Governing an agent that rewrites itself
An agent that edits its own code needs the same review as any other change. What safety research says about reward hacking, sandboxes, oversight, and typed contracts.
· 8 min read
Topics
Code graphsAI agentsCoding agentsSelf-improving modelsSelf-evolving agentsAutonomous agentsProduct ReleasesLatest
A prepaid wallet for agent spendCloud sandboxes for coding agentsDecisions in one place: the Console and the Work boardEvidence an auditor can checkPaperField manualRSS feedOxagen on Hugging FaceResearch · Pillar
Why agents work better when facts are linked.
A graph database stores each fact with its links to other facts: this function calls that one, writes this table, and changed in this commit. An agent can follow those links in a few steps. A search over text can only find passages that look like the question. Posts cover code graphs, graph retrieval, graph memory for agents, the ontology that types each fact and link, and what the research measures about each.