Comment by storystarling
Comment by storystarling 5 days ago
I suspect the endgame is confidential computing for distributed systems. If you are running high value workloads like LLMs in untrusted environments you need to verify integrity. Right now guaranteeing that the compute context hasn't been tampered with is still very hard to orchestrate.
That endgame has so far been quite unreachable. TEE.fail is the latest in a long sequence of "whoever touches the hardware can still attack you".
https://news.ycombinator.com/item?id=45743756
https://arstechnica.com/security/2025/09/intel-and-amd-trust...