Bittensor Weekly Recap – 26 January 2026: News, TAO.
Big Events in the Bittensor Ecosystem
Opentensor Foundation Announces Judging Panel for Bittensor Subnet Ideathon
The Opentensor Foundation has revealed the judging panel for its Bittensor Subnet Ideathon, a competition organized with HackQuest. The event targets developers interested in building subnets on the Bittensor network. The competition offers over $18,000 USD in cash prizes. Selected projects may also receive accelerator access and potential investment from ecosystem funds.
The panel includes five figures from the Bittensor ecosystem. Kibibyte serves as CTO of Latent Holdings and previously contributed to core development at Opentensor Foundation. Latent Holdings maintains the Bittensor documentation and develops ecosystem tools including TAO.app. Sami Kassab is Managing Partner of Unsupervised Capital, a liquid token fund focused on the Bittensor ecosystem. He transitioned from aerospace engineering to cryptocurrency in 2021. Chris Zacharia, known as Macro Zack, founded Bitstarter, a crowdfunding platform for Bittensor startups. He also serves as CMO at Macrocosmos AI, which operates multiple subnets on the network. Pepeleplutus co-founded Leadpoet AI, operating Subnet 71 as a decentralized sales intelligence platform. Mason LeGrange holds the CTO position at BT Labs.
Competition Structure and Prizes
The competition runs in two rounds. Round One focuses on subnet design and closes February 25, 2026. Teams submit design proposals covering incentive mechanisms, miner and validator roles, and business strategy. No testnet deployment is required at this stage. Seven teams advance to Round Two, running March 2 through March 30, 2026. Selected participants must implement their proposals on the Bittensor testnet.
The Hackathon Winner receives $10,000 USD plus entry into the Bitstarter Accelerator and a pitch session with Bittensor co-founder Jacob Robert Steeves or Opentensor Foundation President Etienne. The Runner-Up receives $3,000 USD. Five additional awards of $1,000 USD each recognize notable subnet designs. Unsupervised Capital has indicated it may consider a discretionary investment of up to 1,000 TAO for standout projects. At current prices around $270 USD per token, this represents approximately $270,000 in potential funding.
Registration remains open through February 25, 2026 via HackQuest. The Bittensor network currently hosts over 100 active subnets producing various digital commodities including AI inference, data storage, and compute resources.
CrunchDAO Opens Bittensor Mining to Academic and Enterprise ML Scientists
CrunchDAO announced plans to facilitate Bittensor subnet mining for its community of machine learning researchers. The platform connects over 11,000 ML engineers and 1,200 PhDs with decentralized AI infrastructure, allowing them to participate without managing blockchain complexity directly.
The initiative aims to address what CrunchDAO describes as a talent bottleneck in decentralized AI. Traditional academic and institutional ML scientists possess model development expertise but often lack familiarity with Web3 infrastructure requirements like managing mining slots, staking mechanics, and blockchain operations. CrunchDAO plans to operate as coordinators on Bittensor subnets, handling technical infrastructure while contributors focus on model development. The approach abstracts blockchain complexity so participants can submit models without becoming blockchain experts. According to founder Jean Herelle, the platform enables data scientists from enterprise and academic backgrounds to contribute to decentralized AI while focusing on what they do best.
The company frames Bittensor subnets as another category of customers for its collective intelligence network. CrunchDAO already serves clients including ADIA Lab and the Broad Institute of MIT and Harvard through structured modeling competitions. Bittensor’s subnet architecture provides a compatible environment for deploying ensemble models built from aggregated contributor submissions.
CrunchDAO’s method involves aggregating models from hundreds of independent engineers into ensemble predictions. The company claims combined outputs from diverse contributors often outperform individual expert forecasts. This collective intelligence methodology would translate to Bittensor mining through coordinators that aggregate community submissions into ensemble models for the network. The structure allows any subnet creator or coordinator to establish a modeling market that purchases mining slots and functions as a meta-layer. CrunchDAO refers to this as a “super-miner” approach.
CrunchDAO raised $10 million in total funding, including a $5 million round in October 2025 co-led by Galaxy Ventures and Road Capital with participation from VanEck and Multicoin Capital. The company was founded in 2021 in Paris and was selected for the Solana Incubator’s second cohort earlier in 2025.
Subnet Updates
Score (SN44) Announces Manako Platform to Simplify Bittensor Vision AI Development
Score, the team behind Bittensor SN44, announced plans to launch Manako in Q1 2026. The platform aims to allow users to create custom computer vision models through conversational interfaces rather than requiring machine learning expertise.
Manako builds on Score’s existing Bittensor infrastructure. The subnet currently operates as a decentralized computer vision framework focused on sports analytics, particularly football match annotation and analysis. Score claims their technology reduces costs by 10x compared to traditional solutions while improving processing speed. The team has secured partnerships with organizations across multiple industries, including Reading FC in the UK, a European petroleum company operating thousands of gas stations, and Two-a-Day, an African fruit production company.
The new platform attempts to lower barriers to computer vision development. According to Score, users would describe what they want to detect or analyze, provide images or video, and Manako handles optimization, deployment, and execution. The system abstracts model architecture, pipeline configuration, and infrastructure management behind SDK and API interfaces.
Score describes three core components of the platform. First, a curated library of vision building blocks that the team maintains and optimizes. Second, an AI orchestrator that interprets user intent and assembles appropriate vision components into execution graphs. Third, simplified developer interfaces where users interact only with outcomes and endpoints while internal complexity remains abstracted.
The Bittensor Subnet 44 would continue feeding new computer vision models into Manako through ongoing competitions among miners. This creates a system where participants compete on model quality while Score focuses on data collection and productization.
Score positions Manako as a way to reach smaller companies, startups, and developers who previously could not access advanced computer vision capabilities due to cost or technical requirements. The enterprise partnerships would continue separately on dedicated plans.
Registration for early access is available at manako.ai. The initial launch will have limited user capacity.
Templar (SN3) Completes 72-Billion Parameter Model Training on Bittensor Network
Templar, operating SN3 on Bittensor, has completed pre-training of Covenant72B. The 72-billion-parameter model was trained entirely on decentralized infrastructure without a data center, central server, or single point of control.
The training run unfolded across an open network where contributors provided heterogeneous hardware and operated under real-world constraints. Latency, synchronization costs, and partial failures were normal operating conditions throughout the process. The run demonstrated that Bittensor infrastructure could manage these challenges long enough to complete frontier-scale pretraining.
With pretraining complete, Covenant72B now moves toward post-training phases including fine-tuning, alignment, and evaluation. The next stage, called Crusades, shifts focus to optimization where participants submit training code and validators benchmark implementations.
Read full article about Covenant72B
Metanova Labs (SN68) Summarizes First Year of Decentralized Drug Discovery on Bittensor
Metanova Labs published a year-in-review for NOVA, the drug screening platform operating on Bittensor Subnet 68. The system launched in March 2025 as what the team describes as the first decentralized drug discovery platform on the network.
NOVA currently runs two parallel incentive mechanisms. NOVA Compound focuses on molecule discovery where miners compete to identify compounds with high affinity for weekly targets and low affinity for defined anti-targets. The system scores submissions using Boltz-2, developed by MIT and Recursion Pharma. NOVA Blueprint operates as a separate competition for search algorithms, where miners submit code designed to explore chemical spaces more effectively.
Over the past year, the platform expanded its search space to over 65 billion molecules and screened compounds across tens of thousands of mammalian proteins. The team added anti-targets to enforce selectivity, diversity bonuses to encourage exploration, and combinatorial chemistry with five reaction sets. They also integrated and optimized Boltz-2 scoring beyond Nvidia’s paid inference service.
Metanova signed partnerships with Diagen AI in Canada to develop an automated hit-picking system and Yalotein Biotech in Shanghai to expand into nanobody therapeutics. The Diagen collaboration aims to create curated compound libraries for licensing and benchmarking. The Yalotein partnership marks NOVA’s first step beyond small molecules into biological therapeutics.
For 2026, Metanova outlines plans to move from prediction to validation. The roadmap includes systematic wet-lab testing of NOVA-identified compounds, deployment of the automated hit-picking tool with token-gated access to premium libraries, and expansion into nanobodies with an initial batch of approximately 50 candidates entering laboratory testing through the Yalotein partnership.
The team states that experimental results will allow them to quantify hit rates relative to existing discovery approaches, providing metrics for evaluating the Bittensor-based platform’s performance against traditional methods.
New Subnet Launches:
Djinn (SN103) Launches on Bittensor as Prediction Intelligence Marketplace
Djinn has launched on Bittensor as Subnet 103, creating a marketplace where sports prediction experts can sell signals with TAO staked as performance guarantee. The project raised 600 TAO in 51 minutes through Bitstarter and filled all 256 UIDs.
The mechanism splits participants into Geniuses who provide predictions and stake collateral, and Idiots who purchase signals and execute bets. If predictions underperform committed thresholds, buyers receive refunds from the seller’s stake.
TAO Market Update
Price: $230–$240
Weekly result (7d): -6% to -4%
Ranking: #35–#54
Market Cap: $2.1B–$2.5B
24h Volume: $90M–$140M
As of today, TAO is trading approximately in the $230–$240 range, reflecting current market volatility and active trading participation. Latest market data shows TAO ranking between roughly #35 and #54 by market capitalization, with an estimated valuation of about $2.1B–$2.5B. The 24-hour trading volume remains elevated, roughly in the $90M to $140M range, indicating sustained liquidity and ongoing engagement from market participants.
Top Gainer Subnet: Kinitro SN26
Kinitro (SN26) is a decentralized AI project built as a Bittensor subnet focused on advancing robotic and embodied intelligence through competition-based incentives. It functions as an “AI Olympics” platform where developers submit autonomous agents that are evaluated by validators, with rewards distributed on-chain based on performance.
The project is developed openly by the ThreeTau collective and implemented as an open-source infrastructure integrating Bittensor, agent evaluation environments, and validator logic. Its goal is to create a market-driven system that accelerates practical AI and robotics research rather than static model development.
Weekly change: +42.48%
Price: $1.735391
Market Cap: $6.31M
Volume (24h): $716.87K
