Solana has a development worth watching, but don’t mistake devnet testing for a mainnet launch. According to a report by Jinse Finance (Beijing time, November 15, 2026), Solana Confidential Compute, an open-source tool developed by the Solana Foundation and the Nito team, includes an SDK and an on-chain verifier. It is designed for private AI applications and verification of computing hardware. It is currently available only as a beta on devnet; no date has been announced for the mainnet version, and it has not yet been audited.
Under the mechanism, AI computation remains off-chain, and the output is signed with a key. After verifying the signature, an application can check the on-chain record. Private inputs are not made public, but that does not mean all information is invisible. The report notes that an application server can see visitors’ IP addresses, while Nito can see accounts, models, request times, and ciphertext sizes. The privacy boundary depends on the details; simply calling something “private” is not enough.
The market is a separate matter: over the rolling 24-hour period, $SOL -10.36%. Over the rolling 24-hour period, $JUP -6.60%. Over the rolling 24-hour period, $RAY -9.73%. These are changes in the respective spot prices, not data on tool usage, and there is no evidence that this development caused the market moves. Trading access, aggregation, and liquidity address different parts of the process; opening a test tool does not by itself indicate real usage or retention.
Next, I’ll be watching how the beta evolves, whether an audit is completed, and whether the mainnet plan becomes clearer. If developers integrate it, we’ll also need to see whether their applications solve real problems and whether users keep using them. Without such evidence, for now it’s best to treat this as a development tool under testing, rather than a new capability already deployed across the entire ecosystem.
Under the mechanism, AI computation remains off-chain, and the output is signed with a key. After verifying the signature, an application can check the on-chain record. Private inputs are not made public, but that does not mean all information is invisible. The report notes that an application server can see visitors’ IP addresses, while Nito can see accounts, models, request times, and ciphertext sizes. The privacy boundary depends on the details; simply calling something “private” is not enough.
The market is a separate matter: over the rolling 24-hour period, $SOL -10.36%. Over the rolling 24-hour period, $JUP -6.60%. Over the rolling 24-hour period, $RAY -9.73%. These are changes in the respective spot prices, not data on tool usage, and there is no evidence that this development caused the market moves. Trading access, aggregation, and liquidity address different parts of the process; opening a test tool does not by itself indicate real usage or retention.
Next, I’ll be watching how the beta evolves, whether an audit is completed, and whether the mainnet plan becomes clearer. If developers integrate it, we’ll also need to see whether their applications solve real problems and whether users keep using them. Without such evidence, for now it’s best to treat this as a development tool under testing, rather than a new capability already deployed across the entire ecosystem.
