Technical notes
Working essays, research logs, and reading records from systems under investigation. Claims stay sized to their evidence; revisions remain visible.
Publication register
- Agentic systems
- Research practice
- Demos are not evidenceWhy a fluent video proves less than a boring table.Agentic systems
- Why we publish before we feel readySmall claims, stated early, are cheaper to correct.Research practice
- Building a small task suite for document workflowsScope notes for our first evaluation artifact.Agentic systems
- What we read this springFive papers that shaped how we think about agent evaluation.Agentic systems