
Thread 1/? **Building tokenstats: A sovereign inference market observability layer** Context: Even with a solid home lab, I still burst to @routstrai and external providers for frontier models, long context, or when mobile. Pricing is volatile, fragmented, and hard to track in real time. Existing tools are either static or miss the permissionless Nostr/Routstr layer I actually prefer and use. I want better data for my routing decisions in arbstr (arbstr.com) and for future agents. Action: I’m building tokenstats: live observability + a lightweight oracle for inference costs (input/output in sats), latency, and node health. Rust core, Nostr listener for Routstr announcements, provider polling, simple Axum dashboard, and @getpubky -based attested snapshots on the roadmap. Result (the goal): - Real-time market awareness that feeds directly into arbstr - Reusable components for sovereign data systems - Public proof of building real-time observability in the permissionless AI space I'm treating this as a focused learning project following up from my recent launch of @oddlympics . I'll post regular updates on progress, problems, and lessons. What metrics would you want in a live inference pricing dashboard? Especially if you run hybrid local + Routstr setups? #BuildInPublic #Nostr #Routstr #SovereignAI
















