Inside the runtime
Full discovery lifecycle.
Protean is a controlled scientific runtime. It acquires evidence, forms memory, asks hypotheses, designs computational experiments, generates constrained candidate fields, validates them, and packages review-ready research artifacts.
runtime contract
Bounded autonomous research pass
Autonomous runtime walkthrough
Inside Protean’s full discovery lifecycle.
Protean operates as a controlled scientific runtime: evidence enters, memory forms, hypotheses shape computational experiments, candidates move through deterministic gates, and each cycle becomes reviewable research state.
50
Ranked candidates
latest bounded cycle
15
Candidate explanations
top review set
6
Hypotheses
reviewable research questions
43
Sequence clusters
mapped candidate field
32
Source catalog
provenance-scored sources
6
Runtime modes
planned research emphasis
active stage 01
Scientific Data Acquisition
Literature, evidence records, failures, patents, peptide databases, and planned assay or structure sources enter as provenance-bearing records.
A scientific runtime is only as credible as its source discipline. Protean treats source origin, freshness, duplication, and contradiction as part of the research state.
input
papers, protein records, failure signals, patent context
control
source trust scoring, deduplication, bounded ingestion
output
curated evidence plane
runtime causality map
boundedcycle contract
Control surfaces
Autonomy works because the system is bounded.
The lifecycle is designed around controlled scientific motion. Each adaptive layer writes artifacts, warnings, and review context rather than silently changing the engine underneath it.
Deterministic authority
Proposal systems can suggest candidates and language, but deterministic validators, residue rules, warning burden, and scoring contracts decide what advances.
Failure memory
Rejected motifs, instability patterns, degradation signals, contradictions, and negative evidence remain active signals in future prioritization.
Bounded adaptation
Learning adjusts prioritization conservatively, logs prior state and rationale, preserves normalized weights, and never rewrites scoring logic.
Artifact ledger
Every cycle leaves a trail.
Protean’s public explanation can be sophisticated without exposing moat-critical internals. The important surface is the controlled chain of state: evidence, candidates, rankings, explanations, learning reports, hypotheses, experiments, provenance, and handoff packages.
01
Acquire
curated evidence plane
02
Remember
motif, contradiction, lineage, and exploration memory
03
Hypothesize
bounded hypothesis set
04
Plan
reviewable computational experiment plans
05
Generate
candidate field
06
Validate
validated candidate set
07
Rank
ranked candidate slate
08
Explain
briefs, reports, and candidate assessment papers
09
Handoff
review-ready handoff batch
10
Trace
provenance-aware review package
11
Learn
conservative ranking adjustment
12
Orchestrate
next investigation plan
Scientific boundary
Computational prioritization is not wet-lab proof.
The runtime improves review quality by making uncertainty visible. Candidate rankings, generated assessments, and hypothesis plans remain computational artifacts until controlled assays, comparison groups, and human scientific review establish downstream evidence.