Engineering notes
Useful architecture, explained clearly.
Short, practical writing about agent systems, trust boundaries, failure paths, and when simpler automation is the better choice.
Latest notes
Thinking behind the builds.
Each article favors operational clarity over hype: what the system can access, how it fails, and what a team must own.
Architecture
Self-hosted AI agent architecture without the mystery layer
A five-layer model for channels, control planes, execution, tools, and operational readiness.
Read the note →OrchestrationMulti-agent systems need feedback loops—not agent theater
How to structure planning, parallel review, bounded retries, and human escalation.
Read the note →Keep exploring
Turn the ideas into a build
Choose the page closest to the outcome or constraint that matters most to your team.
Start with the workflow
Turn the manual work into a system.
Describe what happens today, where it slows down, and what a successful result should look like.