Nonclinical · v1

Peptide research,
replayable.

Twelve agents. Hashed before. Revealed after.

Read the protocol
The Loop5 stages
  1. 01Commit
  2. 02Ingest
  3. 03Reason
  4. 04Critique
  5. 05Reveal
commit · reveal · verify
12
Agents
10
Stages
A — X
Grades
sha-256
Commit

Five stages.
One hash.

Protocol
  • 01
    Commit
  • 02
    Ingest
  • 03
    Reason
  • 04
    Critique
  • 05
    Reveal

Twelve specialists.

Architecture
  • 01

    Literature Miner

    upstream · PubMed
  • 02

    Sequence & Structure

    upstream · UniProt · AlphaFold
  • 03

    Target & Pathway

    upstream · OpenTargets
  • 04

    Variant Linker

    upstream · ChEMBL
  • 05

    ADMET Developability

    upstream · Heuristic
  • 06

    Novelty Scout

    reasoning · Whitespace
  • 07

    Patent Competitive

    reasoning · Lens.org
  • 08

    Thesis Generator

    reasoning · kie · gpt-5-4
  • 09

    Evidence Grader

    reasoning · A — X rubric
  • 10

    Red Team

    reasoning · 3 personas
  • 11

    Synthesizer

    output · Cross-pool
  • 12

    Dossier Assembler

    output · Markdown

The output is a single dossier.

Output
dossier · schema
markdown
## Question
<the original mission query, verbatim>

## Cross-pool consensus
- Literature       PubMed citations
- Sequence         UniProt + AlphaFold
- Target/pathway   OpenTargets
- Chemistry        ChEMBL ligand prior

## Open questions
## Risks
## Recommended next steps
sha-256 commit per runsalt revealed on completion
  • Cited.
    Every claim has a real PMID.
  • Hedged.
    No human-use claim, ever.
  • Hashed.
    Replayable end-to-end.

Built so you don’t have to trust us.

Principles
  1. 01

    Falsifiability is a first-class artifact.

  2. 02

    Evidence is graded, not gestured at.

  3. 03

    The red team can actually block.

  4. 04

    A clean falsification is a successful run.

Now

Watch the swarm work.