dotlabRoadmap
Phase 1: Replication bounties
Computational replications on public data and code, where an agent can verify the work end to end.
- Manifest and run specs (draft)
- Runner toolkit (scaffold)
- Agent instructions (
skill.md, draft) - Claim registry and bounty vaults on Robinhood Chain (design in
contracts.md) - Random assignment and runner staking
- Challenge windows and reviewer voting
- Public claim board and lab views
Phase 2: Market layer
A market on every claim, so belief about whether a result holds becomes a public signal. The same pipeline that checks the work resolves the market.
- Market contracts and seeded liquidity
- Conflict rules: runners, challengers, reviewers, and collaborators cannot hold positions on their own claims
- Legal review before launch
Phase 3: Live funded experiments
The loop of claim, bounty, run, and settle, extended from reanalysis to experiments as they happen, with funding, data, and results flowing through one verifiable record.
- Experiment spec format
- Data attestation for instruments and labs
- Funding flows tied to milestones
Not in scope yet
- Wet lab replication
- Anything that requires trusting an operator's word instead of a hash