𝐒𝐮𝐛𝐧𝐞𝐭 𝟔𝟐 𝐑𝐢𝐝𝐠𝐞𝐬 – 𝐎𝐩𝐞𝐧 S𝐨𝐮𝐫𝐜𝐞 A𝐠𝐞𝐧𝐭𝐬 O𝐮𝐭𝐩𝐞𝐫𝐟𝐨𝐫𝐦𝐢𝐧𝐠 𝐭𝐡𝐞 W𝐨𝐫𝐥𝐝

Here’s a tweet made by RvCrypto on X:

In less than four months since launch, Ridges has become the clear leader in building state-of-the-art AI coding agents. By harnessing Bittensor’s incentive mechanisms and open-source competition, Ridges is driving an unprecedented rate of progress, and now outperforms every other effort in the world, including those backed by companies worth hundreds of billions.

1. New state-of-the-art

Latest Ridges agent score: 86.4% on SWE-Benchmark. 𝐓𝐡𝐢𝐬 𝐢𝐬 𝐭𝐡𝐞 𝐡𝐢𝐠𝐡𝐞𝐬𝐭 𝐬𝐜𝐨𝐫𝐞 𝐞𝐯𝐞𝐫 𝐚𝐜𝐡𝐢𝐞𝐯𝐞𝐝 𝐨𝐧 𝐭𝐡𝐞 𝐛𝐞𝐧𝐜𝐡𝐦𝐚𝐫𝐤, 𝐚𝐡𝐞𝐚𝐝 𝐨𝐟 𝐚𝐥𝐥 𝐨𝐩𝐞𝐧-𝐬𝐨𝐮𝐫𝐜𝐞 𝐩𝐫𝐨𝐣𝐞𝐜𝐭𝐬 𝐚𝐧𝐝 𝐞𝐯𝐞𝐧 𝐩𝐫𝐨𝐩𝐫𝐢𝐞𝐭𝐚𝐫𝐲 𝐥𝐚𝐛𝐬 𝐥𝐢𝐤𝐞 𝐀𝐧𝐭𝐡𝐫𝐨𝐩𝐢𝐜 𝐚𝐧𝐝 𝐎𝐩𝐞𝐧𝐀𝐈.

Ridges is now proving that open-source agents, powered by decentralized incentives, can outperform Big Tech.

2. Explosive miner & agent growth

With rewards up to $40k/day, 300+ new agents are submitted daily of which compute partners Chutes and Targon handle 400M+ inference tokens per day, growing the ecosystem from the inside out.

All agent code is open source, miners compete and collaborate, absorbing each other’s innovations at record-breaking speed.

3. Incentives innovation v4

In the new incentive system, a new agent has to beat the top agent by a clear performance threshold. Over time, the threshold decays, so no agent monopolizes emissions forever. This balances rewards for big breakthroughs and incremental gains, keeping the improvement flywheel spinning.

4. Mixed evaluations = No shortcuts

The new mixed evaluations system not only relies on the SWE-Bench but also on Polyglot. New agents must be top across multiple benchmarks to claim #1.

Result: only true generalist agents that can reason, debug, and adapt across domains will rise to the top.

5. Next: real products and revenue

Benchmarks are just a proxy. The real test: can these agents create value in production?

First product: Ridges Terminal → putting coding agents directly into developers hands.

I think this will turn benchmark dominance into enterprise adoption and revenue.

𝐂𝐨𝐧𝐜𝐥𝐮𝐬𝐢𝐨𝐧

Ridges is no longer just catching up — it’s leading the world, with open-source agents scoring 86.4% and surpassing even Big Tech firms worth hundreds of billions.

With Incentive v4, Mixed evaluations, and the Ridges Terminal on the horizon, this is the clearest proof yet that decentralized, incentive-driven AI can outperform traditional labs and deliver real-world value. 𝐎𝐧𝐜𝐞 𝐥𝐢𝐯𝐞 𝐭𝐡𝐢𝐬 𝐢𝐬 𝐠𝐨𝐢𝐧𝐠 𝐭𝐨 𝐛𝐞 𝐚 𝐫𝐞𝐯𝐞𝐧𝐮𝐞 𝐦𝐨𝐧𝐬𝐭𝐞𝐫.