The hosts open with travel/gossip (Mark dialing in from Dubai) and a firm PSA about why they won’t confirm wallet/subnet rumours, They then bring on Yoav, Garrett and Gyles from Tensor Group to debut tao.com, a rebuilt Bittensor wallet (iOS live now; Android targeted for Q1) that makes onboarding dead simple: fiat/TAO toggles, in-app TAO purchase via Coinbase Pay (debit or Coinbase), one-tap staking (abstracts “root/validators”), biometric Secure Enclave keys, clean portfolio/history, and rich subnet pages (stats, team, roadmap, news, search/sort) designed to drive conviction; UK rollout awaits FCA tweaks while US users can update today, CSV export/desktop and Ledger-style “power user” features are planned, and subnet teams will be able to self-manage profiles and potentially promote content; the group discusses growth levers (recommendations, short founder videos, category labels, “did-you-know” hooks), a playful idea of tiny “starter alpha,” and a serious unmet need for institutional/custodial support so PLCs can hold alpha, with Tensor saying they’re self-funded (validator/mining), now raising to scale the broader tao.com/Tensor stack.

Mark and Siam open with a candid TAO update (over-levered loans due, price drawdown, macro notes on TGA and QT) then bring on Akshat from Dippy, who explains Dippy as an AI-friend entertainment app (~8.6M users, ~1 hr/day engagement) adding tap-to-video (not real-time yet) and imminently voice calls; Dippy now runs all text inference on Bittensor SN-4 (Targon) via a six-figure deal and is pivoting SN-11 into a fast, cheap media-inference “studio” (deterministic TensorRT pipeline), currently serving ~2% of Dippy images with plans to ramp to 100% and open a self-serve API; they’ll resume consistent SN-11 alpha buybacks by redirecting $5–10k/month formerly spent on centralized providers, outline 18+ moderation/privacy measures, note revenue of ~$40–60k/month while prioritizing retention and future in-message ads, and share a longer-term vision for interactive “generative worlds,” plus team hiring and platform constraints (Android feasible “Jarvis,” iOS restricted).

Mark and Siam join Max (Score / Subnet 44) for an update on Score’s shift from “just sports” to a broad computer-vision platform: they’ve built a new incentive mechanism that uses VLMs to generate pseudo-ground-truth and run twin tracks—an open, verifiable Hugging Face competition and a private client track—driving rapid gains toward a football “gold line” benchmark. Score’s first featured client, cricket strategist Nathan Leamon (Cards), explains how Score will replicate/extend Hawkeye-style ball-tracking from standard broadcast footage and power decisions from scouting/auctions through in-game tactics. Max outlines real-world uses beyond sport (petrol forecourts, retail, fruit grading, car washes), a 60-day trial motion to win enterprise data and contracts, and the upcoming “vision GPT” product where an agent reads video, recommends/dispatches models as subnet tasks, and ties revenue to ALPHA via a burn-and-mint “scoronomics” loop. The team notes a fresh DSV/Astrid-127 OTC with Score, and—crucially—Score’s first recurring five-figure invoice, underscoring real revenue traction.