AUTONOMOUS COMPARATIVE ONCOLOGYCANINE HSA × HUMAN AS

An engine that tries to be wrong.

It generates a hypothesis, runs the compute to test it, and tries to falsify it first — then packages the evidence and compounds. We pointed it at canine hemangiosarcoma and human angiosarcoma.

PIK3CA→endothelial axis·human AS, p=0.001 alpelisib–PI3Kα·~11 nM predicted vs 4.6 nM measured canine pocket 100% conserved·H1 supported docking·redock RMSD 1.8 Å 3 hypotheses falsified·then sharpened real-GPU checkpoint/resume·durable handoff

Most systems are built to agree with you. This one is built to disagree with itself — cheaply, first, on purpose.

THREE PRIMITIVES
01 / SELF-FALSIFICATION

It tries to disprove itself.

Every hypothesis ships with the cheap, pre-registered test that would kill it — run before anything expensive. The engine falsified its own flagship idea and caught a tumor-purity confound that would have sent us chasing a ghost. Being wrong fast is the feature.

RECEIPT — PIK3CA→immunosuppression: refuted · reframed to the endothelial axis · replicated in human AS at p=0.001
02 / PROOF CAPSULES

Evidence becomes an object.

A result isn't a message — it's a portable, verifiable unit: what was run, against what, the directional signal, the confidence, the limitations. Capsules move between machines and people without losing their meaning. Evidence you can audit, carry, and trust on its own.

RECEIPT — alpelisib–PI3Kα: iptm 0.99 · ~11 nM · supports · off-target VEGFR2 drops to 0.11 · refutes
03 / THE FLYWHEEL

It compounds. (3,3)

Every run gives back — open datasets, signature registries, models — and each contribution lowers the cost of the next. A research engine that gets more valuable the more it gives away. Positive-sum by construction, not by goodwill.

RECEIPT — cross-species cohort + 16-panel canine TME registry (both genome assemblies), published
EVERY NUMBER IS REAL

It earns the claim, then says how far it goes.

0
Cross-species replication. PIK3CA→endothelial axis in human angiosarcoma, p=0.001 — after the immune story was falsified.
0
Co-folded alpelisib–PI3Kα vs 4.6 nM measured — within ~2.4×, iptm 0.99.
0
Canine PI3Kα pocket conserved. The drug's target is the same protein in the dog — H1 supported.
0
Docking validated — alpelisib redocked to its crystal pose, RMSD 1.8 Å.
THE LOOP

Hypothesize. Compute. Falsify. Capsule. Compound.

01

Hypothesize

With the test that would kill it.

02

Compute

Pluggable GPU lanes — dock, co-fold, MD, omics.

03

Falsify

Cheap pre-registered crux, run first.

04

Capsule

Portable evidence: signal, confidence, limits.

05

Compound

Publish datasets & models; the next run gets cheaper.