← back

m625 is not a clinical decision tool · not a publisher · not an expert authority · not a simulator. It is a cross-domain Roll-formalized evidence atlas with public audit + attack surface.
persona: VC · domain: cancer · tier: 1
ferry target: m624 · fixture id: DEMO-003
self-audit grade: A

i_demo_003 · VC × cancer · in-vivo CAR-T cross-domain transferability

Why a VC would come to evidence.x1000.ai

A VC evaluating in-vivo CAR-T (or any phi_direct_rewrite modality — gene editing, RNA gene-writing, in-vivo CD19 platforms) needs to assess: how robust is the cross-domain transfer story when a company pitches "we built it for oncology, neuro is next"? Pain points (see 05_docs/M625_PERSONAS.md):

m625 does not advise on any specific pipeline. **m625 shows what a Roll-formalized cross-domain transferability claim looks like, and what would invalidate it**. The VC is invited to use this framework against any incoming pitch.

This is not an investment recommendation. This is not an evaluation of any specific company. (See "What this demo will not say".)


The Roll-formalized object (anchored on m610 i_case_cancer_450)

FieldValue
Upstream cooked casem610 i_case_cancer_450_v0.2.md (R/R DLBCL CD19 CAR-T · TMU-C-major + TMU-A-minor R-CHOP)
Domainhematologic oncology (DLBCL) — proven transfer; chronic pain / autoimmune — proposed transfer per m620 ladder
Roll TMU typeTMU-C-major (phi_direct_rewrite) + TMU-A-minor (R-CHOP frame)
K4.1 cross-period 3-bitK4.1 = (1, 1, 1) per m604 Framework §4.5 — the only m610 base case with this bit pattern
Cross-domain verificationm620 02_exec_003_DLBCL_CART (per W0 pre-flight L1 disk verified 2026-05-22) — verification ladder anchors the transfer claim
m610 manual auditAUDIT_006 92/100 (training_data_recall from 01_materials_cache/A3_m610_cooked_cases_anchor.md)
m610 machine audit91 → 98/100 (v0.5.1 calibration ±1; training_data_recall)

m625 does not modify m610 or m620 artifacts. References are to the cooked sovereign paths above.


How a VC should read this evidence (≤ 5 minutes)


Cross-domain transferability declaration (per W6.3)

The K4.1 = (1, 1, 1) phi_direct_rewrite pattern is anchored in m610 i_case_cancer_450 (hematologic oncology). m620 02_exec_003_DLBCL_CART is the verification anchor in oncology. Proposed extensions (e.g., to autoimmune via in-vivo CD19 platforms) are open conjectures, not m625-asserted facts. transferability_pct: the bit-pattern match is necessary. The sufficient condition is an m620 verdict on the proposed extension target, which is per-pipeline open work. m625 does not assert any specific pipeline qualifies.


What this demo will not say (anti-drift · per 01_materials_cache/C2_m625_anti_drift_manifesto.md)

If a VC reads any of the above into this demo, that is a failure of this demo and should be filed under `_ops/BLOCKED_LOG.md` for next-round revision.

Falsifying prediction (per W6.5 + Roll EPP-3)

The K4.1 = (1, 1, 1) phi_direct_rewrite pattern, as Roll-formalized in m610 i_case_cancer_450, would be falsified by either of:

The cross-domain transfer claim (to chronic-pain autoimmune or other domains) is falsified by m620 verdict < 70 at the cross-domain exec dir (e.g., `05 AUTISM`, `06 CKD`).

Attack surface entries (per 05_docs/M625_ATTACK_SURFACE_v0.1.md)


Ferry target (per W6.6)

Frontmatter `ferry_target: m624`.

Honest verdict (per W6.5)

- *"VCs need company names."* — Filed under `05_docs/M625_PERSONAS.md` anti-drift; m625 will not name pipelines. VCs build their own watch-list against this framework. - *"What if no pipeline ever passes m620?"* — Then the framework correctly says no specific pipeline has demonstrated cross-domain transfer. That is honest. - *"Why is `(1,1,1)` privileged?"* — It is not; it is rare. Cytostatic `(0,1,1)` and synthetic-lethality `(1,1,0)` patterns are also commercially significant. See i_demo_001. - *"Reproducibility?"* — `cd /data/projects/m625_roll_evidence_atlas && python3 -m 02_audit.audit --artifact 03_demos/i_demo_003_VC_cancer_invivo_carT/README.md` · expected grade A · 0 vetoes.