Proposed · Design proposal
Reproducible task execution
Where do cross-device differences originate?
Freeze inputs, dependencies and Runner versions; compare row-level outputs and failures.
Genesis / Prototype Stage / RESEARCH
Keep important questions
within reach of evidence.
These directions come from unresolved protocol questions. They describe research to be conducted, not conclusions already reached.
RESEARCH AGENDA
Keep important questions
within reach of evidence.
These directions come from unresolved protocol questions. They describe research to be conducted, not conclusions already reached.
Proposed · Design proposal
Evidence and provenance
How much context does acceptance need?
Test whether provenance, revocation propagation and exports support outside review.
Proposed · Design proposal
Verification cost
Could review cost more than execution?
Measure reference reruns, human support and failure recovery across the full workflow.
Proposed · Design proposal
Contribution attribution
How do authors, operators and reviewers differ?
Separate roles and control relationships; evaluate copying, outsourcing and collusion.
Proposed · Design proposal
Controlled local AI execution
Can bounded local AI workloads be reproduced?
After the text-rule loop passes, study versions and measurement policies for authorized small-model tasks.
Proposed · Design proposal
Open network security
What attacks emerge with wider participation?
Test lease abuse, duplicate submissions, malicious inputs, leakage and verifier collusion.
Proposed · Design proposal
Economic necessity before tokenization
Why use a Token when ordinary settlement is enough?
Compare real demand, operating costs and ordinary payments before justifying an economic layer.
A research output should expose its question, method, input provenance, failure conditions and reviewable evidence. Completing a design document does not complete an experiment.
GET INVOLVED
Start with a concrete
piece of work.
Formal testing is not open yet. You can prepare the following materials or review the existing design.
For task submitters
Task submitters
Prepare a low-risk, licensed offline task with clear acceptance criteria. List input permissions, rule versions, expected results and resource limits.
For developers and researchers
Developers and researchers
Review the protocol, Runner limits, data model, recovery and verification methods. State the issue, reproduction conditions and supporting evidence.
For controlled testers
Controlled testers
Join cross-device reproduction after resource and permission boundaries are established. Prepare environment details, a resource budget and exit conditions.