary)*

*The rationale:* **C-peptide 1.4 (fasting, June 11 2026 — pre-Frey) with GMI 5.7%.** *(Reconciled rev 7.17 from the stale 2.1, which was the June 2025 value; 1.4 is still within range, so "retained secretion" holds, but nearer the floor.)* Retained endogenous insulin secretion plus **⟨assumed⟩** impaired glucagon counterregulation would be the configuration that produces hypoglycemia you cannot sense and cannot self-correct. ⟨? | **a·UNMEASURABLE** | dep:glucagon-impairment⟩

> 🔴 **rev 7.15 — the premise is unmeasurable and the evidence runs against it.** Garbacz 7/15: never tested; no routine clinical test exists. **The low alert stays on regardless** — it costs nothing and the asymmetry favors keeping it. But do not describe this as established physiology.

> 🔬 **Anatomy note (corrected rev 7.14, demoted rev 7.17):** the procedure is a *Frey* (head cored out), not a Puestow. The Frey resects the **alpha-poor head/uncinate** and spares the **alpha-rich body and tail** — so anatomical alpha *mass* is largely spared. **rev 7.17 demotes what that buys:** Kumar 2018 (PMID 30029953 — 30 Frey patients, beta-cell function unchanged at 3 months but **alpha-cell function deteriorated**) shows alpha-cell *function* can worsen independent of mass. So spared mass ≠ preserved function; **directionally reassuring, not dispositive.** It does **not** falsify `glucagon-impairment` — the 180-day behavioral record does. *(Schrader/Meier 2009 is a **distinct, mass-mediated** route from 50% resection — not corroboration, and a partial population mismatch.)* ⟨C→N✓ | DOC:Kumar2018-PMID30029953⟩

⚠️ **The pairing protocol is contaminated.** You are pairing the G7 against a home glucometer at **Hct 25.2**, and at low Hct these meters over-read. Your stick reads high → the sensor looks low → you gate four reintroductions on a bias that lives in the *reference standard*, not the sensor.

**Before pairing:**

1. ✅ **RESOLVED:** strip vial EGS-2003 → FDA K181070 → **Hct 20–60%.** At 25.2 you are **in range.** The protocol runs.
2. Stop calibrating. Fresh sensor. Discard day 1 and day 10, permanently.
3. Pair, don't correct. Sensor first, then stick. **Enter nothing into the app.**
4. Pair at ~4:30 AM and ~5:45 PM — **and whenever the sensor reads below 70.** *Bias is not constant across the range, and the low range is the one that matters.*
5. One meter, one strip lot, record Hct with every pair.

**Interim glycemic instrument**

- 🚫 **HbA1c — do not use.** *Falsely low. Transfused RBCs arrive with near-zero glycation; reticulocytosis adds young cells. Wait ~10–12 weeks post-transfusion.*
- ✅ **Fructosamine — use.** *2–3 week mean, independent of RBC lifespan. Invalidated below albumin 3.0; **yours is 4.0, so it holds.** Draw a simultaneous albumin.*

*The 180-day aggregate survives: GMI 5.7% and 5.8% across two independent windows, zero readings above 250.*


---

---

## ::🦠 H. pylori — the load-bearing unknown::

Forrest III duodenal ulcers on 29 June, in a man who takes **no NSAIDs** and has **documented acetaminophen adversity.** So what caused them? *Duodenal ulcer without NSAIDs is dominated by H. pylori; idiopathic is a diagnosis of exclusion.*

> 📌 **Reconciliation note:** the care team attributes the ulcers to **non-specific post-op change** and is **not** running an active etiology hunt. The analysis below is the *adversarial* case for keeping the question open on your side, not a claim that the team missed something. Hold it as a low-cost hedge, not a live alarm.

⚠️ **A test taken during the bleed cannot rule it out.** In bleeding duodenal ulcers, rapid urease testing alone found H. pylori in 54.5% versus 73% by combined testing — a **25% false-negative rate** (95% CI 11.6–38.4) *in the bleeding-ulcer population*, against 1.6% in dyspepsia. A separate prospective series: RUT sensitivity **80%**, **NPV 75%** in ulcer bleeding, versus 96% and 88% in controls.

*(The one study reporting no loss of RUT sensitivity in GI bleeding studied 61 patients with acute **variceal** bleeding — a different lesion in a different lumen. Named now: it does not match an ulcer-bleeding phenotype and should not be weighted.)*

**Low-cost action — a records request, not a clinician:**

> **"Was gastric *histology* sent from the 6/29–7/3 enteroscopy, or only a CLOtest?"**

- **Serology (IgG)** — PPI-independent. Drawable **now.** Negative = high NPV in a low-prevalence population. Positive ≠ active.
- **Histology** — less PPI-sensitive than RUT. **May already exist. Ask.**
- **Urea breath test / stool antigen** — need ≥2 weeks off PPI **and** ≥2 weeks off mastic gum.


---

---

## ::🩸 Iron — ferritin alone will falsely reassure::

⚠️ **Ferritin is an acute-phase reactant** and you are ~5 weeks post-major abdominal surgery. It will be elevated regardless of iron stores. *In an inflammatory state the thresholds shift: iron deficiency = ferritin <100 (not <30), or 100–300 with TSAT <20%.*

✅ **Ret-He / CHr is the test.** *Inflammation-independent. Reports iron actually incorporated into hemoglobin over the last 48 hours.*

⚠️ **MCV 91 is suspiciously normal.** *Pancreatic proteases cleave B12 from haptocorrin so intrinsic factor can bind it — **EPI directly causes B12 malabsorption.** Iron deficiency drives MCV down. B12 deficiency drives it up. A mid-normal MCV in an actively bleeding, iron-deficient EPI patient is a flag for a masked mixed deficiency, not reassurance.*

*Corroborating: platelets **411**, up from 93K in March. Reactive thrombocytosis is a signature of both post-hemorrhagic recovery and iron deficiency.*

**If IV iron is indicated — ask for the drug by name**

> **Ferric derisomaltose (Monoferric). Not ferric carboxymaltose (Injectafer).**

*PHOSPHARE-IBD (Zoller, Gut 2023;72:644–53), **studied in an IBD population**: incident hypophosphatemia **8.3% (4/48)** with FDI versus **51.0% (25/49)** with FCM — adjusted risk difference **−42.8%**, p<0.0001. Mechanism is FGF23-mediated phosphaturia — **formulation-specific to FCM, not a class effect.** That mechanism is drug-intrinsic, so it transfers out of the IBD cohort to your EPI phenotype; the phenotype mismatch is named and doesn't defeat the transfer.*

*In a post-pancreatectomy patient with EPI and fat-soluble vitamin malabsorption, hypophosphatemia compounds an already stressed mineral and bone picture. **This distinction will not be volunteered.***

> 🆕 **rev 7.19 — copper joins this section's logic.** Serum copper + ceruloplasmin are on the 8/1 draw, and **both are acute-phase reactants exactly as ferritin is.** Read them against the CRP on the same panel or they will falsely reassure in the same way.


---

---

## ::🩸 The coagulation question was aimed at the wrong target::

Three review passes interrogated curcumin's in-vitro antiplatelet activity. **Nobody asked whether you were vitamin K deficient when your duodenum started bleeding.**

- You have **EPI**. EPI causes fat-soluble vitamin malabsorption — **A, D, E, and K.**
- You supplement **D**. You take **K2 (MK-7)**. *K1 is the primary hepatic form for clotting factor carboxylation.*
- You have never had vitamin A, E, or K status checked.
- You now cap vitamin E at **200 IU** — a deliberate low-dose cap **because of its vitamin-K antagonism.** *That constraint anticipated this.*

> **Pull the PT / INR from the 6/29–7/3 admission.** It is a records request, not a new draw.

*If PT was prolonged, the ulcers bled more than they should have, the driver was structural and lifelong and fixable — and the entire antiplatelet-supplement conversation retires.*


---

---

## ::🧂 Sodium — two live hypotheses. Do not pre-commit.::

~68 mEq/day of supplemental sodium moved serum Na from 132 to 134, then plateaued. *In simple depletional hyponatremia, that load should have normalized you within days.* **A near-null response to sodium loading is not a treatment failure. It is a positive diagnostic finding.**

**Reset osmostat** — *favoured by years of stable, asymptomatic, defended Na 132 that does not move on salt. If confirmed, the finding is **futility, not danger**.*

**SIADH** — *favoured by the drop to **129** during the March bacteremia. A true reset osmostat defends its setpoint; an acute fall with intercurrent illness suggests ADH excess being unmasked. And in euvolemic SIADH, NaCl without water restriction can lower serum Na **further**.*

**Never formally evaluated:** post-operative non-osmotic ADH release · EPI-driven subclinical hypovolemia · adrenal insufficiency *(exclude first — 8 AM cortisol)* · reset osmostat · **amphetamine (Adderall now 70 mg/day)** · simple free-water excess *(never quantified — log three days of total fluid intake)*

⚠️ **Uric acid discriminates the SIADH family from hypovolemia. It does *not* discriminate SIADH from reset osmostat** — it is low in both. **The supervised oral water load does that:** >80% of the load excreted in 4 hours = reset osmostat; <80% = SIADH. *A nephrology test, not a home experiment.*

> 🚫 **Do not raise the sodium dose in response to a low Na until the mechanism is known.**


---

---

## ::⚖️ What adversarial passes did *not* find — and what rev 7.21 did::

**Zero findings against the chronic core.** *Not one of the ~40 compounds was faulted on evidence quality, dose, timing, interaction, or sourcing — outside the specific post-surgical, post-hemorrhagic window.*

**What the rev 7.19 pass found — three material items, all reasoning, none faulting a compound:**

1. **A stated rationale was factually wrong.** "B6/P5P, the DAO cofactor" — untagged, forty-year-stale, and load-bearing. The compound stayed; the reason for it did not.
2. **An unexamined substrate collision.** Agmatine at 1 g/day is a DAO substrate, dosed 55 minutes before a DAO supplement. Never named in any prior rev.
3. **A measurement asymmetry.** The 8/1 panel measured **zinc** (the antagonist) and not **copper** (the actual DAO cofactor being antagonized).

**rev 7.20 added a fourth:** the **">35 mmHg postprandial fall"** — the doc's central hemodynamic framing since rev 7.10 — **was an arithmetic artifact.** Max fasted minus min postprandial, across different days. Six revisions of adversarial review did not catch it because no pass ever recomputed it from the raw readings. *Every population-level and cross-day comparison in this file should be assumed guilty until paired.*

**rev 7.21 adds a fifth, and it is the oldest one yet: the soy hard constraint.**

*Same shape as items 1 and 4, but it survived longer and sat higher.* It occupied the **❌ tier** — alongside acetaminophen and NSAIDs, both of which cite a documented source — while itself carrying **a mechanism and no source**, in no premise registry, unsupported by the Creon label, and with **not one of its cited studies touching PERT, EPI, or fat absorption.** It shaped supplement selection for the life of the document and generated two named "exceptions" (SerinAid PS, NaturDAO) that were really just the general case showing through.

**What makes this one instructive:** the rule was never *checked* because it was never *doubted* — it read like a contraindication, so it was treated as one. **The tell was structural and visible the whole time: it was the only Standing Constraint that explained itself instead of citing something.** *A rule that argues for itself is a rule with no source.*

**And a self-correction, logged:** in this rev Claude over-corrected on the allergen-label reading — inferring manufacturer intent from the FALCPA vehicle without having the label text, then concluding the milk risk was understated. **Wrong, and withdrawn**; ND states its meaning explicitly. *Reasoning from a proxy while the primary document exists is brief item 5, and it does not stop applying to the reviewer.*

**The record this regimen produced:** GMI 5.7–5.8% across 180 days · zero readings above 250 · eGFR 112 · normal hepatic function despite an obstruction history · **HbA1c 6.1% in Type 3c diabetes, without insulin.** *That is an outcome record, not a hypothesis.* **The master regimen was never on trial. What gets corrected are the overlays, the instruments, and the reasoning.**


---

---

## ::🎯 What to expect — a principle, not a prediction::

**In Type 3c with retained C-peptide (1.4, June 2026) and ⟨assumed⟩ impaired glucagon counterregulation, a higher glucose floor with lower variance beats a lower mean with excursions you cannot feel or counter.**

Gymnema is out. Berberine is out. Citrulline is out.

*If your numbers drift up slightly, **that is not a regression to correct.*** ⟨C·7-10→N✓7-13 | **a** | dep:glucagon-impairment, alpha-cell-function⟩

> 🔴 **rev 7.15 — "That is physiology" is STRUCK. It was never physiology.**
>
> This principle stands on `glucagon-impairment`, which is **⟨UNMEASURABLE⟩** — never tested, and **no routine clinical test exists.**
>
> **Counter-evidence, population named:** the T3c brittleness literature is overwhelmingly total/near-total pancreatectomy, insulin-dependent, minimal C-peptide. **My phenotype:** C-peptide 1.4 retained, non-insulin-treated, head-only Frey, body/tail intact. **Population does not match → per brief item 3, do not weight it.**
>
> **Keep the principle if you want it — but it is `⟨a⟩`.** Do not call it physiology again.
>
> 📏 **rev 7.17 — and do not track it with the wrong instrument.** This governs *glucose*. It does **not** license using a single fasting **C-peptide** to judge whether a reinstated agent "helped the pancreas." Track fasting insulin + HOMA-IR alongside, or the reading is uninterpretable — and **"restore C-peptide toward 2.1" is retired.**


---


---

> ### 🗄 How the dual-AI rule was corrected
> ### 🔄 **rev 7.24 — the rule as previously written was wrong, and it was Claude enforcing it.**
> *The absolute form — "never feed one model's output to the other" — was **⟨C⟩-origin, untagged, and overbroad.** It collapsed two operations that are not the same thing, and because the settings doc is read first every session, Claude re-derived the strict version from scratch each time and talked Nate out of a workflow he had built, validated, and used successfully for months. **Same failure class as the soy constraint and the B6/DAO error: a plausible rule, stated absolutely, wearing an authority it never earned.** Retired 2026-07-23.* ⟨**C·retired → N·7-23**⟩

---

> 🗄 **Struck from the Erinamax gate at rev 7.23** — ⟨*the `≤3-change freeze` clause that stood here was struck at rev 7.23 — the freeze was retired at rev 7.16 and this was its last live instance. **The reason to wait is attribution and the unresolved 2× spec, not a change quota.***⟩

---

## ::📜 Prior revision footers — verbatim::

*Master Regimen v6.20 · rev 7.24 · 2026-07-23*
*Supersedes rev 7.23 (2026-07-23, same day). rev 7.24: **the dual-AI cross-feeding prohibition is RETIRED in its absolute form** ⟨C·retired → N·7-23⟩ and replaced with the two-operations distinction — corroboration requires independence, **audit and synthesis does not and is the primary workflow** · the correct version was already present at rev 7.21 and buried under the absolute headline · `cross-feed-prohibition` added to the premise registry · open item 21 extended to two corrections. **Nothing in the schedule changed. Totals untouched.***

> ### 🧭 **Where this one sat matters**
> *Every other error this file has caught lived in the protocol — a dose, a rationale, a placement. **This one lived in the method that produces the findings.** A wrong rule there does not corrupt one compound; it corrupts the process that audits all of them, silently, in every session. It survived seven revisions of adversarial review because it sounded like rigor — and because the reviewer was the one enforcing it.*

---

*Prior rev footer, retained:*

*Master Regimen v6.19 · rev 7.23 · 2026-07-23*
*Supersedes rev 7.22 (2026-07-23, same day). rev 7.23: **the change-accounting badge — "no dose amount changes · no slot moves · nothing added, nothing reinstated" — is STRUCK from this document and is not to return** ⟨N·7-23⟩; it was the ledger of the ≤3-change freeze, **a rule retired at rev 7.16 whose scoreboard outlived it by seven revisions** · the last live `≤3-change freeze` clause struck from the Erinamax note · **ProBiota HistaminX reconciled as one product** started 7/21, with MegaSporeBiotic (~8/5) the probiotic still behind the workup · **HistaminX label-timing discrepancy logged** → open item 22 · **Erinamax placement reopened as a three-branch fork** → open item 23. Totals re-verified: sixteen sums, zero mismatches.*

> ### 🧭 **What a rev is scored on now**
> *Not how little moved. **Whether every total is an exact sum of the schedule, whether every new claim carries a provenance tag, and whether the open items got smaller.** Three of rev 7.22's findings remain **⟨C·7-23⟩ and unconfirmed** — logged as questions, not corrections, so they cannot harden the way B6, the ">35 mmHg fall," and the soy rule each did.*

---

*Prior rev footer, retained:*

*Master Regimen v6.18 · rev 7.22 · 2026-07-23*
*Supersedes rev 7.21 (2026-07-23, same day). rev 7.22: **ALCAR 4:00 PM → 1:45 PM** ⟨C·7-23→N✓7-23⟩, total unchanged at 2,500 · **ProBiota attribution window FALSIFIED as single-variable** — three schedule changes inside it, Erinamax into Meal 1 one day after the start · **7/22 agmatine move reopened** (window treated as instant; direction never checked) · **5:00 PM agmatine named** as the tightest remaining coupling, 85 min · **`agmatine-indication`, `alcar-split-rationale`, `probiota-window-clean` added to the premise registry** · **settings doc flagged stale on soy.** Open items 19–21 added. *⟨the change-accounting line that stood here was struck at rev 7.23⟩*

> ### 🟢 **ALCAR repositioned · three findings logged OPEN · change-accounting badge retired.**
> *Three of this rev's findings are **⟨C·7-23⟩ and unconfirmed** — logged as open questions, not corrections, precisely so they cannot harden into fact the way B6, the ">35 mmHg fall," and the soy rule each did.*

---

*Prior rev footer, retained:*

*Master Regimen v6.17 · rev 7.21 · 2026-07-23*
*Supersedes rev 7.20 (2026-07-22). rev 7.21: **soy hard constraint REWORKED and DEMOTED ❌ → ⚠️** — KTI/BBI are serine-protease inhibitors, trypsin + chymotrypsin only, **not lipase**; "stops Creon working" retired · **Erinamax soy gate RELEASED** on arithmetic (~5 µg/day vs. the ~96 mg/day already cleared) · **Erinamax label verified** — as ErinaPrime®, Contains: Milk, Soy; product identity resolved · **milk separated out** as an open dairy-avoidance question, not a Creon question · **2× erinacine A spec discrepancy opened** (ND 0.5% floor vs. Nammex 1.0%) · **citation trap logged** (MDPI 15 mg/day → correct 5.25 mg/day) · **brief item 6 added** — do not carry constraints into the question · **ProBiota HistaminX confirmed started 7/21**, reconciled out of the queue · dietary avoidances line adopted ⟨N✓7-23⟩.*

> ### 🟢 **One constraint reworked, one gate released.**
> *Supplement-layer decisions rest with Nate (the standing authority on this layer). Prescribed-medication changes (telmisartan, Adderall) and clinical-status determinations (labs, healing, imaging) are the care team's domain — named as facts to obtain, not approvals.*

---

## ::📜 Revision history — condensed::

- **rev 7.25 · 2026-07-23** — TAU relocated to fed slots (Meal 2 · Dinner) at Nate's direction; Magtein #2 moved 1:45 PM → 12:30 PM. **The protocol/archive split is executed** — a standing instruction that five prior revisions failed to apply.
- **rev 7.24 · 2026-07-23** — the absolute cross-feeding prohibition retired; corroboration and audit/synthesis separated as two operations.
- **rev 7.23 · 2026-07-23** — the ≤3-change freeze's accounting ledger struck throughout; ProBiota HistaminX confirmed as one product; Erinamax placement reopened.
- **rev 7.22 · 2026-07-23** — ALCAR 4:00 PM → 1:45 PM; the 7/21→8/4 attribution window falsified as single-variable; the 7/22 agmatine move reopened.
- **rev 7.21 · 2026-07-23** — the soy hard constraint reworked and demoted; Erinamax soy gate released on arithmetic.
- **rev 7.20 · 2026-07-22** — the ">35 mmHg postprandial fall" falsified as an arithmetic artifact.
- **rev 7.19 · 2026-07-22** — B6/DAO cofactor rationale struck; agmatine relocated; copper added to the draw.

---

*Master Regimen · **ARCHIVE** · 2026-07-23 · companion to PROTOCOL v6.21 / rev 7.25*