WHAT NEW IN X10.31 SAT

AI collections from X · 22 collections · 62,354 posts

jevVIDEO

REAX SN92: an open network hosting Jev-style decision models on Bittensor

REAX subnet 92 lets miners serve and validators verify Jev-style decision models on GPUs, and its DecisionModels product already has paying customers who have burned millions of tokens, plus an EU-compliant tier.

Jon Cipher
@_joncipherArchived 15 days agoOriginally posted 2026.10.15
Jon Cipher @_joncipherOriginally posted 2026.10.15
Open original on X

💫 REAX SN92 - THE REFLEX LAYER 💫 The Quiet Launch That Could Matter Most people are still arguing about which chatbot writes the best email. Meanwhile a small team just quietly stood up something more interesting on Bittensor: a dedicated network for the next layer of AI - the fast, structured decision layer that sits underneath the chatbots. They’re calling it REAX. Subnet 92. Powered by DecisionModels. Here’s the simple version. Normal AI models are great at writing. They’re expensive and slow when all you need is a clean yes/no, a routing decision, a score, or a quick judgment with a probability attached. A new class of models (people call them decision models, System 1 models, or Jev-style models) is built specifically for that job. One forward pass. Structured answer. Much faster. Much cheaper. REAX is the open network that hosts and verifies those models. Anyone with a decent GPU can mine. Validators check that the miner is actually serving the correct model faithfully and quickly. The first product - DecisionModels - already has paying customers who have burned millions of tokens. There’s an EU-compliant tier with real contracts and EU-hosted GPUs for companies that can’t touch the usual AI APIs. Florian (ex-Chutes, the guy who runs the independent JevBench that even Microsoft referenced) is running the project. Const holds the coldkey. Mining is live. Their main validator comes online soon. They’re already talking about competitions to make the inference engine even faster, and they’re pushing an unusually simple onboarding path: point an AI agent at the mining docs and let it set you up. This is early. Very early. But the category itself is only a few weeks old, the product already has organic paying usage, and the team has real experience shipping decentralized inference. Enterprise conversations are apparently already happening. While most of the market is still focused on who can generate the longest text, this one is trying to own the fast, cheap, structured decisions that agents and companies actually need thousands of times a day. Worth watching. #TAO #Bittensor #AI #DecisionModels #REAX #SN92 $TAO $SOL #Solana $COIN #Coinbase $HOOD #Robinhood #RobinhoodChain

x.com/_joncipher/status/2110540424641126872Archived 2026.10.16 10:06