Specialized miner workloads
We prioritize tightly scoped tasks like retrieval, ranking, and verification so miner output is measurable, comparable, and worth paying for.
Subnet Portfolio
Subnets that earn their place in the network.
Rendix runs live Bittensor subnets where incentives, utility, and technical defensibility reinforce one another — three in production today, with more in development.
Design criteria
Useful outputs, healthy participant incentives, and enough architectural clarity that the subnet can improve over time.
Live Subnets
Subnets built by Rendix Network
Three live Bittensor subnets — Leoma (SN99), Eirel (SN36), and NexisGen (SN70) — built for durable utility and strong mechanism design.
LiveSubnet 36
eirel.ai
A Bittensor subnet where miners submit competing AI agents that plan, create, and ship across language, vision, audio, and code. Validators score them with an LLM judge and hard-gated, multiplicative scoring, so the best agent for each task wins.
LiveSubnet 70
nexisgen.ai
A Bittensor subnet that runs like a commissions house for AI training data: buyers commission a dataset spec, a decentralized network of producers fulfills it, and validators certify every record. You get the output with full, on-chain provenance.
LiveSubnet 99
leoma.ai
A live Bittensor subnet for AI video generation, where miners run Text-Image-to-Video models and validators score every output with Gemini. The best miner each round earns the subnet's alpha — winner-take-all.
Focus Areas
Where Rendix can create leverage
Rendix focuses on subnet ideas where miner behavior, validator scoring, and end-user utility reinforce each other over time.
Validator-grade evaluation
Each concept starts with scoring design: reproducible benchmarks, adversarial checks, and incentive logic that rewards honest validation.
Developer-facing utility
A subnet should expose practical outputs through clean APIs and integration paths so builders can use it directly in real products.
Build Principles
How Rendix decides what to ship.
Our subnet roadmap favors mechanisms that can survive open competition: clear tasks, transparent scoring, and repeatable user value.
Incentives before features
We define the reward surface first, then design features around the exact miner and validator behavior the subnet needs.
Defensibility through data
Durability comes from better data pipelines, tighter evaluation loops, and operations that compound quality every tempo.
Operational readiness
We plan for validator uptime, monitoring, abuse resistance, and fast iteration so the subnet performs under real network pressure.
Back to the brand story
Return to the homepage for the full Rendix narrative.
See how the homepage frames Rendix: a team building practical Bittensor subnets with strong mechanism design and execution.