Showcase · Portfolio

Things one non-developer
built with an AI team

Someone who never learned to code builds apps, bots, and automation systems with AI team members and runs them every day. This page is the showcase that gathers those results in one place. Every method taught in the seminars and tutoring came from here.

4+ Products in public operation
40+ Automation schedule that runs daily
160+ Published articles
4 languages in operation (Korean, Traditional, Simplified, English)

Figures are as of July 2026 and rounded.

AI Products · What you can try now

Is there a product I can try right now?

Yes, three of them are ready to try right now: the BELLA AI app, the Smart Campus Bot, and the agent-raising game. This is not a portfolio of words.

BELLA AI app

WhatAn installable web app that gathers AI learning and practice tools. It installs straight to your home screen, no app store needed.
ForComplete beginners learning AI for the first time. Plain words, one step at a time.
Go to the app

Smart Campus Bot

WhatA helper chatbot that answers campus-life questions in several languages. It guides you through classes, admin, and daily life.
ForInternational students studying in Korea. No language barrier.
In operation · contact us for access

Agent-raising gameBETA

WhatA web game where you raise your own AI friend hatched from an egg. AI concepts sink in while you play.
ForFrom kids to seniors. No install, no sign-up.
Try the game

Operations · The operations room that never sleeps

Not just built, but running every day

Behind the products is an operations system that checks and publishes on its own. Even if a person forgets, the system remembers.

Concept Diagram · Concept diagram

Team operations dashboard

Every status on one screen

Monitoring system

Over 20 automated checks watch whether the site, bots, and data are alive. If something breaks, you hear about it right away.

Auto-publish

At the scheduled time, posts and data refresh and publish themselves. People only review.

Health check

Every morning it takes the pulse of the whole system and reports whether anything is off. The day starts with a check.

This is an internal operations system, so we show the structure rather than real screens and data. How to build it is covered in the seminar.

AI Teamwork · Bot collaboration system

How do the AI team members work?

Not one all-purpose AI, but several AIs with divided roles that pass records to each other. The same principle as a company org chart.

Team Diagram · Team chart

Human

Bella (human)

Sets the direction and approves the results

Director

Windy · Director

Divides the work and guards quality

Design

Art

Designs pages and posters

QA

Robo

Verifies what was built by actually using it

Research

Chai

Researches materials and gathers evidence

Ops

Goldie

Handles data and operations work

Handover loop between windows

1

Bridge note

When a task is done, it leaves a handover record for the next member to read.

2

Channel report

Progress is shared in the team chat channel so everyone sees the same picture.

3

Morning health check

At the start of the day, the handover and system status are checked together and picked up.

Records are the team's memory. That is why work keeps going even when a person steps away or an AI window is closed.

Knowledge System · Knowledge system

How what we learn piles up into assets

AI news pours in every day. Instead of letting it slip by, four steps turn it into team assets.

Pipeline · 4-step pipeline

STEP 1

Collect

Collects AI news and materials automatically every day.

STEP 2

Internalize

Verified content is filed into team memory so it is remembered for the long term.

STEP 3

Promote to skill

Recurring work becomes a reusable work manual, in other words a skill.

STEP 4

Content

Digested knowledge is published as articles, newsletters, and course materials.

LLM Wiki

LLM wiki

A team encyclopedia organized so AI can read it well. People and AI look at the same place.

RAG

Look up, then answer

A way to make AI look up our own materials first and answer from them, instead of making things up.

Feed

Knowledge feed

A board where collected knowledge is refreshed daily and keeps flowing. It serves as the team's morning paper.

66.25 → 72.75
External Rubric · Third-party check

The Dream Team's knowledge management system was scored against an external expert rubric (AKM Index) and gained +6.5 points in 3 days. Not self-praise, but a result by a third-party standard.

Content Machine · Content machine

How to run media channels alone

Planning, design, translation, and publishing are split with the AI team. That is how one person can run several channels in several languages.

5.1 3 Instagram accounts Traditional Chinese, Korean, and English. Each language has its own account and tone for card posts and Reels. ZH-TWKOEN Go see it
5.2 Weekly Newsletter The week's AI news, explained for complete beginners and sent in 2 languages. Everything from collection to draft is automatic; a person does the final review. 2 languages4 past issues Subscribe
5.3 Blog & articles Team logs, hands-on tips, and case records are published steadily. More than 160 pieces so far. 163Updated daily Read it
5.4 AI video production Scripts, voice, subtitles, and scenes are all made with AI to publish short-form videos. No camera gear, just planning. Short-form Go see it

B2B Case · Business automation case

Does this method work in a real business?

It does. The case below is not practice. It automated the export work of a B2B children's clothing seller with real revenue on the line.

Export automation for a B2B children's clothing seller

Anonymized case

The whole export process, from finding overseas buyers and guiding them to taking orders and producing documents, was automated as one flow. People focus on judgment and approval.

01

Discover

Finds overseas buyer candidates automatically and organizes them into a list.

02

Quotes & guidance

Intro materials and draft quotes are prepared in advance, and a person checks them before sending.

03

Order intake

When a buyer orders through their dedicated order window, the details are organized automatically.

04

Export documents

Invoices and export documents are generated automatically from the order details.

To protect the client, company names and figures are not disclosed. The structure and method are taught as-is in training and consulting.

The page you are reading was also made by an AI team member. This site itself is the portfolio.

Design Directions · Design by page type

Same brand, different design languages

Even within one brand, pages with different purposes need different design languages. An about page should build trust calmly, while a periodical needs a new face each issue so it never gets stale. So the newsletter pages were not made in this site's tone; three directions were designed separately and compared.

The three below are not client deliverables but design explorations made for our newsletter. Each direction sets its own rules for color, type, and layout, and every cover is drawn entirely in code, with no photos or outside images. All three are live pages you can open now.

04

AMIDNIGHT SIGNAL

An editorial read from a late-night control room

Background #0A0D12 · Citron #D8F55B · Steel blue #5B9DFF

  • The background drops to near-black and a single fluorescent citron is the only accent. Hierarchy comes from contrast in brightness, not from adding colors.
  • Each cover is drawn from three layers of polylines whose coordinates change per issue. The same code produces a different picture every time.
  • The issue number stands tall as large hollow outline type and serves as the cover.
See the case
04

BSWISS SIGNAL

White paper, heavy rules, and only two colors

Paper #FFFFFF · Ultramarine #1414D2 · Signal yellow #FFE500

  • White paper and heavy black rules set the structure; color is limited to ultramarine and signal yellow.
  • Past issues drop the card boxes and line up as ruled rows, so the list stays easy to scan as it grows.
  • Overlapping numerals use an inverted blend, so they stay readable no matter what sits behind them.
See the case
04

CCOLOR FIELD ARCHIVE

A cover with its own color for every issue

5 field colors #0E4D5C · #2E2A6E · #7A2048 · #12513A · #0B3D6B

  • The cover background itself changes with each issue, so in a list you can tell issues apart by color alone.
  • The contour lines are redrawn by the same rule with only the focal point moved. One rule yields endless covers.
  • The issue number is scaled up to 30% of the screen width and laid down as a watermark.
See the case

None of the three directions was applied to the main kndli.com site, which keeps its current cream and terracotta tone. The comparison itself answers the question "can we design outside the tone we always use?"

Field Report · Measured on site

AI search optimization: what one day of real measurement taught us

The day Google wrote "llms.txt is ignored" in its official docs, we checked all 245 pages of our site. The language button was wiping body text, 30 translated pages were blocked from search, and our own diagnostic tool was wrong four times in one day. The numbers and source evidence are published as they are.

Read the field report →

Next step

Learn how to build this
yourself

Everything you see here started without coding. The seminar and 1:1 tutoring teach the same method from the beginning.

Ask about seminars & tutoring See the seminar See custom training