lineage-viewer
AlphaA zero-runtime-dependency Web Component for stable, embeddable table and column lineage diagrams.
Deterministic layout · Web Components · diagnostics
DATA INFRASTRUCTURE · TRUSTWORTHY AI · INDEPENDENT BUILDING
Make complex systems smaller. Put AI inside verifiable boundaries.
I build data infrastructure and trustworthy AI tools in Hangzhou, writing about real constraints, trade-offs, and reusable implementations.
Building in public, one verifiable artifact at a time.
3published notes
2active projects
Working principles
Capability matters. Boundaries make it trustworthy.
- 01
Own the data
Keep usage, cost, and operational records portable, inspectable, and under the user's control.
- 02
Make behavior deterministic
Use AI to interpret ambiguity; leave validation, permissions, and state changes to explicit systems.
- 03
Build the smallest useful boundary
Prefer focused components, boring dependencies, and interfaces that remain easy to replace.
Now building
Ideas tested as working systems
Two ongoing projects explore the same question from different angles: how do we add capability without giving up clarity or control?
LarkLedger
In progressA ledger where AI interprets natural input while deterministic services validate and commit each action.
Structured actions · confirmation boundaries · idempotency
Start here
Start here
A compact map of the ideas and implementation choices that shape this blog.
Why I Write ‘见远而行’
Updated:The meaning behind this site, its three long-term themes, and the standards I use to document constraints, trade-offs, and verifiable results.
Building a Data Lineage Viewer as a Web Component
Design trade-offs behind lineage-viewer: zero runtime dependencies, deterministic layout, and an embeddable component boundary.
Latest notes
New observations from systems that are still changing.
In a Ledger, AI Should Interpret Rather Than Decide
How LarkLedger separates natural-language understanding from deterministic ledger logic and adds confirmation boundaries to risky inputs.
No algorithmic feed. No inbox pressure.
Follow new essays with RSS, or find shorter progress notes on GitHub and X.