Ridges Uses Open Incentivized Tournament to Produce Intelligence

The video features a discussion between Sammy from Unsupervised Capital and Shaq, the founder of Ridges (SN62), a subnet on the Bittensor. The conversation focuses on Ridges’ achievement in developing a state-of-the-art software engineering agent through an open, incentivized tournament model.

Key points include:

1. Background and Journey: Shaquille shares his unconventional path from dentistry to coding, trading, and eventually founding Ridges. He emphasizes his lack of formal AI experience and venture capital, yet managed to create a leading agent with just three interns.
2. Ridges’ Achievement: In less than four months, Ridges produced an agent that outperforms both open-source and closed-source models like Anthropic’s Claude 4 on the SWE-Bench dataset. This was achieved without any external funding, showcasing the power of community-driven innovation.
3. Incentive Mechanism: The project operates on a winner-takes-all model where developers submit agents daily, with the top performer earning a significant prize pool (currently $55,000 per day). This continuous competition drives rapid improvements as submissions are open-sourced, allowing others to build upon them.
4. Economic and Open-Source Impact: Ridges addresses traditional open-source challenges like funding and alignment by using tokenomics. Contributors earn tokens that represent ownership in the system, incentivizing long-term commitment and collaboration. The project has distributed over $1.5 million in rewards, attracting global talent and fostering a competitive environment.
5. Cost Efficiency: The Ridges agent is remarkably cost-effective, solving all 500 SWE-Bench problems for just $1.26, compared to $94 for Claude Opus, highlighting a 90% cost reduction.
6. Future Plans: The team aims to develop a product similar to Claude Code or OpenAI’s Codex, leveraging the cost efficiency and performance of their agent. They also plan to explore new user experiences that integrate AI more seamlessly, potentially transforming how coding agents are used.
7. Challenges and Insights: Shaq discusses the initial challenges of operating in a competitive environment where participants try to exploit the system, but notes that stripping down to pure incentives can lead to more efficient outcomes than traditional corporate structures.
8. Bittensor’s ($TAO) Role: The video underscores Bittensor’s potential as a platform for innovative AI projects, with Ridges possibly becoming a flagship example of crypto-AI success.