Reversibility, single-driver, and the honest kernel boundary (MALT, anal; GIST/NET/small-bowel/cholangio)

The kernel's signature is reversibility: gastric MALT lymphoma is an H. pylori barrier-scale reduction that regresses when eradication restores the scale (RR≈6→1). Anal squamous carcinoma is a single sustained HPV bias, monotone in exposure. Four sites stay out-of-kernel with stated obstacles — GIST and NET (gene-key), small-bowel adenocarcinoma (needs the immune layer), cholangiocarcinoma (hepatobiliary). Reversibility is [V]; absolute incidence [O].

Two sites sharpen the kernel's signature, then it states its own boundary. The strongest falsifiable claim is reversibility: a barrier-SCALE-driven cancer must regress when the scale is restored. Gastric MALT lymphoma is the clean test — an H. pylori g-reduction (RR~6) that regresses on eradication (~90% Hp+, ~77.5% remit on eradication alone, Zullo 2010); restoring g drops the rate monotonically to one. The ~22% t(11;18)/API2-MALT1 non-responders are a g-INDEPENDENT driver, outside this kernel and owned by disease_wp (the boundary the model should draw). Anal squamous carcinoma is the single-driver case: a sustained HPV E6/E7 bias, ~90% HPV-attributable (De Sanjosé 2019), monotone and vaccine-preventable. Treatment reading: remove the scale driver (eradication) or never apply the bias (vaccination) — prevention is the lever, established-tumour therapy is out-of-model. The honest boundary: the carcinogen-Kramers kernel does NOT cover GIST or gastroenteropancreatic NET (gene-key / neuroendocrine → disease_wp), small-bowel adenocarcinoma (needs the C1 immune layer to supply its barrier-lowering driver), or cholangiocarcinoma (needs a hepatobiliary + immune layer; circulatory seam). MALT reversibility and the anal monotone are [V] against cited anchors [L]; absolute incidence [O]; the four out-of-kernel sites are honest [O] with stated obstacles — stating the boundary is the result.

Two more sites sharpen the kernel's signature, and then it states its own boundary honestly. The kernel's strongest falsifiable claim is reversibility: if a cancer is driven by a barrier-scale reduction, then restoring the scale must drop the crossing rate. Gastric MALT lymphoma is the clean test — an H. pylori chronic-inflammation g-reduction (RR ≈ 6.0) that regresses when eradication restores g: about 90% of cases are H. pylori-positive and roughly 77.5% remit on eradication alone (Zullo 2010). Restoring the scale drops the rate monotonically to one:

The ~22% of MALT cases carrying t(11;18)/API2-MALT1 that do not respond are a g-independent oncogenic driver — outside this barrier-scale kernel and owned by disease_wp (gene-key), which is exactly the boundary the model should draw.

Anal squamous carcinoma is the single-driver case: a single sustained HPV (E6/E7) bias h, ~90% HPV-attributable (De Sanjosé 2019), monotone in exposure and preventable by vaccination — the bias removed before it is ever applied.

Treatment (model reading). Both readings make the same point: remove the barrier-scale driver (eradicate H. pylori; MALT regresses) or never apply the bias (HPV vaccination; anal carcinoma prevented). Prevention/risk-reduction is the lever the kernel supplies; therapy of an established tumour is out-of-model.

The honest boundary. The carcinogen-Kramers kernel does not cover every GI cancer, and the extension states where it stops and who must supply the missing layer rather than over-reaching:

MALT reversibility and the anal monotone are verified [V] against cited anchors [L]; absolute incidence stays open [O] (external population calibration). The four out-of-kernel sites are honest [O] with stated obstacles: GIST and gastroenteropancreatic NET are gene-key / neuroendocrine (disease_wp); small-bowel adenocarcinoma needs the C1 immune layer to supply its barrier-lowering driver; cholangiocarcinoma needs a hepatobiliary + immune layer (a circulatory seam). Stating the boundary is the result here — the kernel earns trust by drawing its own edge.