Bittensor recently released Null Consensus, a major protocol update expanding subnet capacity from 256 to 2,500 participants. It introduces a new way to determine miner rewards, giving subnet owners an alternative to Yuma Consensus.

Targon's new GPU capacity is being booked by enterprise customers within a day, often under year-long contracts. On the latest Novelty Search, @TargonCompute explains how Subnet 4 is building a confidential AI cloud that protects models and data even on third-party hardware. The team also unveiled a public revenue dashboard, with bare-metal servers, virtual machines and AI-agent sandboxes already available, and more cloud services on the roadmap. Hosted by @const_reborn Full episode in the first comment

Novelty Search :: Bittensor Subnet 4, Targon xitter.cf/i/broadcasts/1aKbdErpe…

> Documentation: bittensor.com/docs/guides/nu… > Release: github.com/RaoFoundation/sub…

The update also introduces optional proof-of-work registration, allowing miners to join subnets without spending TAO upfront. Instead of paying a registration fee in TAO, they can solve a computational challenge using their hardware. This opens the door for more miners, including autonomous AI agents, to join Bittensor using compute alone.

Yuma remains the default. Subnet owners can opt into Null or keep the existing multi-validator consensus system. Null offers a simpler way to determine miner rewards, designed for much larger competitions. With both mechanisms available, Bittensor can support a wider range of subnet designs, each using the reward model that best fits its needs.

Without Yuma’s consensus and bond calculations, Null can handle much larger subnet competitions. It also supports precise integer weights, allowing validators to preserve very small reward ratios across thousands of participants. The 2,500-participant limit applies to subnets using one reward mechanism.

Null removes that consensus step. The registered participant with the highest stake weight becomes the only validator allowed to submit weights. The protocol uses those weights to determine miner rewards. Other participants with stake still receive proportional validator dividends.

Under Yuma, validators evaluate miners and assign weights based on their performance. The protocol compares these weights through stake-weighted consensus, limiting scores that exceed what the majority supports. Those adjusted weights determine how miner rewards are distributed.

Open robotics is bringing Bittensor and @peaq together. @openroboto (Subnet 80) now uses peaqOS to link robot training footage to capture devices with verifiable identities on @solana.

This Thursday on Novelty Search :: Subnet 4 :: @TargonCompute Targon is building a decentralized GPU cloud on Bittensor, providing on-demand compute for AI training and inference. Its confidential VMs use Intel and NVIDIA hardware security to protect sensitive models and data, even from the providers operating the servers. Live on Thursday :: 9PM UTC :: via Bittensor Discord Hosted by @const_reborn

We said cheaper storage was on Bittensor. @Hippius_cloud beats AWS, Google Cloud, R2, Wasabi and Backblaze on price at every size tested.

Looking for a cheaper way to put leading AI models to work on your computer? Bittensor subnet @say_gm_ just released a desktop AI assistant app for Mac, Windows and Linux. You can switch between Claude, GPT, Gemini and more, even mid-chat, at discounted rates and without a subscription. Bittensor’s growing product ecosystem is getting ready for the mainstream.