Open research / independently reproducible

The apparatus,
in the open.

A static public site and a separate research toolkit. Inspect the code, preserve the provenance, and test the boundaries before collecting evidence.

Start with the source

A small, inspectable
research system.

The repository contains the artistic materials, protocols, allocation and configuration tools, schemas, tests, and deployment documentation. The public website holds no provider credentials.

Open the repository ↗

A dry run validates software structure. It does not demonstrate an effect, prove provider isolation, or authorize a real study.

01
Understand the design.
Read the protocol, scientific objections, and the study runbook before configuring a run.
Study runbook ↗
02
Inspect the configuration.
Use the toolkit’s deterministic research allocator. Pin the provider, exact model, reasoning effort, sampling settings, tasks, and memory policy.
Dry-run example configuration ↘
03
Validate the boundaries.
Test state isolation, corpus hashes, trace completeness, context assembly, and blinded export. A local dry run makes no inference calls. The current trace schema is diagnostic; a validated live trace contract and provider adapter remain prerequisites.
Diagnostic trace schema ↘
04
Freeze before collecting.
Complete the pilot launch gates. Preregister confirmatory hypotheses, analysis, exclusions, and a justified sample size before confirmatory data collection.
Preregistration draft ↗