Act III · Method

Method

Not just results, but methods worth reusing
Starting from my own practice, keeping the process and the judgment

AI workflowDeep among the lotus

Every work here I made on my own with AI
AI is the brush; the judgment and the choices are mine

  1. Root

    Pick the question

    What I do
    Find a question worth working on, and who it is for
    What AI does
    Gathers material and lists similar work that already exists
    How I check
    Is the need real, and has someone already solved it
  2. Leaf

    Learn first

    What I do
    Find great examples and take apart why they work
    What AI does
    Reverse-engineers structure, looks up open source code, sums up the approach
    How I check
    Open the examples myself instead of trusting a summary
  3. Stem

    Set the rules

    What I do
    Write the project rules first: goal, structure, limits, what done means
    What AI does
    Works within the rules and stops to ask at any red line
    How I check
    Are privacy and publishing covered by the red lines
  4. Bud

    Prototype

    What I do
    Build a first version with real content and point out what is wrong
    What AI does
    Writes the code, prepares assets, gets it running
    How I check
    Walk through it myself on phone and desktop
  5. Bloom

    Polish

    What I do
    Pick at every detail and decide what stays
    What AI does
    Screenshots its own work, fixes issues, tests screen by screen
    How I check
    Passing the checks is not the same as done
  6. Seed

    Ship and review

    What I do
    Decide when to go public and state where things stand
    What AI does
    Prepares the launch checklist and the test record
    How I check
    Write the review into my notes so the next project starts ahead

Case · This website

In
One idea, a few reference sites, an AI illustration I directed, a song I sang
How
Rules and a design spec first, then eight rounds, each checked with screenshots on phone and desktop
Out
A bilingual static site with no third-party libraries, from phones to 4K screens

I

From sources to usable information

  1. Define the question
  2. Gather sources
  3. Sort and verify
  4. Write it readably

Care about where information comes from, where it stops and when it changes, not just a generated answer
China AI Model Radar came out of this path

II

From references to digital experiences

  1. Take references apart
  2. Prototype with real content
  3. Design phone and desktop separately
  4. Walk the whole journey

First understand why others made their choices, then redesign around my own content
This portfolio was made the same way

A full AI office workflow case is being written up
with input material, steps and sample output to come

NextAct IVPingchengAnd a city I am still telling