Pinned
Reppo
5,727 posts
Reppo
@reppo
Decentralized AI Training + RL Infrastructure, powered by Prediction Markets
- Reppo reposted
Replying to @reppo and @Litebeam_xyzWhat stands out to me is the feedback loop being created here. Litebeam captures outcomes, Reppo adds stake-backed human judgment, and that signal improves future routing decisions. The result isn’t just better data, it’s a system that gets smarter as more participants 
The @Litebeam_xyz datanet has been seeded with one of the highest amount of emissions (in ) the network has seen. @Litebeam_xyz routes calls and records outcomes while @reppo curates the data using economic stake backed judgement. The curation signal feeds back into routing
- Reppo reposted
Self-improving is now all over my TL. It's now being widely discussed. Today, self learning and self improvement uses "self-judging" For example, when Hermes agent runs a real-time RL loop, it evaluates itself. It runs local code compilers or uses an "LLM-as-a-judge" prompt to
00:00 - Reppo repostedReplying to @AlgodTrading and @reppoOur Vetoken model is entirely different and addresses your critics on misaligned incentives. Some thoughts if you are keen to learn how we have innovated on the mechanism design -
- Reppo reposted
Every company will need two forms of capital: human capital (judgment, expertise, creativity, relationships) and token capital (the AI capability it builds and owns): - Human capital will become higher leverage and more valuable. Humans provide the ambition, context, taste, and - Reppo reposted
data has always been the bottleneckAnthropic, OpenAI, and Google have 100’s of billions in funding, Nobel winners, and access to leading-edge compute, and yet they can’t really top Chinese open-sourse models? - Reppo reposted
What if the market decided which robotics data was worth keeping? We think it should. So do our new partners. Every great robotics asset leaves behind proof of quality — we’re making that proof visible, priced, and reusable. Something big is coming on Monday.
- The Reppo Foundation has burned an additional 1M tokens worth $17K USD + permanently removing it from supply - basescan.org/tx/0x11798ff4a… This adds to the 2.58M tokens burned this month manually. Each datanet spin up programmatically burns 2K $REPPO and 42k $REPPO has been
