Bittensor Weekly Recap – 12 January 2026: News, TAO.
Big Events in the Bittensor Ecosystem
SimplyTAO Beta Is Progressing Successfully
SimplyTAO is currently in beta, and the early phase is progressing exceptionally well. A growing number of users have already placed their trust in the platform, actively engaging with its features and confirming that we are building in the right direction.
Development continues at a strong pace, guided by real user feedback and real usage. As a result, SimplyTAO is steadily evolving into a reliable and insightful hub for the Bittensor and TAO ecosystem.
Limited beta access remains available. If you would like to join early, explore the platform ahead of public release, and be part of its growth, feel free to contact us via private message.
Join to SimplyTao Early Access now!
Subnet Updates
Fictio by Chutes launches in early access as a new AI character roleplay platform
Chutes AI has launched Fictio, a new AI-powered character roleplay platform designed for users who want to chat, create, and build narrative worlds without heavy restrictions. The platform is currently available in early access and targets creators looking for greater control over their stories and characters.
Fictio focuses on long-form character roleplay, allowing users to interact with AI-driven characters while maintaining narrative consistency and creative freedom. The platform emphasizes user ownership of characters and worlds, positioning itself as a creator-first alternative to more constrained AI chat systems.
The project is developed by Chutes AI, a company known for its decentralized, serverless AI compute infrastructure used to deploy and scale large language models. By building Fictio directly on its own infrastructure, Chutes highlights scalability, flexibility, and performance as core strengths of the platform.
At launch, Fictio does not require a paid subscription and offers free AI generations during early access, lowering the barrier for experimentation. Chutes is also running creator and referral contests with rewards of up to 500 US dollars, signaling a strong focus on community-driven growth.
An official Discord community supports early users with feedback channels and development updates. While detailed technical documentation and a public roadmap have not yet been released, Chutes has indicated that Fictio will continue evolving into a broader world-building and creative roleplay ecosystem.
Entering a competitive market, Fictio positions itself as a flexible and creator-focused AI roleplay platform, and represents Chutes’ first major consumer-facing product built on its AI infrastructure.
Templar outlines low-bandwidth LLM training and Covenant72B progress
Templar has shared new details on its approach to low-bandwidth pre-training of large language models, alongside an update on the final training phase of Covenant72B. The announcement highlights ongoing efforts to make large-scale model training viable outside traditional, high-bandwidth data centers.
The core focus of the work is heterogeneous distributed training, a method designed for environments where compute resources and network speeds vary significantly. Instead of requiring all machines to communicate constantly at high throughput, Templar’s approach allows high-bandwidth nodes to operate normally, while lower-bandwidth participants contribute through compressed and staged communication. This reduces network load while preserving training quality.
By combining communication-efficient data parallelism with pipeline model parallelism, the system enables multiple weaker nodes to jointly handle parts of a single model replica. Data exchanged between stages is heavily compressed, making training practical even under limited network conditions.
This research aligns with Templar broader goal of incentivized, internet-scale model training, where contributors can participate from diverse hardware setups rather than centralized infrastructure. The update also confirms continued progress on Covenant72B, a seventy-two-billion-parameter language model approaching the final stages of training.
Together, these developments signal a push toward decentralized and more accessible LLM training, positioning Templar as a project focused on scaling large models beyond traditional data center constraints.
Loosh AI launches Cognition Engine Beta on Bittensor Subnet 78
Loosh AI has launched the public beta of its Cognition Engine, now live on Bittensor Subnet 78, offering an early look at a new architecture designed for robotics and agentic systems rather than traditional chat-based AI. The release is positioned as a transparent cognition layer where reasoning, memory, and constraints are fully observable.
The Cognition Engine operates through multi-stage reasoning with tool usage, allowing users to observe tasks breaking down in real time and follow execution across multiple steps.
The beta showcases Loosh’s core cognitive services, including working memory, long-term memory, embeddings, and an explicit ethics layer. This ethics system is designed to actively constrain behavior using rights-based, deontological, and virtue-based evaluators, enabling users to test how conflicting constraints influence outcomes in practice.
Loosh AI describes the beta as a hands-on environment for early adopters, builders, and participants in the Bittensor ecosystem who want to explore how cognition operates as an integrated architecture. In parallel, the team confirmed progress on its validator and miner infrastructure, marking the first stage of running real inference workloads through the Loosh network.
Overall, Loosh positions its Cognition Engine as a foundational layer for autonomous agents and robotics, emphasizing transparency, inspectability, and ethical constraints as core design principles rather than afterthoughts.
404 announces new mesh generation competition on Bittensor Subnet 17
404 has announced the launch of its second major competition on Bittensor Subnet 17, shifting focus toward native 3D mesh generation for real production use. The initiative marks a strategic move away from an exclusive emphasis on Gaussian splats and toward assets that integrate directly into modern game and 3D pipelines.
Until now, Subnet 17 has concentrated on Gaussian splats, a technique well suited for visualization and world reconstruction but limited in practical deployment across existing engines. That work enabled Unity integrations, the creation of one of the largest 3D datasets, and early applications with real users and organic traffic. However, according to the team, developer demand is now clearly centered on production-ready mesh outputs.
The new competition is designed to accelerate progress toward native mesh generation that can be dropped directly into game engines and creative tools. By prioritizing meshes, 404 aims to unlock immediate commercial utility while maintaining longer-term research into world models, digital twins, and Gaussian splat-based representations.
The competition-driven format allows Subnet 17 to adapt quickly to shifting goals, rewarding miners who deliver practical results while keeping innovation pressure high. With several open source mesh generation models released in recent weeks, the timing reflects a rapidly rising baseline for quality and performance.
Set to launch in the coming weeks, the mesh generation competition signals 404’s intent to position Subnet 17 as a core source of usable 3D assets, bridging the gap between experimental 3D AI research and real-world production workflows.
New Subnet Launches: LuminarNetwork SN87
Luminar Network is an emerging technology project focused on transforming raw video footage into verified and timestamped digital evidence. Built with transparency and decentralization at its core, the project addresses a critical issue in modern video infrastructure: the vast majority of recorded footage is never reviewed or meaningfully analyzed.
Industry data highlights the scale of the problem. Up to 95% of body camera footage and 98% of CCTV recordings remain unseen. This gap limits accountability, reduces situational awareness, and weakens the value of video as a source of reliable evidence.
Luminar Network introduces a modular system of vision intelligence agents designed to operate across diverse hardware environments, from body cameras to city-wide surveillance systems. These agents analyze, verify, and structure video data so it can be trusted and acted upon in real time.
A major milestone for the project is the recent mainnet launch on Bittensor (SN87). By building on Bittensor, Luminar leverages a decentralized intelligence network that allows video processing and validation to occur without centralized control points. This approach improves resilience, transparency, and scalability while aligning incentives across the network.
By unifying fragmented video ecosystems through Bittensor-based decentralized intelligence, Luminar Network positions itself at the intersection of blockchain, computer vision, and real-world data. The project’s long-term vision is clear: to make video evidence more trustworthy, accessible, and operationally useful for institutions, cities, and organizations worldwide.
As demand for reliable visual data continues to grow, Luminar Network demonstrates how Bittensor can serve as a foundation for next-generation video intelligence infrastructure.
TAO Market Update
Price: $280-$295
Weekly result (7d): +7% to +9%
Ranking: #49-#51
Market Cap: $2.75B-$2.9B
24h Volume: $140M-$180M
Over the last 24 hours, TAO traded in approximately the $280-$295 range, reflecting sustained intraday volatility and balanced two-way participation. Current market data places TAO around the top-50 by market capitalization, with a total valuation in the $2.75-$2.9 billion range and 24-hour trading volume between $140 million and $180 million, confirming continued liquidity and strong market engagement.
Top Gainer Subnet: sundae_bar (SN121)
sundae_bar is an AI agent marketplace that helps businesses discover, test, and use AI agents built by real developers. The platform connects practical AI use cases with decentralized intelligence, including its presence on Bittensor subnet 121, making Bittensor-based innovation easier to access. By focusing on discovery and real-world adoption, sundae_bar acts as a bridge between decentralized AI networks like Bittensor and the growing demand for AI agents as digital workers.
Weekly change: +25.67%
Price: $3.261754
Market Cap: $6.32M
Volume (24h): $388.02K
