underrated AI protocols i’m interested in first thing first, a very simple thesis keep me in touch useful AI needs cheaper compute, systems that learn from experience, agents that can operate real markets i’m interested in the protocols building those layers and how their tokens capture the resulting demand $VVV, $NEAR, $TAO offer useful precedents and they all outperforming - VVV has subscription-funded buybacks - NEAR connects staking to AI credits - TAO subnets supply useful AI workloads that’s why these 3 caught my attention: [1] @rei_labs | $REI adaptive AI that learns from measured outcomes its robotics simulation research shows improvement on unseen layouts [2] @dphnAI | $POD cheap decentralized AI inference powered by idle GPUs from the team that co-developed Venice’s uncensored model commitment to use inference revenue for POD buybacks [3] @Moshdottrade | $BUN AI market makers managing permanently locked launch capital opening-bundle backers earn trading fees, changing their incentive from selling an allocation to supporting an active market all eyes on ‘em if you missed VVV