Nieuwe data-activaklasse die zich voor onze ogen ontwikkelt. Denk niet dat er iemand beter gepositioneerd is dan Shaga om te profiteren van gaming frames data. Stuur cloud gaming naar de massa's.
🇮🇹 uncleG 'latency guy' | ☁️🎮 Shaga.xyz (d/acc)
🇮🇹 uncleG 'latency guy' | ☁️🎮 Shaga.xyz (d/acc)12 aug, 17:01
Frames+controls is the scarce feedstock for Genie3‑class world models. Not scrapable. Must be mined live. Shaga to date: 236,936 hours since March. Current run rate: - 520.6 hours/day. - 165 straight days. 6,067 users. - Registrations up 510% (10.1/day -> 61.7/day). - Node:client 6.7:1 so capacity headroom is real. Avg session 5.1h ≈ 1.10M frames at 60 fps. Value ladder for the network: $1/hour -> $236,936 implied $5/hour -> $1,184,680 Early broker quotes $50–100/hour -> $11.85–23.69M Unit economics: - 500 W rig at $0.20/kWh ≈ $0.10/hour power - $1/hour GAP -> ~$0.90/hour gross per active rig - $5/hour -> ~$4.90/hour Why human > synthetic for this modality: Synthetic is not instant. You still simulate and render every frame. That is compute/render bounded. Humans already render at home for fun. Frames are identical whether the policy is human or bot; only the actions can be synthetic. You still pay the rendering bill either way. (!!!) Scale math: - 1 player at 60 fps = 216,000 frames/hour - 1M players x 1 hour/day = 216B frames/day - 720p60 storage on the order of a few PB/day Freshness compounds daily as patches and metas shift. That is the moat. Bottom line: Frames+controls is a real data asset with pricing power. Imagine getting paid to play Cloud Gaming... Neural Game Codecs
6,42K