Building
- Safer interfaces between AI agents and production infrastructure
- Durable tools that improve continuity across coding-agent sessions
- Personal agent systems that remain local, inspectable, and useful
Investigating
- How approval policies should evolve as agents gain more context
- Which forms of memory create signal instead of accumulating noise
- How small teams can adopt AI without creating an invisible operations burden
Writing
Field notes on agent architecture, workflow automation, and practical engineering trade-offs.