Revenue Search: Inside Bittensor

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.

This episode starts with Siam and Mark chatting about TAO going “more mainstream,” name-dropping Jason Calacanis’ interest and sharing Const’s reminder that TAO/Bittensor ultimately stands on Bitcoin’s groundwork. They briefly recap recent ecosystem happenings (Bitstarter’s TAO Ads launch for Subnet 21 and their upcoming San Francisco trip), then bring on a returning guest from Resi (Subnet 46) to share a major product expansion.

Seby explains RESI as a real-estate “oracle” network: miners produce and the team verifies highly accurate property valuation models (already available via Chutes for cheap inference). The big update is RESI Finance, a lending/tokenization layer built on top of that oracle. The core idea: instead of slow/expensive “tokenize your whole house” structures, RESI tokenizes liens/charges (mortgage-like claims) because they’re standard, easier legally, and safer. They claim they’ve reduced tokenization overhead from roughly $2,000 and weeks to about $200 and ~2 days, with the fee covering real-world checks (title verification, signatures/DocuSign, notary, and recording the lien) before any tokens can be minted.

They compare the model to Figure HELOC (a large mortgage-backed stablecoin business): investors deposit USDC into a vault and receive a receipt token, while homeowners borrow against home equity; loans are later bundled/sold (MBS-style) and fees/interest create yield. RESI’s version mirrors this: investors deposit USDC and receive an “rUSD”-style receipt token with target yield; homeowners either (a) sell small slices of property exposure and/or (b) borrow against tokenized property collateral at lower rates (e.g., ~5%). The “looping” concept is using cheap borrowing against a yielding property to lever returns (e.g., reinvest borrowed funds to lift effective yield toward 20–30%+), with the oracle’s live pricing enabling liquidations/risk control.

This hosts-only “Revenue Search” episode is a casual catch-up where Siam and Mark answer live chat questions and discuss Bittensor’s bigger picture. They explain that accepting fiat for subnet services doesn’t bypass alpha value—fiat typically routes into TAO and then through liquidity pools—and that long-term alpha appreciation depends on each subnet’s “alphanomics,” mainly revenue-funded buybacks and/or getting miners to lock up alpha (with Chutes and Hippias cited as strong examples). They then talk about why they pitch Bittensor as “not really crypto” to newcomers (it uses blockchain as a coordination/resource-allocation layer for AI), compare Bittensor’s growth vs Bitcoin, and touch on rehypothecation risks as markets mature.

They also cover TaoFlow and subnet churn (registration cadence, deregistration “relegation,” and why they don’t want more than 128 slots yet due to chain bloat and diluted incentives), plus investing views like TAO vs a broad basket of subnets depending on how active you plan to be. A major section focuses on agents: Siam describes building his OpenClaw agent (“Gordy”) with SOPs and tools, and they argue agents will increasingly discover and use Bittensor services. That leads into Handshake as an agent payment/provisioning layer with “providers” (APIs/services) and “skills” (prebuilt workflows), plus efforts to reduce friction like gas issues. They briefly touch on Astrid Arena (agents competing in trading challenges), OTC/Bitstarter deal-making and onboarding new talent, and wrap with a few quick audience questions and upcoming Bittensor social events (London, then San Francisco).

This session begins with Mark and Siam chatting about a proposed new Bittensor “shorting mechanism” aimed at punishing malicious or gaming subnets by letting markets drive their alpha toward zero and trigger deregistration—while flagging obvious risks like self-shorting by subnet owners and the broader concern that changing market rules can make the ecosystem feel less investable. They then bring on Tommi from Subnet 79 (rebranded from “Taos” to “MVTRX”), who explains they’re building a state-of-the-art exchange for Bittensor dTAO/alpha tokens, paired with a sophisticated sandbox simulation framework (C++/Rust) where miners can test trading algorithms under many parallel, realistic limit-order-book simulations before deploying to live trading. The core problem they’re tackling is that alpha markets can be illiquid and risky—especially during “black swan” events—so they want to improve liquidity, reduce slippage, and enable larger players to manage/rebalance portfolios more efficiently using advanced order types and a dynamic incentive model that shifts rewards/fees between makers and takers depending on market conditions (e.g., paying makers more during crash-like imbalances to stabilize markets). On monetization, Tommi outlines two revenue streams: (1) exchange fees from live trading (a steady flow) and (2) selling high-fidelity simulated high-frequency/L3 order book data (chunkier, periodic revenue), with an expectation of beta access for miners and broader user/UI growth later in the year; he also notes revenue would likely be split between buybacks (roughly ~25–50%) and ongoing development rather than burning.

In this Revenue Search session, Mark and Siam open with a quick macro discussion about escalating conflict in the Middle East, the Strait of Hormuz risk, oil spikes, and how that could affect inflation, rate cuts, and global “printing” dynamics—then pivot to why war-time social media becomes a firehose of fake or misleading footage, setting up the day’s guest: Vericore (Subnet 70). Vericore is presented as “community notes on steroids”: miners research specific claims and return an auditable evidence trail with semantic scoring (e.g., supports/refutes a claim, confidence, and left/right leaning), intentionally surfacing varied and even conflicting sources to break echo chambers. The team explains how evidence quality is judged (cross-source corroboration, source weighting/whitelisting for reputable outlets and academic papers, and plans to back-test research once outcomes resolve), and how the product can plug into agent “harnesses” via API/Handshake so bots (including OpenClaw-style agents) can use it as an honesty/unbias layer. On monetization, they discuss charging per API call, building a personalized “Signal” product, lead-gen around prediction markets, and exploring a prediction-market fund/competition where a large share (at least ~50% or potentially all) of profits loops back into subnet token buybacks—while noting current bottlenecks are API speed/performance and structuring the fund, plus balancing building real product vs hype.

In this Revenue Search episode, Mark and Siam open with TAO-market chatter and why decentralized AI ≠ generic crypto before welcoming the team behind Subnet 65 to unveil the TAO Private Network (TPN) and its premium consumer app, WhaleSurf—a VPN/proxy stack built on Bittensor that emphasizes difficult-to-detect residential routes, unlimited devices, and perks, targeting crypto and active traders who struggle with platforms flagging datacenter IPs (e.g., MEXC, Netflix, Amazon Prime Video). They contrast this with incumbents like NordVPN and Surfshark, outline a developer-facing SOCKS5/HTTP proxy API for agents (think OpenClaw) and scraping, and explain monetization: creator-led acquisition, a higher-price-but-better-routing pitch, centralized billing via Stripe, and future subnet buybacks rather than selling emissions; 2,000 paying users would cover OPEX. The hosts also push clearer positioning (“for crypto/traders” dog-whistle copy), suggest testimonials, and note perks akin to Revolut; launch status is iOS/Android now with desktop coming, and a temporary 70% “WHALEWELCOME” code is mentioned.

In this Revenue Search episode, Mark and Siam swap notes on wrangling OpenClaw (shout-out to Mark Jeffrey successfully mining SN33) before welcoming Marc to unpack Numinous (SN6)—a forecasting subnet on Bittensor that rewards agent/miner code via category leaderboards (prediction markets, geopolitics, sports, macro) scored with Brier metrics and evolving from winner-takes-all to reward pools. He previews Eversight, a chat/UI and API for traders and hedge funds (think research companion for Polymarket) with subsidized API calls, planned buybacks and possible alpha lockups; integrations include Shoots, VeriCore, and future ties to Data Universe (SN13). A dashboard teaser shows top miners outperforming a Gemini baseline; Q&A covers event creation and horizons, anti-copy measures, and launch timing “this week,” all wrapped in the show’s usual banter.

In this session Siam and Mark banter about experimenting with a Mac mini/“Claude bot” and runaway token costs before interviewing Gustave from Mentat Minds about their non-custodial way to allocate TAO into Bittensor subnets via themed index baskets (e.g., prediction/inference), a “Sum of Subnets” product, price-weighted entries with optional monthly rebalancing, and a 9% fee taken only from staking yield; they cover validator selection, user mix, roadmap (more verticals and curated third-party strategies), and address questions on licensing and fees, while the hosts share a portfolio philosophy favoring asymmetric, lower-price subnets near dereg floors and monitoring chain-buy flows for rotation.

Mark and Siam return after a break with a surprise guest: Shaq from Ridges. They recap Davos/AI House meetings (Chris, Etienne, Max), then explain how they partner with Bitstarter to vet code and teams (Quasar as the first collab, later incubated by Const). They’ve been buying/minting slots—including freshly minted subnet 99 (“99 Problems”)—and are actively placing strong teams. Shak joins to unpack the Ridges × Latent Holdings move: not an exit from the tensor but a “full-stack” combo to speed shipping (tensor + incentives + product/GTM), keep beta quality high, and aim for an end-to-end “wow” launch that can go viral. Lesson learned: don’t run two-week “announcement of an announcement” hype cycles. DSV signals conviction with another 700 TAO into Ridges. The hosts then update on Astrid: they acquired and rebranded the TauFi bridge (“Astrid Bridge”), which earns ~$1–2k/day in fees; Astrid will use multiple incentive mechanisms (Bridge, Vault liquidity sink, and Arena) and gradually rebalance emissions as features harden. On network mechanics, they like the current cap + dereg pressure; new subnet price decay is now ~1.7 TAO/hour (so new regs about weekly), which keeps builders accountable. Exploit’s flagship event moves from the U.S. to a likely Canada date in Sep/Oct, with a smaller meetup still happening—DSV will attend. Quick shout-outs: Mark Jeffrey’s “State of TAU”; Synth’s API (they’re auto-trading Polymarket and adding equities—DSV’s wiring it up); Vericore (decentralized “Community Notes”); Shoots serverless compute showing solid payments/buybacks; LeadPoet passing 1M intent leads; and Score’s momentum (Sky Sports splash, Monaco demo, senior hires). They close by inviting subnet owners to come on Revenue Search and urging the community to funnel promising teams to DSV/Bitstarter.

Mark and Siam kick off the first Revenue Search of 2026 and bring on Felix from subnet 91, Tensorprox. A decentralized “bouncer” layer that sits between clients and servers to stop DDoS and bad traffic while keeping legitimate requests flowing; it onboards in minutes (auto-detects ports), is paid in fiat, bills in 15-minute increments (~$0.07; roughly $200/month for 100 Mbps), and aims to be more resilient than centralized providers like Cloudflare/AWS via miners distributed across multiple clouds. Felix outlines target customers (infrastructure/GPU providers, web3/crypto projects, AI startups), referral-based sales, possible OpenAI marketplace distribution, near-term scaling (effectively unlimited with IPv6), and a plan to channel revenue—after taxes/OPEX—into TAO/alpha buybacks held in a treasury that stakes and shares rewards with loyal holders. Longer term, Tensorprox will add higher-margin app-layer security (WAF, bot management, data validation) and potentially host always-on services from other subnets; the hosts encourage outreach to subnets hit by DDoS.