Four active projects, presented through their actual architecture, constraints and validation records. No invented impact numbers and no placeholder showcase media.
The method behind the workHow I think and build息壤 / LOCAL-FIRST AGENT
A personal agent that can be proactive without taking control away from its owner.
A useful personal agent needs context and initiative, yet the same capabilities can make it opaque, invasive or unsafe.
智能体脉搏 / DEVELOPER TOOL
A privacy-first observability layer for coding agents working across devices.
Parallel coding agents create an attention problem, but conventional telemetry would expose exactly the private material the tool is meant to protect.
知芽 / AI LEARNING PRODUCT
A full learning loop where planning, focus, evidence, review and reward reinforce one another.
Learning apps often optimize either planning or motivation; the difficult part is preserving evidence, feedback and memory across the whole loop.
Personal Lab / HUMAN-AI PORTFOLIO
A portfolio that behaves like a small, inspectable product.
A portfolio has to communicate taste and proof without becoming a fragile performance or a collection of empty promises.