A Bittensor Subnet Proposal | Subnet Ideathon Round I
Live Dashboard • Technical Documentation • GitHub Source
ReasonForge is a next-generation Bittensor subnet designed to incentivize structured, auditable reasoning chains. While existing AI subnets often prioritize raw generation speed or generic quality, ReasonForge focuses on the integrity of the thought process.
By leveraging a mathematically rigorous protocol, ReasonForge transforms raw compute into Verifiable Intelligence. Whether it's complex logical proofs, causal inference, or high-stakes financial modeling, ReasonForge ensure that every answer is backed by a transparent, verifiable chain of logic.
- Verifiable Chains: Every response includes an auditable reasoning path.
- Mathematical Rigor: 13 unique formulas governing every aspect of the network.
- Validator Accountability: Quadratic slashing and reputation-weighted rewards.
- Trap Detection: Integrated mechanisms to catch lazy or malicious actors.
Experience the power of the ReasonForge protocol locally in minutes.
Clone the repository and install the core protocol package:
pip install -e .
python -m reasonforge.run --epochs 10 --emission 100 --output results.json --seed 42Spin up the FastAPI server to interface with the simulation engine:
pip install fastapi uvicorn
uvicorn api.server:app --reload --port 8000Launch the high-fidelity visualization suite:
npm install
npm run devReasonForge is built on a modular "Separation of Concerns" architecture, ensuring scalability and ease of integration.
| Component | Tech Stack | Responsibility |
|---|---|---|
| Core Engine | Python |
Implementation of all 13 whitepaper formulas. |
| Epoch Simulator | Python |
Full-cycle network simulation (Miners/Validators). |
| Interactive UI | React/TS |
Real-time visualization of network health and rewards. |
| Plagiarism Guard | Python |
Jaccard-based similarity detection for network integrity. |
| Validation Layer | Pytest |
Comprehensive suite of 60+ tests for formula accuracy. |
ReasonForge is governed by 13 core formulas that ensure game-theoretic stability and fairness.
- Q: Path Quality
- A: Accuracy
- N: Novelty
- Eff: Efficiency
We simulate a diverse range of 12 miners to test protocol resilience:
- Elite (m-001, m-002): High-performance reasoning nodes.
- Adversarial (m-011, m-012): Testing plagiarism and spam detection.
6 distinct validator archetypes:
- Honest (v-001, v-002): High stake, high accuracy.
- Lazy/Malicious (v-005, v-006): Triggering slashing and penalty mechanisms.
- Phase 1: Core Mathematical Engine & 13 Formulas.
- Phase 2: Multi-Epoch Simulator & CLI.
- Phase 3: FastAPI Integration & SSE Streaming.
- [wip] Phase 4: Mainchain Integration & Bittensor Onboarding.
This project is licensed under the MIT License.