Mechanism-verified, private-infrastructure drug discovery for the agent era.

Peptiter is an agent-native drug-discovery platform — across peptides, designed proteins, and small molecules — that lets pharma run mechanism-verified, audit-by-construction discovery loops on infrastructure they control, and that machine-checks whether a cross-modal combination is a Lyapunov-stable controller. Strongest in peptide discovery.

First: private data cannot leave the building.

Internal compound libraries, target hypotheses, patient-derived data, unpublished assay results, and competitive mechanism maps are not assets pharma wants routed through hosted AI APIs. The winning discovery stack has to run inside the customer's environment.

Second: AI-ranked candidates are not enough.

Regulators are increasingly focused on how AI is used across the drug-development lifecycle. FDA has published draft guidance on AI used to support regulatory decision-making for drug and biological products, and CDER reports a sharp rise in submissions containing AI components. EMA's AI reflection paper covers AI and ML across the full medicines lifecycle, from discovery through post-authorization. The direction is clear: "the model ranked it first" will not be a sufficient explanation for high-stakes candidate decisions.

Third: generative discovery is better at proposing than refusing.

Most AI discovery platforms optimize for novelty, binding, or phenotype lift. They are weaker at saying: this mechanism is unsupported, this claim is overextended, this safety mode is unresolved, this assay is not informative, or this candidate should not advance.

That creates a gap. Pharma does not only need more generated molecules. It needs AI systems that can operate privately, reason mechanistically, produce audit trails, and enforce safety boundaries in code.

1. Formal verification of biology

Every mechanism claim is treated as an artifact that must be checked, not as prose from a model. DiscoveryLab binds mechanism hypotheses to machine-checkable proof receipts. The model can propose a candidate, a pathway rationale, or an assay plan, but the platform marks it as unverified until it crosses the audit boundary.

The reviewer is not asked to trust the model. The reviewer gets the claim, the evidence path, the proof receipt, the failure modes, and the human approval trail.

2. Agent-native, private-infrastructure deployment

DiscoveryLab is built to be driven by AI agents, not merely used through a dashboard. A researcher or pharma customer can run the system inside their own environment — laptop, private cloud, secure cluster, or air-gapped server. An agent host can drive literature search, mechanism construction, candidate generation, verification, prioritization, and lab orchestration without sending proprietary data to a third-party model API.

The strategic bet: the next discovery platform is not just SaaS. It is an agent-operable scientific system that runs wherever the customer's data lives.

3. Audit boundary as architecture

DiscoveryLab encodes the audit boundary into the system itself:

• model output starts as an unverified candidate;
• mechanism claims require proof receipts;
• lab jobs require human sign-off;
• recursive improvements require evaluation and approval;
• promotion creates an auditable artifact;
• rollback is part of the lifecycle.

The result is a discovery loop that is not only AI-assisted, but reviewable, defensible, and operationally constrained.

Peptides are more constrained than large protein design, often more mechanism-readable than small molecules, and naturally suited to receptor/pathway biology, protein-protein interfaces, signaling motifs, and short regulatory sequences.

Peptiter is not trying to win by generating the largest possible molecule universe. The platform is designed to win where mechanism, traceability, and candidate discipline matter. Where much of the field competes on generative breadth, Peptiter competes on:

• constrained design space;
• pathway-aware candidate generation;
• mechanism-first prioritization;
• private deployment;
• formal audit trails;
• human-gated lab execution;
• recursive improvement from verified outcomes.

Peptides are the calibrated reference modality and the standout strength — but the headline differentiator, the machine-checked combination-controller proof, is modality-general: it composes peptides, designed proteins, and small molecules on orthogonal mechanism points. Peptides are where we lead, not the ceiling of what the platform certifies.

condition
→ pathway map
→ peptide hypothesis
→ mechanism claim
→ proof receipt
→ candidate ranking
→ assay plan
→ human approval
→ lab outcome
→ calibration update
→ next hypothesis

Every failed proof, rejected mechanism, reviewer correction, assay outcome, and regression result feeds the Recursive Discovery Loop. The system improves its candidate ranking, retrieval policy, mechanism templates, assay selection, and calibration metrics — but it cannot rewrite safety gates, approval boundaries, or claim thresholds.

The platform recursively improves its scientific process, not its authority.

The investor story should include this because it explains why Peptiter is more than a candidate generator. TensFormer-style world models let the system propose an intervention, predict a typed body-state shift, and expose the mechanism as a bounded artifact that can be checked before anyone treats it as scientific evidence.

The language layer is TensorLang: an experimental tensor-equation language for explicit finite domains, tensorized relations, semiring-style reasoning, small trainable fragments, and bounded dynamical models. In DiscoveryLab, TensorLang is not the wet-lab truth source and not the frontier neural model. It is the typed middle layer that turns AI inventions into executable, hash-bound mechanism claims.

AI proposes candidate mechanism
 TensFormer world model predicts intervention-conditioned state shift
 TensorLang emits typed tensor/relation artifact
 Lean verifies finite exported claims
 wet lab decides biological truth

That is a sharper investor wedge than "we use AI." The platform can invent, but the invention has to pass through a typed world model, an audit boundary, and eventually a biological experiment.

1. Inflammation / IL-23 axis

The original DiscoveryLab vertical focuses on pathway-aware peptide discovery in inflammation, using the IL-23 axis as a mechanism-grounded starting point. It is the first proof that the system can connect condition-first discovery to formal mechanism checks.

disease condition
→ pathway graph
→ peptide candidate
→ mechanism proof
→ assay proposal
→ auditable promotion path

2. Verified Epigenome Rescue / longevity biology

The second vertical extends the same substrate into a much more ambitious domain: verified epigenome-rescue peptide discovery. The longevity angle is not a generic anti-aging claim. It is a falsifiable cell-state rescue program inspired by epigenomic-fidelity theories of aging: preserve or restore identity, chromatin order, function, and senescence balance while refusing unsafe reprogramming paths.

DiscoveryLab treats this not as a slogan, but as a falsifiable discovery program:

Can we discover peptides that restore a declared aged cell state toward a safer youthful identity attractor by modulating chromatin readers, repair-chromatin coupling, metabolic-to-epigenome signaling, senescence pathways, or nuclear delivery — without triggering dedifferentiation, DNA damage, oncogenic proliferation, or broad chromatin disruption?

• histone-tail mimetic peptides;
• chromatin-interface modulators spanning readers, erasers, and writers;
• SIRT6 / heterochromatin fidelity candidates;
• DNA-repair–chromatin coupling peptides;
• metabolic-epigenome signaling peptides;
• senescence-state peptides;
• nuclear-targeted delivery motifs.

The objective is not vague "anti-aging." It is verified epigenomic fidelity:

restore cell identity
+ improve chromatin order
+ rescue function
+ reduce senescence / inflammatory drift
- avoid pluripotency
- avoid DNA damage
- avoid cancer-like proliferation
- avoid global chromatin disruption
- reject clock-only wins

The benchmark path is EpiCheck v1: each row should resolve to promote, revisit, rejected_safety, or rejected_clock_only before the company makes any longevity efficacy claim.

This vertical proves DiscoveryLab is not hardcoded to one biology story. The same agent-native, mechanism-verified infrastructure can move from inflammation to epigenome repair without forking the platform.

Production-grade platform

DiscoveryLab is already built as a deployable system, not a slideware concept. Single deployable binary, 240+ passing tests, registered biology graphs, and an agent-driven loop demonstrated end-to-end.

Two discovery verticals

Original inflammation / IL-23 program and a second verified epigenome-rescue program for longevity biology. Same core platform. Different biology. No engineering fork.

Closed recursive discovery loop

The system tracks calibration and regression metrics, proposes typed improvements, routes those improvements through evaluation and approval, and preserves auditability across promotion and rollback.

Private-infrastructure thesis

Designed for the procurement reality of pharma: discovery agents must work where the data already lives.

• We have not yet run a prospective wet-lab assay against a non-fixture candidate.
• We are not claiming clinical validation.
• We are not claiming epigenetic rejuvenation.
• We are not claiming lifespan extension or healthspan extension.
• We are not claiming to be the broadest protein-design platform.
• We are not trying to outspend the large generative-biology companies.

The next milestone is obvious: prospective wet-lab validation of a platform-generated candidate. That is the highest-leverage use of capital.

Generate:Biomedicines has raised nearly $700M in equity financing since 2020 and describes its platform as machine-learning-powered generative biology across protein therapeutics. Recursion, Insitro, Schrödinger, Cradle, and others are pursuing image-based phenotyping, computational biology, physics-based modeling, generative protein design, or large-scale discovery automation.

Peptiter's wedge is different:

agent-native
+ private infrastructure
+ peptide-first
+ mechanism-verified
+ audit-by-construction
+ human-gated lab execution
+ recursive evidence improvement

We are not trying to be the biggest generator. We are trying to be the most defensible discovery system for organizations that need AI agents to work inside controlled environments, produce reviewable scientific claims, and support regulator-facing audit chains.

1. Pharma adoption of AI agents has hit the procurement-blocker wall.

Hosted demos are easy. Production deployment inside a pharmaceutical organization is not — data residency, IP control, validation, and audit requirements stop almost every vendor at procurement. The winning system has to respect data boundaries from day one.

2. Regulatory expectations for AI-derived submissions are sharpening.

FDA and EMA are actively building guidance for AI use across drug development and the medicines lifecycle. Platforms that can produce audit chains — not just rankings — will be the ones whose outputs survive contact with a regulatory package.

3. The model context layer for agents has standardized.

MCP-style context standards make it practical, for the first time, to deploy a multi-agent discovery loop on infrastructure the customer already controls — without rebuilding the agent stack per environment. Private-infrastructure agent deployments have moved from "research" to "shippable."

AI agent capability
+ private deployment
+ formal mechanism review
+ audit-chain architecture
+ wet-lab learning loop
= 12–18 months ahead of obvious copies

1. Prospective wet-lab validation

Run the first prospective assay with a partner CRO against a non-fixture peptide candidate. Converts the platform from in-silico rigor to biological credibility.

one candidate
one mechanism hypothesis
one assay plan
one pre-registered success criterion
one audited result

2. First private pharma deployment

Deploy DiscoveryLab inside a pharma customer environment, driving an existing peptide or pathway program behind the customer's firewall.

agent-driven discovery workflow
private data retained in customer environment
mechanism proof receipts generated
candidate decisions audited

3. Regulatory-grade audit-chain package

Develop the documentation layer that compliance teams can use when AI-influenced discovery decisions become part of later regulatory submissions.

claim lineage
model versioning
mechanism proof receipts
human approvals
assay provenance
promotion/rollback history
candidate decision log

4. Verified Epigenome Rescue proof-of-concept

Advance the longevity vertical from internal scaffold to a benchmarked, assay-ready campaign.

• SIRT6 / heterochromatin fidelity peptides;
• chromatin-interface histone-tail and protein-interface mimetics;
• senescence-state peptide modulators;
• metabolic-to-epigenome signaling peptides.
• EpiCheck v1 benchmark rows for promote / revisit / safety-reject / clock-only-reject outcomes.

cell type selected
aging failure mode defined
peptide intervention class selected
epigenomic readouts specified
safety gates encoded
EpiCheck row drafted
candidate set generated
assay plan ready

1. Enterprise deployment license

On-prem or private-cloud deployment for pharma and biotech teams.

2. Discovery partnership

Joint peptide programs where Peptiter contributes the platform, mechanism verification, and candidate generation.

3. Internal asset creation

Use the platform to generate proprietary peptide candidates in selected verticals, starting with inflammation and epigenome rescue.

The first revenue path should be enterprise deployment or paid pilot. The largest upside path is internal asset creation once the first wet-lab validation is complete.

Use of funds:

• Prospective wet-lab validation with a partner CRO.
• Two senior biology hires to convert platform workflows into pharma-ready discovery programs.
• Private deployment readiness for the first on-prem or private-cloud customer.
• Regulatory audit-chain package to lead the conversation while the category is still forming.
• Verified Epigenome Rescue program design to convert the longevity vertical into a benchmarked, testable peptide campaign.

The platform is built. The next milestone is wet-lab credibility. The longer-term opportunity is a new category: mechanism-verified, private-infrastructure AI discovery.

Most AI drug discovery platforms ask pharma to trust the model. Peptiter takes the opposite position:

Do not trust the model. Verify the mechanism, audit the decision, gate the experiment, and learn from every result.

That is why peptides are the right first modality, why private deployment matters, why formal verification is not academic decoration, and why the Recursive Discovery Loop is the core product.

Peptiter is building the discovery platform for the moment when AI agents enter pharma — and pharma asks the only question that matters: Can we run it here, can we audit it, and can we trust the decisions it helps us make?