Cortex is building an autonomous research network where reproducible findings can improve shared methods, rather than remain isolated model checkpoints.
| Audience | Read first | Next |
|---|---|---|
| Investors and new readers | Project overview | Current capabilities and limits |
| Research contributors | Getting started | Proof |
| Bug hunters | Bounty | Troubleshooting |
| Validators | Validator guide | Architecture |
| Developers | Contributing | Architecture, naming and compatibility |
| Operators | Deployment | Runbooks, security checklist |
These documents keep the exact rules needed to operate and verify the network. You do not need to read them to understand the project.
- Proof: topic publication, evaluation, and research rewards.
- Bounty: pairing, bug reports, adjudication, and rewards.
- Bundle specification: signed results and reward verification.
- Site API: public website data.
- Threat model: security guarantees and their limits.
- Repository cleanup notes: what was removed and what must stay.
The whitepaper, A Proposal for a New Incentive Mechanism on Cortex, explains why the network exists. Start with §2 for the checkpoint and static-evaluator critique, and §7 for research reuse and the proposed synthesis agent.
The plain-language reading guide and implementation comparison separates that proposal from current code. The paper is a primary source for the vision, not a deployment-readiness report or a replacement for current scoring contracts.
Design, Prism, Prism recipes, the retired miner pages, evidence, and experiments provide historical background. They are not a list of open challenges.
Frozen specifications and operational evidence remain available for verification and old links. Use the guides above for current participation and deployment.