Protean Labs
What we are,
what we are not.
Protean Labs is a private research lab building computational infrastructure for peptide discovery. The runtime ingests evidence, proposes candidates under explicit constraints, ranks them against deterministic validators, and prepares review-gated handoff to wet-lab partners.
We are a research institution, not a wet-lab provider. We are not a peptide generator, not a chatbot, not a dashboard, not a token site. The website is the public surface of a private scientific system. It describes architecture and commitments — not internal operating mechanics.
The runtime is local-first by default and bounded by design. Models propose, embed, rerank, and explain. Deterministic validators, failure memory, bounded learning rules, review gates, public/private boundaries, and human scientific review remain authoritative.
Computational rankings are research-prioritization signals. They do not establish biological activity, efficacy, safety, patentability, or clinical readiness. Wet-lab review remains authoritative.
01—Posture
How we
work.
Four operating principles that shape every system we build. They are not aspirational — they are enforced in code, in the workflow DAG, and at every publication boundary.
01
Local-first runtime
Default execution is on-device. External providers are escalation paths, reviewed and bounded — not default scientific infrastructure.
02
Bounded orchestration
Workflows are DAGs with declared inputs, outputs, allowed write roots, failure policies, and review gates. Step cap of 16, zero recursion.
03
Scientific object truth model
Candidates, hypotheses, experiments, evidence bundles, papers, and runtime cycles are typed objects. Truth lives in the object graph.
04
Review-gated escalation
Wet-lab submission, scoring mutation, validator mutation, and code rewriting are not permitted as autonomous actions. Ever.
02—What's next
Infrastructure
expands only
where the science can hold it.
Roadmap items are described by direction, not by date. Reserved design space is preserved as optionality, not promised as feature work.
Now
Sealed cycles, public anchors
Continuous cognition loop, sealed cycle snapshots, redacted public provenance graph, and on-chain anchoring on Base mainnet.
Next
Deeper scientific memory
Contradiction graph as a default cycle stage, richer assay feedback pathways, and per-step enforcement of the workflow DAG.
Then
Programmable wet-lab handoff
Simulated dry run, real provider adapter, on-chain payment, and provider packets linked into the public provenance graph as typed edges.
Reserved
Network-aligned coordination research
Counsel-reviewed coordination surfaces over public-safe state. Reserved design space — not a committed mechanism.
03—Contact
For scientific
and engineering
collaboration.
We work with wet-lab partners, computational chemists, and infrastructure engineers on review-gated cohorts. If you study peptide discovery, programmable wet-lab execution, provenance infrastructure, or bounded scientific cognition — start with the architecture paper, then the live provenance graph, then write to us.
Boundary
The runtime proposes; deterministic code decides; the reviewer remains authoritative. This boundary holds across every part of the system. In code, in documentation, in published artifacts, and in the copy you are reading right now.
Protean Labs is a research lab — not a wet-lab provider, not a registered investment vehicle, not a clinical trial sponsor. Computational rankings are research prioritization, not biological proof.