Technology

Structure before code, delivery is the goal — from Web to desktop to embedded.

Jingxin

A minimal meditation guide app — breathing, mindfulness, guidance, and relaxation.

FengOffice

Mail, CRM, and Newsletter in one connected pipeline.

Search King

An unbeatable human-machine search tool — images and text in one grab.

Question First, Build Later

No ambiguous requirements. Think it through before building.

System as Compound Interest

PDCA drives progress, layer by layer.

One Dimension is Bias

The real world is a web, not a spreadsheet.

Tools Follow Problems

Open-source first, AI-accelerated, choose what fits.

Delivery is the Start

Code is just a byproduct; delivery is the start.

Investment Framework →

Tools need a business context to create value.

Explore

One App, Four Ways to Breathe

Jingxin — a minimal meditation guide app, breathing, mindfulness, guidance, and relaxation in four modes.

Proof of a Cross-Platform Mobile Build

Built with Flutter — one codebase that runs on Web / Android / iOS / Windows. A minimal interface that stays focused on meditation itself, undisturbed by feature clutter.

  • Data model: a MeditationSession records duration / mode / completion state.
  • State management: Provider manages the chosen duration and mode, with crisp UI responses.
  • Breathing animation: a custom BreathingCircle component guides the breathing rhythm.
  • Four modes: breathing / mindfulness / guidance / relaxation.

Mail, CRM, and Newsletter in One Chain

FengOffice — AI-unified mail and CRM, folding the infrastructure of content operations into one automated pipeline.

Where Engineering Delivery Meets Marketing Operations

One CLI manages mail, one Docker stack runs Twenty CRM — mail discovery → reply → CRM record, the whole chain wired up. Newsletters are distributed automatically by the Listmonk system.

  • fengmail.py: a CLI mail client, IMAP receive + SMTP reply.
  • Twenty CRM: one-command docker-compose deploy, port 3002.
  • Newsletter: a Listmonk + Postgres subscription system, with AI scripts to add, unsubscribe, and send campaigns.
  • Four-tier mail classification: docs/email-classification.md defines the classification system.

Eighteen Engines, the Best One Wins

Super-Invincible-Search-King — a Playwright-powered all-in-one search tool: images, text, web reading, deep crawling, and anti-bot site scraping.

Treating "Being Blocked" as the Norm

One machine, eighteen engines. CAPTCHAs are skipped automatically, the engine chain falls back by itself, and a real-human browse mode is the final backstop.

  • Five behaviors: image / text / read web / deep crawl / platform fetch (Xueqiu, Bilibili).
  • Twelve image engines + six text engines, covering mainstream and regional search.
  • Six backends: from headless browser to real browser, layered fallbacks.
  • Traceable copyright: each result carries its source and license record.

Question First, Build Later

No ambiguous requirements. Break it down until it's measurable and verifiable — that's clarity.

How I Usually Approach Software Projects

  • Define the real problem first: What are we solving, who will use it, and what does "done" really mean?
  • Choose the right stack: Prioritize deployment and maintenance over tech for tech's sake.
  • Build in layers: Get the structure and flow right first, then refine visuals and experience.
  • Deliver with launch in mind: Documentation, naming, navigation, meta information — all part of shipping.

Advancing with PDCA: plan first, then execute in stages, check, and adjust — never jump into writing messy code. Decouple only after the structure is clear. Don't cram everything together.

System as Compound Interest

PDCA drives progress. Break down, research, plan, execute, review — layer by layer.

Advancing with PDCA

  • Plan first: Define the problem, break down tasks, research solutions, then start building.
  • Then execute: Move forward in stages, each step verifiable.
  • Check and adjust: Review after completing each cycle, then enter the next loop.

Research the Structure First, Then Start Building

Decouple only after the structure is clear. Don't cram everything together. Reuse everything reusable — extractable modules, styles, components, and workflows are all made reusable when possible.

  • Reuse everything reusable: Extractable modules, styles, components, and workflows are all made reusable when possible.
  • Value adaptability and aesthetics: Landscape and portrait, responsive layout, rounded corners, shadows, modern interface feel — all considered together.

One Dimension is Bias

The real world isn't a 2D table. Data is a web of relationships, not rows and columns.

I Don't Treat Software as Isolated Code

I consider user needs, deployment paths, maintenance costs, and how tools fit into real workflows — all together. Product plus operations perspective, debugging and iteration discipline, documentation and handoff awareness — these matter more than the code itself.

Data & Integration Work

I often connect APIs, spreadsheets, local files, and prompt systems, turning scattered manual steps into running workflows.

  • JSON / CSV / Excel processing
  • HTTP API & glue code integration
  • Prompt-assisted content & workflow systems
  • web-app-template — full-stack TypeScript project template
  • Acquaintance Relation Table — personal CRM system (full-featured pure frontend)
  • Financial reports & market data automated scraping and analysis

Tools Follow Problems

Open-source first, AI-accelerated. Choose what fits best, not what's flashiest.

Research open-source projects first: If a suitable open-source solution exists, I study how others have done it before deciding. When no open-source project exists, research AI solutions: But AI is an accelerator, not a replacement for structural thinking.

Code isn't the only focus: Whether the functionality is correct, stable, and usable matters more than simply piling up code.

  • Web & Growth: HTML/CSS/JS delivery, GitHub/Cloudflare Pages, SEO & multilingual.
  • Apps & Automation: Flutter, Python, AutoApply (Selenium→Playwright), UnifiedSearch, cc-connect-docker.
  • Data Workflows: API integration, financial report automation, prompt pipeline systems.
Flutter Python JavaScript PowerShell Automation Prompt Engineering

Delivery is the Start

Code is just a byproduct. Documentation, deployment, maintenance, and iteration — that's where delivery begins.

If you want to see specific repositories, code walkthroughs, architecture explanations, or a recruiting-focused technical screening version, I can showcase the most relevant code by scenario: growth tools, internal software, automation tools, or multi-platform builds.

Technical Implementation

Think it, build it

Cross-platform & Mobile Development
Flutter(Android / Windows / Web) Native Android(Kotlin) Responsive Layout Multi-language Internationalization Full Build & Release Pipeline
Desktop Development & Automation
Python(Tkinter) VBA Office Automation PowerShell Advanced AutoHotkey Batch Scripting
Web & Site Deployment
HTML / CSS / JavaScript Independent Website Building GitHub Pages / Cloudflare Pages Domain & DNS Chrome Extension Development SEO Basic Optimization
AI Engineering
Harness Engineering(Multi-model Collaboration) AI Agent Development Docker Containerization Deployment Prompt Engineering Compute Management & Low-temperature Deployment
Tools & Engineering Literacy
Git Version Control Command Line(PowerShell / Bash) Markdown Multi-system Proficiency(Win / Linux / macOS / Android / iOS) SSH Remote Management Network Diagnostics & Packet Capture
Design & Multimedia
Photography(Portrait / Landscape / Documentary) UI Aesthetics & Interaction Design Video Editing & Production Canva / PS Basics

Web3 Deep Research

From technology fundamentals to business scenarios — a systematic study of blockchain, stablecoins, RWA, and decentralized finance

Read Full Research →

AI Engineering & Automation

Harness Engineering, AI Agent development, Prompt Engineering, and multi-model collaboration — AI is not just a tool, it's a new way of working

View AI Topic →

Cloud Services Architecture & Practice

From Docker containerization to Cloudflare deployment — building scalable, high-availability cloud infrastructure

View Cloud Topic →
Photo with Jensen Huang

With Jensen Huang at HKUST — I'm behind him.

He stood before me, yet I felt we were no different — just two people on different pages of the same story. He may be the most important person in the world, but his time bank has fewer deposits than mine.