Shakeel Hussein, founder of Ridges Subnet 62, joins the podcast to unpack how his open-source, incentive-driven approach is creating AI coding agents that are already outperforming competitors like Claude Code. He details his plan to make human coders obsolete and introduces a novel "alpha-to-equity" model designed to bridge the gap between crypto tokens and traditional venture capital.
The AI Coder Revolution
Incentives Fueling Innovation
From Alpha to Equity: A New Playbook
Key Takeaways:
For further insights and detailed discussions, watch the full podcast: Link
This episode reveals how Ridges Subnet 62 is not only achieving state-of-the-art performance in AI-driven software engineering but is also pioneering a novel "alpha to equity" model to directly link token value with company profits.
Introduction: The Race to Automate Software
Shakeel Hussein, founder of Ridges Subnet 62, discusses the rapid progress since open-sourcing his miner competition. He notes that this decision was one of the best they've made, fostering a dynamic where miners both collaborate and compete, building on each other's ideas to accelerate agent quality.
Despite the subnet's meteoric rise to the #2 rank on Bittensor, Shakeel maintains a relentless focus on speed and execution. When asked how his younger self would view his recent success, he remains grounded: “Not much really. I guess just like could keep going faster.”
The Vision for a Ridges Product
Shakeel outlines the initial product vision, which resembles tools like Claude Code. Users would have an application connected to their codebase where they can assign problems to a Ridges agent and observe its entire thought process—from file analysis to final code implementation.
A key challenge is accounting for subjective human preferences in design, which current benchmarks don't capture. The long-term plan is to integrate real user feedback directly into the subnet's incentive mechanism. This would involve running multiple top agents for user queries, tracking which agent's solution is preferred, and rewarding miners based on this real-world performance data rather than just benchmark scores.
Evolving the Incentive Mechanism
Before a full product launch, Shakeel details several upcoming upgrades to the incentive mechanism designed to ensure agents are genuinely capable and not just overfitted to a single benchmark.
The Role of LLMs and Compute Costs
The conversation explores the trade-offs between the underlying Large Language Model (LLM) and the agent's architecture. While more powerful models like Claude's provide a slight performance boost, Shakeel notes they are prohibitively expensive—around 380 times the current cost.
Ridges' core strategy is to build a "thick" agent layer on top of a "thin," cost-effective model layer. This contrasts with competitors who bet entirely on expensive, state-of-the-art models, leading to what Shakeel describes as "massively negative gross margins."
The Path to Fully Autonomous Software Engineering
Shakeel believes that current LLMs are already sufficient to, in the limit, fully automate software engineering. He observes that the internal monologue of the agents is beginning to resemble his own thought process when coding.
The agents currently operate in a highly constrained sandbox environment with no web access, forcing them to rely purely on their own capabilities. Shakeel confirms that removing these "handcuffs," particularly by enabling web search, would provide the single biggest performance boost, allowing agents to access context from sources like Stack Overflow and technical blogs.
A Novel Tokenomic Model: Alpha to Equity
Shakeel introduces a groundbreaking "alpha to equity" plan designed to solve key tokenomic challenges and align incentives between the company, miners, and token holders. This model moves beyond simple buyback-and-burn mechanisms.
Building the Team and Overcoming Challenges
Shakeel discusses his unique hiring strategy, which he calls the "Waterloo funnel." By offering compensation competitive with Canadian firms but lower than top US firms, he attracts elite interns from the University of Waterloo who are highly skilled but may not wish to move to the US. His process involves personally reviewing the GitHub profiles of nearly a thousand applicants.
The biggest operational challenge has been maintaining system stability while pushing frequent updates. Ensuring fair and consistent evaluation for miners is a top priority, and Shakeel sometimes compensates miners from his own wallet when system downtime affects their rewards.
Conclusion: A New Blueprint for Crypto AI Value
This episode highlights a dual innovation: Ridges is building a state-of-the-art AI agent to automate coding while architecting a tokenomic model that directly converts decentralized network value into traditional equity. Investors should monitor the alpha-to-equity implementation and multi-benchmark performance as key indicators of a new value creation model.
