Right now, Alpha seems to be showing some signs of recovery. At least the new tokens have broken free from the 30U curse. New projects are few, but their quality has improved. Unlike some previous LJ projects that launched and then ghosted.
On-chain, the new token Nesa (NES) is set to launch tomorrow at 20:00. The project is focused on a privacy-preserving decentralized AI inference network, and their official Twitter has around 340k followers, indicating some hype. Funding details haven't been released yet; wallet Booster tasks are live now, requiring 2 points to participate, with 50,000 slots available, and each participant gets 20 $NES. The answer to the task is: ACBC.
Today, let's continue discussing the PIPE module associated with @OpenGradient . It hasn't officially launched yet, but my first thought is that if this thing works, the entire on-chain AI game will change.
Currently, on-chain applications using AI follow what path? The oracle model. Your smart contract needs AI judgment, sending requests off-chain. The model runs off-chain, and then the results are fed back on-chain via an oracle. This process involves at least two transactions, with a time lag in between. If the on-chain state changes while waiting for results, your AI judgment could be outdated.
What PIPE aims to do is execute AI inferences directly within the transaction. Your Solidity contract calls a precompiled contract, with model inference completed as part of the transaction state transition. The result becomes part of the on-chain state, not an external oracle callback. This means that AI judgments and on-chain operations are atomic—either both succeed or both fail, with no intermediate state.
Consider how DeFi liquidations are currently handled: the risk control model runs off-chain, the oracle sends back results, and then the contract executes the liquidation. But what if the price changes in those few seconds? Your liquidation could happen at the wrong price. PIPE's atomic execution packages risk control inference and liquidation operations into the same transaction, executing the liquidation the instant the inference result is available, eliminating time lag.
This makes me feel that AI on-chain shouldn't just be about running models off-chain and stitching results together on-chain; instead, the inference itself should be part of the on-chain execution. #OPG is pushing in this direction, and $OPG is fueling the entire verifiable AI inference network. I'll keep a close eye on its value realization.
On-chain, the new token Nesa (NES) is set to launch tomorrow at 20:00. The project is focused on a privacy-preserving decentralized AI inference network, and their official Twitter has around 340k followers, indicating some hype. Funding details haven't been released yet; wallet Booster tasks are live now, requiring 2 points to participate, with 50,000 slots available, and each participant gets 20 $NES. The answer to the task is: ACBC.
Today, let's continue discussing the PIPE module associated with @OpenGradient . It hasn't officially launched yet, but my first thought is that if this thing works, the entire on-chain AI game will change.
Currently, on-chain applications using AI follow what path? The oracle model. Your smart contract needs AI judgment, sending requests off-chain. The model runs off-chain, and then the results are fed back on-chain via an oracle. This process involves at least two transactions, with a time lag in between. If the on-chain state changes while waiting for results, your AI judgment could be outdated.
What PIPE aims to do is execute AI inferences directly within the transaction. Your Solidity contract calls a precompiled contract, with model inference completed as part of the transaction state transition. The result becomes part of the on-chain state, not an external oracle callback. This means that AI judgments and on-chain operations are atomic—either both succeed or both fail, with no intermediate state.
Consider how DeFi liquidations are currently handled: the risk control model runs off-chain, the oracle sends back results, and then the contract executes the liquidation. But what if the price changes in those few seconds? Your liquidation could happen at the wrong price. PIPE's atomic execution packages risk control inference and liquidation operations into the same transaction, executing the liquidation the instant the inference result is available, eliminating time lag.
This makes me feel that AI on-chain shouldn't just be about running models off-chain and stitching results together on-chain; instead, the inference itself should be part of the on-chain execution. #OPG is pushing in this direction, and $OPG is fueling the entire verifiable AI inference network. I'll keep a close eye on its value realization.