← Platform
Platform & handoff
Modality router — modalities and enforced claim discipline.
Which modalities are covered, and how claim discipline is enforced across them.
Modality router · claim discipline
One workbench, many modalities — each gated by what it has actually earned.
DiscoveryLab routes a causal biological hypothesis to a modality, then bounds what it may claim by that modality's evidence. The rule is enforced in code (ModalityBenchmarkRegistry): a modality without a calibrated benchmark can only produce rank-only or governance output — never a discovery claim. Peptides are the mature, calibrated modality; the rest are honestly staged.
modalityclaim levelbenchmarkstatus
Peptide
discovery
PeptCheck v1.9
Calibrated reference modality. Discovery claims permitted; safety axes carry machine-checked Lean receipts.
Small molecule
rank-only
ChemCheck v0
Deterministic developability triage (PAINS/Brenk + Lipinski/Veber), plus a CNS translational layer: hypothesis-tiered mechanism routing and occupancy from supplied affinity + free-brain concentration. Benchmarked from public filters; awaiting calibration on measured outcomes.
Microbiome-mediated ecosystem
rank-only
Microbiome Ecosystem Hypothesis Graph v0
Ecological indirection scaffold: helper organism → target organism/guild → metabolite/signal → host-state hypothesis. Produces measurement-first plans; no colonization, safety, or efficacy claims until calibrated.
mRNA / transient expression
rank-only
mRNACheck v0
Deterministic CDS plausibility (frame/start/stop, GC, homopolymer, innate-immune proxies). Benchmarked; awaiting calibration.
Gene therapy
governance-only
VectorCheck v0
Governance scorecard over declared vector attributes (integration, payload, reversibility). Evaluation/audit only — no design claims.
Biologic · cell therapy · targeted delivery
hypothesis-only
—
No adapter or benchmark yet. Can appear as a routed hypothesis, never a discovery claim.
The split that keeps this honest: a benchmark (often built from public data) supports ranking; only calibration on our own measured outcomes unlocks discovery claims. "Found data online" moves a modality to rank-only, not discovery. Source:
Packages/PeptiterDiscovery/…/ModalityRouting.swift and experiments/{chem,mrna,vector}check-v0/ and experiments/microbiome-ecosystem-hypothesis-graph-v0/. The same rank-only discipline governs route selection: a deterministic DeliveryTriage sorts a candidate into oral / local-gut / alternative-route buckets (grounded in the oral-octreotide and oral-semaglutide precedents) without predicting bioavailability.More on platform
Related on this track
Experiments
Validation and assay planning.
AI Scientist
Research loop and next-action reasoning.
Latest results
BioFoundation, wet-lab prep, and Bio-JEPA run.
Agents
Agent-facing operational surface.
Architecture
System layers and contracts.
Gut–vagus graph
Curated peptide/receptor seed for the vagal axis.
Operations
Run, review, and inspect surfaces.
LabSpace
Partner handoff and returned results.
Snapshot
Observation half of the improvement loop.
Learning loop
Bounded recursive evidence improvement.
Technical credibility
Evidence, calibration, and claims.
Mathematical framing
Why the model is bounded.
Safety
Research-use boundaries.
Demo
Final call to action.
← Back to Platform index
All sections on the platform track.