Building agents that turn experience into capability.
构建将经验转化为能力的 Agent。
I learn by building AI-native systems, with agents as teammates.
I'm exploring how agents use tools, check their work, and carry useful experience into the next task. I'm especially interested in memory and skills—not just how to store them, but how to tell whether they actually help.
Outside agent systems, my interests include self-hosted infrastructure and Minecraft engineering automation.
- Set the direction. Make goals, constraints, and task boundaries clear; let agents plan and act within them.
- Check the work. Use testing, evaluation, and human review to understand what worked and what still needs attention.
- Carry useful lessons forward. Turn experience into reusable memory and skills, then check whether they help on later tasks.
My AI teammates
Tech stack
- Applications · Python / FastAPI · TypeScript / React
- Infrastructure & data · Linux · Docker Compose · PostgreSQL · Redis · Qdrant
- Knowledge & versioning · Obsidian · Git
Happy to compare notes on agent harnesses, memory and skills, and AI-native applications.
