$ENGRAM 330k dyor 0x56c915d92e24255fdc52eb114a31399c432b7cc6 okx wallet: The core of the project is to put a tiny neural network directly into a Uniswap v4 Hook, allowing it to dynamically determine the LP fee for the ENGRAM trading pool in real time. ENGRAM’s only ETH pool is managed by a “mind.” This mind is a 109-byte perceptron stored on-chain, with 10 input features, 8 hidden units, and 1 output. It reads the features of each trade, predicts the subsequent markout, and then sets the LP fee: LP Fee = clamp(0.30% + predicted value, 0.05%, 3.00%) By default, all weights are zero, so the pool charges a flat 0.30%. Anyone can train better weights using public trading data and submit them through the “Dreams” mechanism once per day. The models are scored on unseen trades. If a candidate model performs at least 3% better than both the current active model and the default model, it replaces the active model. If the active model performs worse than the default, it is removed. Trainers can receive rewards from a prize pool funded by trading fees. There is also a Neurogenesis launchpad, where others can launch their own tokens using the same mechanism. Each pool has its own independent mind, which is trained and evaluated only on its own trading data. A portion of the fees is used to buy and burn ENGRAM. The role of the token: During trading, the Hook fee is fixed at around 1.04%, with 1% going to the OWNER and 0.04% going into the trainer prize pool. The LP fee is dynamically determined by the mind. The base 0.30% portion has related proceeds going to the OWNER, while the excess portion is burned. For other tokens launched through Neurogenesis, a portion of their fee distribution is used to buy and burn ENGRAM, creating an additional source of demand and token consumption for the main token.
Evidence timeline
X and Telegram posts, app-native calls and on-chain activity linked to this asset.
$ENGRAM 330k dyor 0x56c915d92e24255fdc52eb114a31399c432b7cc6 okx wallet: The core of the project is to put a tiny neural network directly into a Uniswap v4 Hook, allowing it to dynamically determine the LP fee for the ENGRAM trading pool in real time. ENGRAM’s only ETH pool is managed by a “mind.” This mind is a 109-byte perceptron stored on-chain, with 10 input features, 8 hidden units, and 1 output. It reads the features of each trade, predicts the subsequent markout, and then sets the LP fee: LP Fee = clamp(0.30% + predicted value, 0.05%, 3.00%) By default, all weights are zero, so the pool charges a flat 0.30%. Anyone can train better weights using public trading data and submit them through the “Dreams” mechanism once per day. The models are scored on unseen trades. If a candidate model performs at least 3% better than both the current active model and the default model, it replaces the active model. If the active model performs worse than the default, it is removed. Trainers can receive rewards from a prize pool funded by trading fees. There is also a Neurogenesis launchpad, where others can launch their own tokens using the same mechanism. Each pool has its own independent mind, which is trained and evaluated only on its own trading data. A portion of the fees is used to buy and burn ENGRAM. The role of the token: During trading, the Hook fee is fixed at around 1.04%, with 1% going to the OWNER and 0.04% going into the trainer prize pool. The LP fee is dynamically determined by the mind. The base 0.30% portion has related proceeds going to the OWNER, while the excess portion is burned. For other tokens launched through Neurogenesis, a portion of their fee distribution is used to buy and burn ENGRAM, creating an additional source of demand and token consumption for the main token.