Devrandom is infrastructure for AI agents that improve how they work while remaining accountable and within user-defined permissions. A user defines a task, success criteria, and limits. The agent works within those boundaries, preserves execution evidence, and proposes improvements to its instructions, skills, workflows, or memory strategy. Changes must pass independent evaluation and receive authorization from a separate Governor operating under a scoped Promotion Mandate. The agent cannot approve its own changes, expand its permissions, or erase failed evidence. The system is designed to preserve verified progress across restarts and share sanitized improvements without transferring private data, identity, or authority. Built with cryptographic identity and durable, searchable evidence, Devrandom lays the foundation for trustworthy long-running agents, reusable agent behaviors, and future delegated compute.
Team
Placement
Participant
