Second-order systems for
consequential infrastructure
Second Order Labs builds infrastructure where evidence, provenance, review, and memory matter more than the first answer. The company behind Protean Labs.
Where first-order demos end, second-order systems begin
We build at the layer after generation: validation, proof, deployment, and compounding feedback.
Autonomous Scientific Infrastructure
Systems where models propose hypotheses, evidence retrieval validates them, and deterministic validators gate what survives.
Provenance & Public Proof
Every important output is replayable. Provenance trails make computation inspectable and claims verifiable.
Bounded Agentic Operations
Autonomy with constraints. Systems earn authority through demonstrated reliability, not assumed competence.
Research Memory & Feedback
Systems that remember what worked, what failed, and why. Memory as infrastructure, not afterthought.
Review-Gated Deployment
Human review gates before outputs gain authority. Automation proposes; review decides what ships.
Company-Scale Operating Systems
Infrastructure that compounds. Systems designed for the long term, not the demo.
How we operate
Principles that guide how we build and deploy systems.
Observe the System
Understand the domain deeply before building. Map the incentives, failure modes, and existing constraints.
Model the Incentives
Design for how the system will actually be used, not how it should be used. Align outputs with outcomes.
Bound the Autonomy
Give systems authority only where they have demonstrated reliability. Constraints enable trust.
Build the Proof Surface
Make every important output inspectable. Provenance trails let evidence survive the interface.
Review Before Authority
Human review gates before outputs gain power. Automation proposes; review decides what ships.
Compound What Survives
Memory and feedback loops turn validated outputs into institutional knowledge that compounds.
Protean Labs
A provenance-aware scientific operating system for peptide discovery. Models propose, deterministic validators and review gates decide, and public provenance makes the process inspectable.
Evidence Retrieval
Structured scientific evidence informs every hypothesis
Deterministic Validators
Rule-based gates that enforce scientific constraints
Public Provenance
Every claim traces back to its computational origin
Review Gates
Human review before publication or deployment
Note: Computational rankings are research-prioritization signals. They do not constitute biological validation, therapeutic claims, or clinical readiness.
Common questions
Building where consequences matter.
For strategic partnerships, research collaborations, infrastructure conversations, or high-signal operators: let's talk.