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of new subnets being added it can be hard to get up to date information across all subnets, so data may be slightly out of date from time to time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 27

Compute

Emissions
Value
Recycled
Value
Recycled (24h)
Value
Registration Cost
Value
Active Validators
Value
Active Miners
Value
Active Dual Miners/Validators
Value

ABOUT

What exactly does it do?

Compute Subnet 27 aims to boost decentralized computing capabilities, enhancing efficiency and scalability for AI computations. Leveraging the Bittensor framework, they facilitate collaborative computing power across a distributed network, supporting the execution of advanced AI models and tasks. This initiative focuses on creating a more accessible and robust computing platform for AI development in a decentralized environment, enabling exponential growth for Bittensor in computational power.

Subnet 27 marked a significant advancement by enabling platform-native distributed compute for validators on the network. This development liberated Bittensor from relying on external sources like GPU or CPU rentals from companies such as Runpod, Vast Vulture, or Digital Ocean. Subnet 27 empowered users to access compute resources directly on the Bittensor Network, revolutionizing the landscape. The project provides a trustless, validator-verified mechanism for verifying GPU performance, ensuring honest behavior and consistent availability in a decentralized compute network.

Compute Subnet 27 aims to boost decentralized computing capabilities, enhancing efficiency and scalability for AI computations. Leveraging the Bittensor framework, they facilitate collaborative computing power across a distributed network, supporting the execution of advanced AI models and tasks. This initiative focuses on creating a more accessible and robust computing platform for AI development in a decentralized environment, enabling exponential growth for Bittensor in computational power.

Subnet 27 marked a significant advancement by enabling platform-native distributed compute for validators on the network. This development liberated Bittensor from relying on external sources like GPU or CPU rentals from companies such as Runpod, Vast Vulture, or Digital Ocean. Subnet 27 empowered users to access compute resources directly on the Bittensor Network, revolutionizing the landscape. The project provides a trustless, validator-verified mechanism for verifying GPU performance, ensuring honest behavior and consistent availability in a decentralized compute network.

PURPOSE

What exactly is the 'product/build'?

Neural Internet emphasized Bittensor’s permissionless nature, enabling engagement in mining, subnet creation, and contributions without external approvals. There are vast opportunities beyond machine learning with Bittensor developers, advocating for the platform’s versatility and potential impact.

Transparency and clear communication attract committed miners and validators, ensuring subnet longevity and stability within the ecosystem. Efforts have focused on streamlining access for miners, whether owning or renting GPUs, to participate in mining operations. Plugging in hardware is as simple as running a few commands and ensuring Docker is set up, making it accessible for everyone.

The team is simplifying the process further with a one-click deploy feature, reducing entry barriers for mining and validator engagement. Efforts to lower barriers include easier access to GPUs for diverse use cases such as flight simulators, enhancing validator participation. Validators specify GPU requirements and send requests, gaining temporary access to machines for tasks through a streamlined process.

Subnet 27 prioritizes distributed ownership to prevent dominance and maintain network integrity, regulating IP usage for fairness. It boasts a well-distributed ownership structure among miners, with IPs and cold keys maintaining a balanced ownership of uid slots, ensuring fair participation in the mining pool. The community aims to prevent the use of the same IP address by multiple users to avoid issues with consensus and incentives. Miners rent GPUs but inform providers to change IPs to prevent conflicts before entering the network. Emphasis is on aligning incentives to encourage high-performance GPUs like A6000s over older models like 1080s. Efforts involve continuous evaluation and tweaking of mechanisms to prevent domination of the subnet.

Validators choose from available GPUs, including high-end models like the H100, ensuring diverse ownership among miners. Aligning incentives encourages use of high-performing GPUs, ensuring efficient validation in a permissionless network.  Stake-based prioritization minimizes centralization risks, optimizing subnet operations. Validators with more stake have higher priority in compute allocation, reducing centralization risks.

Subnet owners select validators, controlling participant access and ensuring compliance for subnet stability. Validators monetize by creating value across subnets, influencing network economics and user engagement. Subnet 27 ensures fair emission distribution, promoting competitive mining with fair access to compute resources. Validators on the subnet network must comply with the rules set by the subnet owner and miners. Weight setting by validators may become insignificant as network value access could be restricted. Validators can monetize their subnets by allowing access through front ends they build, competing with successful business models already in the market.

Subnet 27 is inherently scalable, with the ability to attract more miners as it expands, ensuring a balanced distribution of low-end and high-end compute resources.

 

Proof-of-GPU Compute System

  • Validators send Merkle-tree based challenge seeds to miners.
  • Miners must execute computationally intensive tasks (like building nested arrays of hashes) using their GPU.
  • The returned results are benchmarked against known GPU throughput (TFlops) profiles.
  • Based on output, validators can accurately determine the type and power of the GPU, e.g., H100, A100, 4090, etc.

 

Compute Access

  • Users can rent containers (Docker-based initially, moving to Kubernetes in 2025).
  • Similar to traditional services like RunPod or Vast.ai but secured and validated on-chain.
  • Compute buyers get SSH access to specific machines.

 

Fraud Prevention & Hardware Consistency

  • Regular challenge-response proof tests discourage miners from hardware swapping or misuse.
  • Plans to implement simultaneous proofs across machines (“raids”) to expose shared-resource cheating.

 

Frontend & Branding

  • Rebranding under NodeXO (from neuralinternet), with new light/dark themed UI targeting enterprise clients.
  • Focus on building corporate trust and usability beyond typical blockchain aesthetics.

 

Infrastructure Plans

  • Full Kubernetes control plane on each minor node.
  • Validators check for availability & stability of the clusters via exposed APIs.
  • Nested orchestration for advanced failover & uptime monitoring.

 

Neural Internet emphasized Bittensor’s permissionless nature, enabling engagement in mining, subnet creation, and contributions without external approvals. There are vast opportunities beyond machine learning with Bittensor developers, advocating for the platform’s versatility and potential impact.

Transparency and clear communication attract committed miners and validators, ensuring subnet longevity and stability within the ecosystem. Efforts have focused on streamlining access for miners, whether owning or renting GPUs, to participate in mining operations. Plugging in hardware is as simple as running a few commands and ensuring Docker is set up, making it accessible for everyone.

The team is simplifying the process further with a one-click deploy feature, reducing entry barriers for mining and validator engagement. Efforts to lower barriers include easier access to GPUs for diverse use cases such as flight simulators, enhancing validator participation. Validators specify GPU requirements and send requests, gaining temporary access to machines for tasks through a streamlined process.

Subnet 27 prioritizes distributed ownership to prevent dominance and maintain network integrity, regulating IP usage for fairness. It boasts a well-distributed ownership structure among miners, with IPs and cold keys maintaining a balanced ownership of uid slots, ensuring fair participation in the mining pool. The community aims to prevent the use of the same IP address by multiple users to avoid issues with consensus and incentives. Miners rent GPUs but inform providers to change IPs to prevent conflicts before entering the network. Emphasis is on aligning incentives to encourage high-performance GPUs like A6000s over older models like 1080s. Efforts involve continuous evaluation and tweaking of mechanisms to prevent domination of the subnet.

Validators choose from available GPUs, including high-end models like the H100, ensuring diverse ownership among miners. Aligning incentives encourages use of high-performing GPUs, ensuring efficient validation in a permissionless network.  Stake-based prioritization minimizes centralization risks, optimizing subnet operations. Validators with more stake have higher priority in compute allocation, reducing centralization risks.

Subnet owners select validators, controlling participant access and ensuring compliance for subnet stability. Validators monetize by creating value across subnets, influencing network economics and user engagement. Subnet 27 ensures fair emission distribution, promoting competitive mining with fair access to compute resources. Validators on the subnet network must comply with the rules set by the subnet owner and miners. Weight setting by validators may become insignificant as network value access could be restricted. Validators can monetize their subnets by allowing access through front ends they build, competing with successful business models already in the market.

Subnet 27 is inherently scalable, with the ability to attract more miners as it expands, ensuring a balanced distribution of low-end and high-end compute resources.

 

Proof-of-GPU Compute System

  • Validators send Merkle-tree based challenge seeds to miners.
  • Miners must execute computationally intensive tasks (like building nested arrays of hashes) using their GPU.
  • The returned results are benchmarked against known GPU throughput (TFlops) profiles.
  • Based on output, validators can accurately determine the type and power of the GPU, e.g., H100, A100, 4090, etc.

 

Compute Access

  • Users can rent containers (Docker-based initially, moving to Kubernetes in 2025).
  • Similar to traditional services like RunPod or Vast.ai but secured and validated on-chain.
  • Compute buyers get SSH access to specific machines.

 

Fraud Prevention & Hardware Consistency

  • Regular challenge-response proof tests discourage miners from hardware swapping or misuse.
  • Plans to implement simultaneous proofs across machines (“raids”) to expose shared-resource cheating.

 

Frontend & Branding

  • Rebranding under NodeXO (from neuralinternet), with new light/dark themed UI targeting enterprise clients.
  • Focus on building corporate trust and usability beyond typical blockchain aesthetics.

 

Infrastructure Plans

  • Full Kubernetes control plane on each minor node.
  • Validators check for availability & stability of the clusters via exposed APIs.
  • Nested orchestration for advanced failover & uptime monitoring.

 

WHO

Team Info

The Neural Internet team operates as a decentralized autonomous organization (DAO), with clear departmental structures and teams handling specific aspects to promote contribution and value creation efficiently. This structured approach from DAO to team level has facilitated rapid scaling.

Neural internet co-founder Hansel’s interest in the intersection of blockchain and AI was ignited in late 2020 through projects like Ocean Protocol and Vector Space, which introduced them to language model technology. Their exploration of OpenAI, particularly with GPT-3, provided hands-on experience in adjusting parameters to understand the model’s functionalities and outputs.

With a background in researching blockchain technology, Hansel shifted his focus to working in the blockchain industry following an internship at State Street. Recognizing the potential of applying blockchain concepts to traditional banking operations, he transitioned to a blockchain software development company to explore product creation within the market.

The integration of OpenAI and GPT-3 streamlined the process of creating white papers from several weeks to just a couple of days, significantly enhancing client interactions. By leveraging AI to generate documents based on prompts, Hansel optimized team operations and insights, catalyzing a shift in the entire company’s workflow toward greater efficiency.

Team

Hansel Melo – Co-Founder

Gunner McLeod – Co-Founder

Arthur Simonian – Co-Founder

Adrian Walker – Co-Founder

Angel Rivas – Co-founder

Matan K – Co-Founder

Felix Peterson – Design and Product Architecture

Alex Kiriakides – Product and Financial Analyst

Douglas Albert – Product Development

Ibtehaj Khan – ML Engineer

Saram Hai – ML Engineer

The Neural Internet team operates as a decentralized autonomous organization (DAO), with clear departmental structures and teams handling specific aspects to promote contribution and value creation efficiently. This structured approach from DAO to team level has facilitated rapid scaling.

Neural internet co-founder Hansel’s interest in the intersection of blockchain and AI was ignited in late 2020 through projects like Ocean Protocol and Vector Space, which introduced them to language model technology. Their exploration of OpenAI, particularly with GPT-3, provided hands-on experience in adjusting parameters to understand the model’s functionalities and outputs.

With a background in researching blockchain technology, Hansel shifted his focus to working in the blockchain industry following an internship at State Street. Recognizing the potential of applying blockchain concepts to traditional banking operations, he transitioned to a blockchain software development company to explore product creation within the market.

The integration of OpenAI and GPT-3 streamlined the process of creating white papers from several weeks to just a couple of days, significantly enhancing client interactions. By leveraging AI to generate documents based on prompts, Hansel optimized team operations and insights, catalyzing a shift in the entire company’s workflow toward greater efficiency.

Team

Hansel Melo – Co-Founder

Gunner McLeod – Co-Founder

Arthur Simonian – Co-Founder

Adrian Walker – Co-Founder

Angel Rivas – Co-founder

Matan K – Co-Founder

Felix Peterson – Design and Product Architecture

Alex Kiriakides – Product and Financial Analyst

Douglas Albert – Product Development

Ibtehaj Khan – ML Engineer

Saram Hai – ML Engineer

FUTURE

Roadmap

Subnet 27 is engineered for scalable growth, designed to accommodate an expanding number of miner slots. Its adaptable configuration allows for the inclusion of more miners, effectively balancing both low-end and high-end compute resources.

Increasing slots enhances distribution and accessibility to diverse GPUs within the subnet, fostering positive ecosystem development. The objective is to facilitate seamless GPU connectivity across various models, ensuring incentives are appropriately aligned regardless of the number of GPUs used. Scalability is supported by an expanding UID count, enabling flexible growth based on validator demand.

Priority is placed on developing tooling that simplifies validator deployment and operation, promoting value creation. Beyond Subnet 27, the team explores opportunities for additional subnets, aiming to facilitate application development such as market prediction tools by standardizing API access across subnets.

Future plans include expanding collaborations and showcasing practical applications, such as deploying a Tiny Box to illustrate tangible use cases. Emphasis is placed on demonstrating the existing utility of subnets and fostering wider community adoption to maximize project impact.

 

Completed / In Progress (2024)

  • Merkle-proof based Proof-of-GPU infrastructure deployed.
  • Early implementation of SSH compute access.
  • Rebranding initiative started (NodeXO), with redesigned UI/UX.
  • Engagement with industry researchers and open-source contributors for compute verification.
  • Running and scaling Subnet 27 on Bittensor.

 

Upcoming Milestones (2025 & Beyond)

Q3–Q4 2025:

  • Launch of full Kubernetes cluster support with minor-level orchestration.
  • Release of NodeXO 2.0 website, with corporate-facing design and onboarding UX.
  • Further development of the GPU-proof benchmarking library (possibly open-sourced in Python).
  • Strengthening reliability of compute access for real-world users (addressing misuse and uptime issues).

 

Compute Ecosystem Expansion:

  • Targeting multi-machine distributed training support (2026).
  • Planning on supporting enterprise-grade training workflows using platforms like Run.AI.

 

Economic & Governance Direction:

  • Explore Revenue Burn vs. Recycling Mechanisms.
  • Burn in bear markets to reduce token supply.
  • Recycle in bull markets to incentivize growth and rainy-day reserves.
  • Experimentation with community-voted tokenomics, especially around incentives per GPU model.

 

Security and Performance Enhancements:

  • Continue refining anti-fraud systems (e.g., against hardware swaps during active rentals).
  • Integrate automatic validator-level monitoring and recovery for validator-cluster reliability.

 

Long-Term Vision:

  • Position as a cornerstone decentralized compute provider, supporting scalable and validated GPU access worldwide.
  • Become a fundamental infrastructure provider across decentralized AI and compute industries.

 

Subnet 27 is engineered for scalable growth, designed to accommodate an expanding number of miner slots. Its adaptable configuration allows for the inclusion of more miners, effectively balancing both low-end and high-end compute resources.

Increasing slots enhances distribution and accessibility to diverse GPUs within the subnet, fostering positive ecosystem development. The objective is to facilitate seamless GPU connectivity across various models, ensuring incentives are appropriately aligned regardless of the number of GPUs used. Scalability is supported by an expanding UID count, enabling flexible growth based on validator demand.

Priority is placed on developing tooling that simplifies validator deployment and operation, promoting value creation. Beyond Subnet 27, the team explores opportunities for additional subnets, aiming to facilitate application development such as market prediction tools by standardizing API access across subnets.

Future plans include expanding collaborations and showcasing practical applications, such as deploying a Tiny Box to illustrate tangible use cases. Emphasis is placed on demonstrating the existing utility of subnets and fostering wider community adoption to maximize project impact.

 

Completed / In Progress (2024)

  • Merkle-proof based Proof-of-GPU infrastructure deployed.
  • Early implementation of SSH compute access.
  • Rebranding initiative started (NodeXO), with redesigned UI/UX.
  • Engagement with industry researchers and open-source contributors for compute verification.
  • Running and scaling Subnet 27 on Bittensor.

 

Upcoming Milestones (2025 & Beyond)

Q3–Q4 2025:

  • Launch of full Kubernetes cluster support with minor-level orchestration.
  • Release of NodeXO 2.0 website, with corporate-facing design and onboarding UX.
  • Further development of the GPU-proof benchmarking library (possibly open-sourced in Python).
  • Strengthening reliability of compute access for real-world users (addressing misuse and uptime issues).

 

Compute Ecosystem Expansion:

  • Targeting multi-machine distributed training support (2026).
  • Planning on supporting enterprise-grade training workflows using platforms like Run.AI.

 

Economic & Governance Direction:

  • Explore Revenue Burn vs. Recycling Mechanisms.
  • Burn in bear markets to reduce token supply.
  • Recycle in bull markets to incentivize growth and rainy-day reserves.
  • Experimentation with community-voted tokenomics, especially around incentives per GPU model.

 

Security and Performance Enhancements:

  • Continue refining anti-fraud systems (e.g., against hardware swaps during active rentals).
  • Integrate automatic validator-level monitoring and recovery for validator-cluster reliability.

 

Long-Term Vision:

  • Position as a cornerstone decentralized compute provider, supporting scalable and validated GPU access worldwide.
  • Become a fundamental infrastructure provider across decentralized AI and compute industries.

 

MEDIA

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 audio interview, Keith interviews AIZorr0 from Neural Internet and delves into Subnet 27 Compute, where miners can leverage their GPUs in the cloud directly on Bittensor to earn TAO.

A big thank you to Tao Stats for producing these insightful videos in the Novelty Search series. We appreciate the opportunity to dive deep into the groundbreaking work being done by Subnets within Bittensor! Check out some of their other videos HERE.

In this session, the team delves into the intricacies of the subnet’s validation mechanism, including the innovative Proof of GPU system, which ensures reliable performance through Merkle tree-based validation. The conversation covers the challenges of providing scalable GPU compute resources for machine learning, addressing issues such as hardware tampering, validation reliability, and incentive alignment within the network. Additionally, the team outlines their roadmap, focusing on improving platform reliability, rebranding efforts, and expanding into distributed model training by 2026. The discussion emphasizes the importance of community-driven governance and strategic decision-making in ensuring the subnet’s success.

NEWS

Announcements

Load More