ATELIER · 01 · /now

For three years,I rebuilt game content as an industry with AI.Next: AI-native games — co-built.

Welded AI into the level / hero / numeric / data / art / tooling layers of a live SLG. Now looking for the team that wants to point this pipeline at a new shape of game.

Shanghai · UTC+8Open to: AI-native game teams.
superman.shipping.diff
↓ scroll for receipts

/ 02 · receipts

The numbers are mine.

Every figure traces back to a live build artifact — no borrowed credit, no narrative inflation.

60+
Heroes covered

Full content + recommendation pipeline for every hero in the live game.

01
100k+
Config rows handled

Across 7 interlocking tables, all flowing through reusable AI workflows.

02
100+
Agent Skills shipped

Six business domains: levels, heroes, numerics, data, art, tooling.

03
122
Auto Exports

End-to-end MD → CSV → multi-language → validated, no human re-typing.

04
118
Analysis reports

Gacha, pool positioning, system teardowns, monetization decisions.

05
~1/3
Project revenue supported

Supergirl SP content I owned end-to-end is the project's single largest revenue contributor.

06

/ 03 · experience

Title, scope, and the P&L line I sit on.

  • 01

    Joined FunPlus in 2023 as a campus-hire level designer; grew into Level Lead. Main campaign, combat experience, combat AI and version drops all ship live — player-facing content with every release.

  • 02

    Walked the three generations of hero design end-to-end myself: Batman (fully manual) → Riddler (AI-assisted, semi-automated) → Supergirl SP (AI-native pipeline). Same person, one pipeline, pushed from zero to production.

  • 03

    Supergirl SP is the flagship hero supporting ~1/3 of total project revenue — and the first proof point that the AI-native pipeline actually ships.

  • 04

    SP Card System Owner — the system itself is a core revenue module. I own teardown, plan, ship, iterate end-to-end and sit directly on its P&L.

  • 05

    Took over a stalled SLG content line; currently rebuilding it with the same AI pipeline playbook to bring it back to stable delivery.

  • 06

    Horizontally: built the AI production line covering levels, heroes, numerics, data, art and tooling — 60+ heroes, 100,000+ config rows, 100+ Agent Skills. Personal capability turned into team infrastructure.

/ 04 · pipelines

AI is a production line, not a prompt.

Six business domains, all flowing through reusable workflows. Each pipeline is something I personally built, dogfooded, and handed off.

pipeline_01

Level Production Line

Spec → level config → monster config → reward config → numeric tuning → ship

7 interlocking tables, 10,000+ rows synchronised in one session.
pipeline_02

Hero End-to-End Pipeline

Brainstorm → design doc → engineer spec → config → CSV export

60+ heroes through the same reproducible flow.
pipeline_03

Hero Guide Auto-System

Recommendation copy + lineup + stats + signature gear + environment + multi-language

Batch output for the full roster — what used to be a 2-week version becomes a 1-week sprint.
pipeline_04

Decision-Grade Data Reports

Raw data → structured analysis → version & monetization decision support

118 reports backing real shipping decisions.
pipeline_05

100+ Agent Skills Library

Reusable skill packs across six business domains

Personal capability turned into team productivity. Skills handed off and used by colleagues.
pipeline_06

Code-as-Spec Decoder

When the doc is missing, AI reads the code and reconstructs the design intent

Unblocked a handover where the original designer had left and no doc existed.

/ 05 · atelier

Things I actually shipped.

Each card is work already running in production with data behind it. Full case-study docs available — just ask.

Hero design · one person, three eras

Three hero generations · manual → AI-native

Batman (fully manual) → Riddler (AI-assisted, semi-automated) → Supergirl SP (AI-native pipeline). Same person, one pipeline, pushed from zero to production.

// metric · 1 person · 3 generations · 1 reusable pipeline.
01

Hero · Monetization core

Supergirl SP content line

Owned design → combat mechanics → skill feel → config → version delivery. The flagship hero where the AI-native pipeline first truly shipped.

// metric · ~1/3 of total project revenue supported.
02

System Owner · Revenue module

SP Card System

A core revenue module of the project. I own teardown, plan, ship, iterate end-to-end and sit directly on its P&L.

// metric · Stable monthly delivery · sits on P&L.
03

AI Pipeline

Hero Recommendation Stack

Structured guides for the full hero roster: text → MD → CSV → multi-language → validation.

// metric · 60+ heroes, single-week turnaround.
04

System Owner · Earlier

Tower System (1700 levels)

Levels, monsters, rewards, numeric balancing, lore, art briefs, milestones, BI requirements — single-owner.

// metric · 1700 levels shipped.
05

System Owner · Earlier

Tag Tower System

2 small towers × 120 levels + 5 big towers × 600 levels, with tutorial, ranking, lore, UE design.

// metric · 240 + 3000 levels.
06

Throughput

Multilingual Sprint

1000 lines of multilingual content shipped in a single day — the old cycle was a 2-week version, the new one is one day.

// metric · 1000 lines / 1 day.
07

/ 06 · about

I'm a player, rebuilding the games I design with AI.

I played SLGs to the point of dissecting their data tables, and then walked into the industry. That's where my taste for what's good comes from — I judge from the player's side of the screen first.

I use AI because I'm allergic to doing the same thing three times. I tear the process apart, let AI handle what's repeatable, and keep the judgment calls for myself.

Next stop: AI-native games — not AI as a sticker on a product page, but games whose mechanics grow out of AI from the bottom up.

/ SLG player credentials

  • Shuaitu Wuji
  • S3 Sango Zhejiang Region · MengMa
  • Bameng Invitational · Champion (Returning)
  • Sanqi Sanmou · Xuantian Si

Top-tier player. That's where my ability to read the math and still feel the pull comes from.

/ 07 · let's talk

If you build AI-native gamesor 3D AI products,we should talk.

Especially if your team thinks of AI as a production line, not as a costume on a product page.

/ channels

Shanghai · UTC+8