Confidential computing in this channel In the conversation with Alex Komoroske, confidential computing appears as a privacy architecture for cloud-based AI — secure enclaves with fully encrypted memory, paired with open attested runtimes, so users can remotely verify through hardware-signed attestation that the system they are talking to is unmodified and private. It is offered as an alternative to assuming privacy means local-first, which is architecturally hard and inconvenient. It also appears as one of the tools for breaking his iron triangle (untrusted code, sensitive data, network access — pick two), enabling contextual flow control. Where it fits - A pillar of private by design intentional technology, letting data be used only for and by the user. - A coordination primitive: many people can all verify the server does only what it says. - A leverage point alongside contextual flow control and information flow control for escaping data gravity and aggregation. [moment: 851] [moment: 1945] [moment: 1771] - Confidential computing = running cloud compute in secure enclaves with fully encrypted memory. - Combined with open attested runtimes, users get hardware-signed proof the server is private and unmodified. - It lets AI assistants stay cloud-based and private-by-design — without local-first complexity.