Apex (Bittensor’s Subnet 1) Shows a New Path for Decentralized AI Evaluation

One of the hardest problems in decentralized AI is not generation, it is evaluation. As AI systems move into open-ended domains like reasoning, creativity, and agentic behavior, judging quality becomes subjective, expensive, and difficult to verify on-chain. Traditional approaches rely on handcrafted metrics, spot checks, or delayed outcomes. All of them struggle at scale. New…

Read More

Bittensor Brief #17: Sundae Bar – Subnet 121

SUMMARY: The video introduces Sundae Bar (Subnet 121), a Bittensor project building a single generalized, autonomous enterprise AI agent that functions as a true digital worker rather than a chatbot—capable of executing end-to-end business workflows with persistent memory, role awareness, and continuous improvement. After pivoting from multiple vertical agents, Sundae Bar now uses a winner-take-all…

Read More

Bittensor Ecosystem Highlights – Jan Week 2

SUBNET UPDATES & ACHIEVEMENTS ➤ @chutes_ai (SN64)Chutes unveiled Fictio, a new character roleplay platform expanding interactive AI use cases within the subnet.Read more: Fictio launch ➤ @tplr_ai (SN3)Templar released a new research paper titled Heterogeneous Low-Bandwidth Pre-Training of LLMs, detailing novel approaches to efficient large model training.Read the paper: Templar research ➤ @babelbit (SN59)Babelbit introduced…

Read More

Loosh AI Debuts Cognition Engine Beta on Bittensor’s Subnet 78

Loosh AI, building on Bittensor’s Subnet 78, has released the public beta of its Cognition Engine, offering an early look at a decentralized system designed to make machine reasoning, memory, and behavioral constraints directly observable. Unlike consumer facing AI interfaces, the Loosh Cognition Engine is built as infrastructure. It exposes the internal processes that govern…

Read More