I Keep Trying to Prove ShrekOS Doesn't Need to Exist

In a candid reflection on his project 'ShrekOS,' the author explores the tension between building a custom operating system and finding existing solutions for securing AI agents. The developer initially set out to create a safer environment for running autonomous agents but found himself architecting a full Linux distribution. Despite questioning the necessity of such a massive undertaking, the author argues that current tools—like containers and sandboxes—lack a cohesive, user-facing layer that allows humans to easily manage and visualize agent permissions. The project aims to create a boundary where agents can operate with specific, transparent access rights without gaining full authority over the host machine. While the author remains open to the possibility that ShrekOS might eventually be reduced to a smaller runtime or a set of best practices, he continues development because he has yet to find an existing desktop environment that treats AI agents as first-class, untrusted actors.