anzhe@portfolio:~/dev $← editorial
●00_home.md
●01_work.json
●02_stack.toml
●03_about.md
●04_book.sh
 ▲ ANZHE.XYZ
 · ai agent engineer

// boot sequence...

[ok] loading ~/.config/anzhe/identity.toml

[ok] mounting anthropic-agents-sdk

[ok] mcp: gmail calendar linear drive

[ok] langfuse · 5 judges loaded

[warn] coffee level critical, refilling…

[ok] ready.

$ whoami --verbose

name     : Anzhe (Barton) Dong
role     : Founding AI Engineer · TripFix
based    : Toronto, ON · Canadian PR
focus    : agentic systems · evals · MCP
contact  : barton97daz@gmail.com
domain   : anzhe.xyz

$ cat ./work/tripfix.md

# An AI co-pilot for flight-delay refund claims.

Reads boarding passes and airline emails, drafts the rebuttal letter, escalates only when uncertain. Built as a small team of specialised agents — not one monolithic prompt — so each piece is testable, swappable, and auditable on its own.

agents
4-agent system
llms routed
14+
eval judges
5 dims
headcount
1 (me)

read full case study →

$ ls -la ./work/

total 4  ·  drwxr-xr-x  anzhe

$ cat ./stack.toml

[workshop]
purpose = "ship like a team of five from one terminal"
pieces  = 8
surface = "claude code · cursor · neovim"
  • 01[Setup]
    Start every machine with CLAUDE.md
  • 02[Workflow]
    Plan mode before code mode — always
  • 03[Agent]
    Specialised subagents on every task
  • 04[Hook]
    PreToolUse hooks for hard-stop guardrails
  • 05[Skill]
    Skills as versioned markdown, not system prompt
  • 06[Memory]
    Auto-memory across conversations
  • 07[MCP]
    MCP connectors as one surface for life admin
  • 08[Workflow]
    Cursor + Claude Code, side by side

workshop --details →

$ contact --all

emailbarton97daz@gmail.com

linkedin/in/anzhe-dong

x@therealbossdong

github@donganzh

book./book.sh --interactive

$

$cd ../editorial$open ./work$open ./stack$open ./about$./book.sh
NORMALmainanzhe.xyz