The demo is not “look at several repos.” It is one operating story: I identify repeated workflow pain, build AI-native tools, publish the proof surface, and write the learning back into a compounding system.
Demo promise
From scattered repos to a visible operating system.
repos / surfaces unified into one demo path
repeatable loop stages from pain to write-back
central narrative: AI-native systems from real workflow pain
Start from real work: code review friction, manager-facing communication, knowledge retrieval, or AI tool routing.
Package the answer as a repo, skill, MCP server, proxy, site page, or repeatable knowledge workflow.
Expose the artifact through docs, visuals, live pages, demos, and writing so a stranger can understand it in minutes.
Useful outputs become notes, code, pages, and reusable operating rules instead of disappearing into chat history.
Suggested 7-minute walkthrough
Each stop answers a different question an evaluator has: What is the thesis? What did you build? Can you reason deeply? Can you package workflows? Does it compound?
A file-first operating layer that turns my knowledge base into a working digital twin for research, writing, site updates, and continuous refinement.
Demo line: Shows the central thesis: durable files + agent execution + write-back loops create a compounding personal OS.
A trace-first standard and CLI for judging whether an AI agent is worth more token budget, permissions, and autonomy in my real workflows.
Demo line: Turns agent usage from vibe-based token burning into a repeatable scorecard for artifacts, verification, busywork, and invest-more decisions.
A proxy server that exposes OpenAI/Gemini/Claude/Codex-compatible APIs for CLI and coding tools, with OAuth login, provider routing, and multi-account load balancing.
Demo line: Turns fragmented CLI/OAuth/provider access into a routeable API surface for coding agents and workflows.
An 8-chapter interactive tutorial deconstructing Claude Code's internal architecture from its leaked source map. Bilingual (中文 / English), free and open-source.
Demo line: Converts source-code archaeology into a public, bilingual learning asset with diagrams and tutorials.
An anti-fluff workplace communication system packaged as both an MCP server and a reusable skill warehouse for manager-facing writing.
Demo line: Packages high-signal manager communication as reusable skills and an MCP-ready workflow.
A quiet runtime layer for querying my Obsidian knowledge base, recording run artifacts, and turning digital-twin experiments into repeatable workflows.
Demo line: Connects Obsidian knowledge, run artifacts, and improvement proposals into a repeatable local execution layer.
My public narrative system: a Next.js site, MDX writing archive, and AI-assisted publishing workflow designed to compound proof over time.
Demo line: Acts as the canonical surface where the work is sequenced, explained, and made externally inspectable.
Closing script
The strongest read of this portfolio is that each project moves the same system forward: clearer access to models, clearer use of personal knowledge, clearer workplace communication, clearer public proof. The repos are not random side quests; they are modules in an AI-native operating layer.
Best next viewer action
If someone only has 90 seconds, send them here first. If they want the deeper thesis, route them to Digital Twin. If they want code proof, route them to CLIProxyAPI or Claude Code Sourcemap.