Revenue Search returns with the usual chaos and banter, then introduces Josh and the launch of Green Compute—a new Bittensor compute subnet designed specifically for enterprise-grade inference workloads. Josh shares his background building and selling GPU infrastructure in the UK since 2017, and explains that Green Compute plugs into an already-profitable compute business with existing customers, contracts, and deployment experience—so the subnet isn’t starting from zero.

The core thesis: bring data centers to constrained renewable energy. Across the UK (and beyond), farms and renewable sites often generate power the grid can’t accept—so it’s wasted. Green Compute turns that stranded power (solar, wind, hydro, and especially anaerobic digestion / biogas) into usable AI compute, offering site owners far higher returns than exporting electricity back to the grid. Unlike “spot” compute markets, Green Compute is aiming at longer-term, high-volume enterprise deals that require symmetry (large clusters of identical GPUs, networking, CPUs/RAM/storage) plus real human support (sales + engineers) and predictable uptime—things many existing marketplaces struggle to guarantee.

They also touch on tokenomics and onboarding: compute can be bought with fiat, but the goal is to push real-world customers toward paying via subnet alpha over time (creating buy pressure). Mining is gated by standards (e.g., high-bandwidth connectivity and matching hardware) to meet enterprise requirements, with a process for miners to apply and be verified—including the “green” power source. The team plans to update naming/branding and community channels shortly, with more details and access via the Green Compute website.

Revenue Search returns with the usual banter (and a bit of tech lag) before welcoming Bitrecs (Subnet 122)—a small, doxxed team building an LLM-powered product recommendation engine for e-commerce, starting with Shopify. Dimitri (CEO), Max (CTO) and Arsham (CRO) explain how their widget boosts store performance by generating smarter “you might also like” suggestions, then cleaning messy LLM outputs with a consensus/ranking layer. They show live examples on real stores, including a unique feature: explanations (“reasoning”) displayed to end-users for why each recommendation was chosen.

They also introduce Bitrecs V2, which separates the product into a fast Web2 inference layer (serving real-time recommendations) and a Bittensor “intelligence” layer where miners compete in a winner-take-all prompt-template (“artifact”) evolution game. Bitrecs shares business traction (~130 customers), metrics (avg ~$32/month ARPU, ~$75 CAC, ~1% lift so far with a goal of 2–5%+), and a clear growth plan: deploy a six-figure marketing budget, aim for ~1,000 stores, ship a self-serve API for non-Shopify/enterprise use, and (once trust + lift improve) transition toward performance-based billing / revenue share so stores pay only on measurable uplift.

Revenue Search is back after the Bittensor San Francisco event, and this episode is a first: a dual-subnet session with Yanez (SN54) and BitMind (SN34). Jose and Ken announce a partnership aimed at tackling the rapidly growing threat of deepfake-driven identity fraud—the kind of attacks that can bypass KYC, liveness checks, and even enable high-value social engineering scams.

In short: Yanez produces high-fidelity, well-annotated synthetic identity/face data and attack vectors, and BitMind uses that to train and improve face-focused deepfake detection models via their subnet. They’ll take the combined “data + detection” stack to enterprise customers (financial institutions and identity providers), typically via licensing/usage-based deals, with both teams reinforcing that real-world revenue supports their subnets (including alpha buybacks into treasury) while keeping flexibility for future DeFi/treasury use.