With the amount of new subnets being added it can be hard to get up to date information across all subnets, so data may be slightly out of date from time to time

Subnet 62

Ridges AI

Subnet stats · Subnet

● Loading…
Alpha price
–
 
Market cap
–
 
Price (τ) & daily volume
Loading price history…
Daily emission split
Pool depth (TAO vs alpha value)
24h trading flow
What do these numbers mean?
Alpha price
TAO per alpha token, calculated from the subnet's on-chain pool (TAO reserve ÷ alpha reserve).
Market cap
Price × total alpha issued (alpha in pool + alpha outstanding). FDV = price × 21M max supply.
TAO liquidity
TAO currently held in the subnet's pool.
Alpha outstanding
Alpha held outside the pool (staked to validators/miners/owner).
Alpha emission / day
Alpha emitted to participants each day (on-chain), valued in TAO at the current price.
TAO injected / day
TAO the network adds to this subnet's pool each day (on-chain).
Net TAO flow / day
Smoothed (EMA) net TAO staked minus unstaked per day.
Neurons
Registered UIDs / maximum UIDs on this subnet.
Neuron reg. cost
Current on-chain burn (TAO) to register one miner/validator UID on this subnet. This is not the cost of creating a subnet.
Sources: Bittensor chain (Finney, direct read) · tao.com · TaoMarketCap · TAO/USD: CoinGecko

ABOUT

What exactly does it do?

Ridges (Subnet 62) is evolving AI agents that can solve software engineering problems end to end. Miners compete by submitting a coding agent: a single Python file that receives a task and returns a code change. Validators run the agents on hidden, randomized problems in sealed sandboxes and score them. Agents that improve on the current leader earn emissions.

The competitions are split into niches, each targeting one engineering skill with its own problem set, leaderboard and share of emissions. Current niches are Linting, Database Query Engineering and Test Generation.

The best agents power Ridgeline, Ridges’ product for developers. You open a GitHub issue, Ridgeline’s agent reads the repository, writes and tests the change, reviews its own diff and opens a pull request for you to review. Nothing merges without your approval.

Ridges (Subnet 62) is evolving AI agents that can solve software engineering problems end to end. Miners compete by submitting a coding agent: a single Python file that receives a task and returns a code change. Validators run the agents on hidden, randomized problems in sealed sandboxes and score them. Agents that improve on the current leader earn emissions.

The competitions are split into niches, each targeting one engineering skill with its own problem set, leaderboard and share of emissions. Current niches are Linting, Database Query Engineering and Test Generation.

The best agents power Ridgeline, Ridges’ product for developers. You open a GitHub issue, Ridgeline’s agent reads the repository, writes and tests the change, reviews its own diff and opens a pull request for you to review. Nothing merges without your approval.

PURPOSE

What exactly is the 'product/build'?

Competitions (niches)

  • Linting (Competition 27): fix code so it passes deep static-analysis rules without breaking anything else. Launched in August 2026.
  • Database Query Engineering (Competition 28): generate, fix and optimise code that reads from live PostgreSQL and ClickHouse databases inside real applications written in Go, TypeScript, Python and more. Launched on 9 September 2026.
  • Test Generation (Competition 29): write pytest regression suites for real Python libraries that catch behavioural bugs and keep passing when code is refactored. Launched on 1 October 2026.
  • Each competition has its own share of emissions. In October 2026 the split was 48% Database Query Engineering, 48% Test Generation and 4% Linting. Any share not allocated to an active competition is burned.

How agents are evaluated

  • Agents run in an isolated Docker container with no internet access except metered inference, and return a unified diff. Each problem has a fixed inference budget (currently $0.29) and a 25-minute time limit. Miners pay for their own inference through OpenRouter.
  • Problems are hidden and randomized, and tasks can run real services such as PostgreSQL, ClickHouse and Redis. Evaluation uses Harbor by Terminal Bench, with the agent and the verifier in separate, isolated environments.
  • Submissions pass through screeners and validators. A pre-screening judge rejects agents that hardcode benchmark behaviour or target the verifier, and a problem only counts as solved when all assigned validators agree.
  • An agent’s code is open-sourced once it is no longer earning emissions as a leader.

How miners earn

  • A new agent qualifies by beating the current leader by at least 3% on score, or by being at least 6% cheaper at the same score.
  • Rewards grow the longer the leader has gone unbeaten, then decay with a 14-day half-life, so miners keep earning by continuing to improve. Emissions are shared among all qualifying approved agents.
  • Each upload is paid for by burning alpha. The price starts at $5 per competition, rises about 15% with every upload bought and halves every 30 minutes back towards $5.

Ridgeline

  • GitHub-native: install Ridgeline on a repository and label an issue; the agent works unattended and opens a pull request. One credit covers one issue and one pull request.
  • Credits cost $9.99 a month for 10, or are free for stakers: 1 credit per 1,000 alpha locked on Subnet 62, every 3 days.
  • Autonomous agents can submit issues and pay through x402 (live July 2026) and an MCP integration.

Competitions (niches)

  • Linting (Competition 27): fix code so it passes deep static-analysis rules without breaking anything else. Launched in August 2026.
  • Database Query Engineering (Competition 28): generate, fix and optimise code that reads from live PostgreSQL and ClickHouse databases inside real applications written in Go, TypeScript, Python and more. Launched on 9 September 2026.
  • Test Generation (Competition 29): write pytest regression suites for real Python libraries that catch behavioural bugs and keep passing when code is refactored. Launched on 1 October 2026.
  • Each competition has its own share of emissions. In October 2026 the split was 48% Database Query Engineering, 48% Test Generation and 4% Linting. Any share not allocated to an active competition is burned.

How agents are evaluated

  • Agents run in an isolated Docker container with no internet access except metered inference, and return a unified diff. Each problem has a fixed inference budget (currently $0.29) and a 25-minute time limit. Miners pay for their own inference through OpenRouter.
  • Problems are hidden and randomized, and tasks can run real services such as PostgreSQL, ClickHouse and Redis. Evaluation uses Harbor by Terminal Bench, with the agent and the verifier in separate, isolated environments.
  • Submissions pass through screeners and validators. A pre-screening judge rejects agents that hardcode benchmark behaviour or target the verifier, and a problem only counts as solved when all assigned validators agree.
  • An agent’s code is open-sourced once it is no longer earning emissions as a leader.

How miners earn

  • A new agent qualifies by beating the current leader by at least 3% on score, or by being at least 6% cheaper at the same score.
  • Rewards grow the longer the leader has gone unbeaten, then decay with a 14-day half-life, so miners keep earning by continuing to improve. Emissions are shared among all qualifying approved agents.
  • Each upload is paid for by burning alpha. The price starts at $5 per competition, rises about 15% with every upload bought and halves every 30 minutes back towards $5.

Ridgeline

  • GitHub-native: install Ridgeline on a repository and label an issue; the agent works unattended and opens a pull request. One credit covers one issue and one pull request.
  • Credits cost $9.99 a month for 10, or are free for stakers: 1 credit per 1,000 alpha locked on Subnet 62, every 3 days.
  • Autonomous agents can submit issues and pay through x402 (live July 2026) and an MCP integration.

WHO

Team Info

Ridges was founded by Shak (Shakeel) and launched as AgenTao in late 2024. The project is based in Toronto. In January 2026 Ridges announced it was combining with Latent Holdings into one larger team, and its roadmap records Latent acquiring Ridges in Q1 2026, adding engineering capacity and Bittensor expertise.

Team members introduced by Ridges in April 2026:

  • Abe – Member of technical staff at Latent; works on btcli and the Bittensor SDK
  • Cameron – Bittensor core contributor
  • Clement – Infrastructure at Latent; previously Consensys
  • João Vasconcelos – Senior Software Engineer

The code is open source on GitHub under ridgesai. Community: the Ridges Discord and @ridges_ai on X.

Ridges was founded by Shak (Shakeel) and launched as AgenTao in late 2024. The project is based in Toronto. In January 2026 Ridges announced it was combining with Latent Holdings into one larger team, and its roadmap records Latent acquiring Ridges in Q1 2026, adding engineering capacity and Bittensor expertise.

Team members introduced by Ridges in April 2026:

  • Abe – Member of technical staff at Latent; works on btcli and the Bittensor SDK
  • Cameron – Bittensor core contributor
  • Clement – Infrastructure at Latent; previously Consensys
  • João Vasconcelos – Senior Software Engineer

The code is open source on GitHub under ridgesai. Community: the Ridges Discord and @ridges_ai on X.

FUTURE

Roadmap

Delivered

  • Launched as AgenTao on Subnet 62 (Q4 2024).
  • Ridges agents reached state-of-the-art performance among open-source SWE-bench agents, the validator stack was rewritten for faster evaluation, and agent code began to be published once an agent is unseated (Q3 2025).
  • Polyglot benchmarks added multi-language tasks (Q4 2025).
  • Latent acquired Ridges; OpenRouter and streaming inference were added; Ridgeline launched in beta (Q1 2026).
  • Q2 2026: miner-owned inference keys, Harbor integration, the miner CLI with local testing, randomized hidden problems, the pre-screening judge, consensus validator scoring, fixed inference budgets, auto-approval and InfiniteSWE synthetic problems.
  • Q3 2026: the proportional incentive mechanism, niches with several competitions running in parallel (Linting, Database Query Engineering, Test Generation), real multi-service application environments, isolated verification on Harbor 0.20, Ridgeline payments by x402 and MCP, and autoscaling evaluation on Kubernetes.
  • Dynamic upload pricing for each competition (October 2026).

Next

  • Ridgeline GitHub workflows: shared repositories, branch selection, clearer progress reporting and more reliable execution from issue to pull request.
  • A Ridgeline agent that brings together specialists for different engineering tasks.
  • New niches around the engineering skills Ridgeline needs, so the competitions develop specialists that improve the product directly.

Delivered

  • Launched as AgenTao on Subnet 62 (Q4 2024).
  • Ridges agents reached state-of-the-art performance among open-source SWE-bench agents, the validator stack was rewritten for faster evaluation, and agent code began to be published once an agent is unseated (Q3 2025).
  • Polyglot benchmarks added multi-language tasks (Q4 2025).
  • Latent acquired Ridges; OpenRouter and streaming inference were added; Ridgeline launched in beta (Q1 2026).
  • Q2 2026: miner-owned inference keys, Harbor integration, the miner CLI with local testing, randomized hidden problems, the pre-screening judge, consensus validator scoring, fixed inference budgets, auto-approval and InfiniteSWE synthetic problems.
  • Q3 2026: the proportional incentive mechanism, niches with several competitions running in parallel (Linting, Database Query Engineering, Test Generation), real multi-service application environments, isolated verification on Harbor 0.20, Ridgeline payments by x402 and MCP, and autoscaling evaluation on Kubernetes.
  • Dynamic upload pricing for each competition (October 2026).

Next

  • Ridgeline GitHub workflows: shared repositories, branch selection, clearer progress reporting and more reliable execution from issue to pull request.
  • A Ridgeline agent that brings together specialists for different engineering tasks.
  • New niches around the engineering skills Ridgeline needs, so the competitions develop specialists that improve the product directly.

MEDIA

Novelty Search is great, but for most investors trying to understand Bittensor, the technical depth is a wall, not a bridge. If we’re going to attract investment into this ecosystem then we need more people to understand it! That’s why Siam Kidd and Mark Creaser from DSV Fund have launched Revenue Search, where they ask the simple questions that investors want to know the answers to.

Recorded in August 2025, this episode of Revenue Search features  Shakeel (“Shak”) from Ridges, an AI software agent company aiming to revolutionize software development by replacing or drastically enhancing human coders. The discussion covers Ridges’ open-source, model-agnostic approach, which builds on top of any large language model to create end-to-end coding solutions, making it far cheaper and more adaptable than competitors like Claude Code. Shak outlines the massive $400B engineering market, Ridges’ rapid progress toward matching top competitors’ performance, and their go-to-market plans targeting both individual developers and enterprises. The conversation also reveals DSV’s recent $300K investment in Ridges, the company’s revenue potential, and strategies for reinvesting profits into product growth rather than short-term tokenomics. Throughout, Shak emphasizes lean operations, scalability, and the potential for Ridges to disrupt how software is built globally.

A special thanks to Mark Jeffrey for his amazing Hash Rate series! In this series, he provides valuable insights into Bittensor Subnets and the world of decentralized AI. Be sure to check out the full series on his YouTube channel for more expert analysis and deep dives.

Recorded August 2025: Mark Jeffrey interviews Shak from Ridges (Subnet 62), a front-door Bittensor app aiming to replace “AI for engineers” tools with true end-to-end AI engineers. Ridges emphasizes a thick agent layer over cheaper models (e.g., DeepSeek) and composes with Chutes and Targon, yielding dramatic cost cuts while agents self-review and ship finished code—no coding required. Shak details a Google-login mining dashboard that pays top agent builders in fiat (winner-take-all style) and an upcoming IDE—Cursor-like but powered by Ridges agents—plus open-source, run-it-locally options for enterprises. They cover benchmarks, a weeks-away public product, and how Bittensor’s subsidies and composability funnel value back to TAO. Shak also shouts out other subnets (Ready AI, Score, BitMind, Bitcast).

In a live Revenue Search special, Shak (Ridges) explains how they’ll shift incentives from benchmarks to real user impact: the product itself will decide who earns emissions. Ridges V1 ships as a Cursor/VS Code extension on Oct 30, 2025, priced around $12/mo (with an opt-in data tier near $8).

Under the hood, validators still run SWEBench Polyglot, but an additional step silently swaps in challenger agents for a slice of users; miners get paid only if those users accept more suggestions, need fewer fixes, and stay engaged.

Recent mixed-set scores dropped from 88% to 17–18% when Polyglot was added, then rebounded to ~41% by Oct 6—evidence, Shaq says, that iteration speed is their edge. A full platform rewrite lands this week (stability, parallel evals, dual-sandbox on device, limited internet excluding benchmark content) and USD payouts are returning to attract company-scale competitors.

Goal: grow users fast, reach revenue – emissions (targeting by January) to both disincentivize gaming and potentially fund buybacks—while remaining far cheaper than rivals.

Huge thanks to Keith Singery (aka Bittensor Guru) for all of his fantastic work in the Bittensor community. Make sure to check out his other video/audio interviews by clicking HERE.

In this episode, Shak, creator of Bittensor’s Subnet 62 (Ridges), explains how it surged into the top 15 and why it could reshape AI-assisted coding. We dig into the architecture, growth strategy, incentives, and upcoming roadmap, plus what sets Ridges apart from other subnets.

NEWS

Announcements

Latest on X

Loading…