AI-native platforms and software · Korea

Five products. Four stages.

Brilliant Tiger builds AI-native platforms and software: a shipped desktop app, an open-source AI knowledge base, a platform in development and two more platforms that are still on paper.

Claude Pet releases in 12 weeks
30
open-source projects under the MIT license
2
of commits in our public repos co-authored with an AI agent
71%
MCP tools in Moshim's server
16

Public-repository figures as of Oct 11, 2026.

Products

Where each one stands

From a shipped desktop app to platforms that are still on paper. Here is what we are building and how far along each piece really is.

Claude Pet

Shipped

A desktop pet that watches your Claude usage.

A free companion that roams your screen and shows session, weekly and per-model usage in one compact pill, with reset timers and a startled look when your token burn spikes.

  • Server-reported usage through your existing Claude Code sign-in, with a local-log fallback.
  • Native on macOS (Developer ID signed, Apple-notarized) and Windows 10/11.
  • Five built-in pets, plus your own with a simple pet.json and spritesheet.

MIT · 30 releases since Jul 2026

Stack
  • Python
  • AppKit
  • Qt

llmwiki_json

Open source

A self-updating knowledge base for your AI.

A JSON-canonical knowledge base after Karpathy's LLM Wiki pattern: the model integrates what it learns once and keeps it current, instead of re-assembling raw sources on every question. What is relevant to a prompt is injected into the AI's context automatically.

Works withClaude CodeCodex

  • A UserPromptSubmit hook picks only the blocks relevant to each prompt and injects them into the AI's context.
  • Built-in 2D and 3D graph view, no Obsidian required.
  • Deterministic builds: the same input gives byte-identical output, checked by parity tooling.

MIT · started Aug 2026

Stack
  • Python
  • SQLite
  • Bun
  • TypeScript
Private preview

Moshim 모심

In development

Hire talent. Book hosts. Get found.

An AI platform that turns your material into a personal site, résumé, slide deck and short-form video, and lets hiring teams, producers and agencies search, cast, recruit and propose in one place.

For members

Personal site, résumé edits by chat, slide deck, vertical shorts (15–60 s)

For finders

Search, casting, hiring and proposals

Running today

  • An MCP server with 16 tools, 5 resources and 3 prompts, so members update their own careers by chatting with an AI assistant.
  • Korean-language search by role, region, career and open dates.
  • Every profile is a static page with JSON-LD, llms.txt and a sitemap, republished the moment it changes.

Kicked off Oct 7, 2026 · 3 pilot members · 69 tests

Stack
  • TypeScript
  • Node.js 22
  • MCP
  • zod

Private preview. Ask us for a walkthrough.

Email us
Private preview

K-Factory Hub working title

Planning

A manufacturing ecosystem where anyone can turn an idea into a product.

A B2B2C platform for Korea that helps aspiring founders and companies find the right factory. The service plan and a six-phase development plan are written. For now it is a plan only: development has not started.

Development plan

  1. P0Setup1 wk
  2. P1Public data + admin CRM9 wks
  3. P2Users + factory search7 wks
  4. P3RFQ + payments6–12 mo
  5. P4RFX + bidding1–2 yrs
  6. P5AI matching2+ yrs

Solid outline: the first three phases, about 17 weeks in the plan.

In the plan

  • Start from public data: Korea's registered-factory records, so buyers can search before a single factory signs up.
  • Sales-led onboarding with an admin CRM, and business verification against national tax records.
  • Korean by design: KS and KC certification, prime–subcontractor structures, government support matching.

Working title · service plan v2.0, Dec 2025

Planned stack
  • Java 21
  • Spring Boot
  • PostgreSQL
  • Next.js

Planning documents are internal. Ask us for an overview.

Email us
Private preview

Instructor Platform working title

Planning

A student-management platform built for instructors, not for academy owners.

A SaaS for Korean academy instructors to manage their students and parents with less effort: vocabulary tests, homework, lecture videos and parent reports in one place. The service plan, business plan and development plan are written. It comes right after Moshim.

Written so farService plan v1.0Business plan v16Development plan v1.5 + 16 appendices

In the plan

  • One place for everything an instructor does for a student: vocabulary tests, homework, lecture videos and parent reports.
  • AI drafts the first round of feedback, and the instructor reviews it before anything is sent.
  • Students join with the instructor's invite code, and parents can optionally join to follow progress.

Working title · plans written May 2026 · 36-week build plan

Planned stack
  • Java 21
  • Spring Boot
  • Next.js
  • Python

Planning documents are internal. Ask us for an overview.

Email us

AI

How we use AI

AI is part of how we build, and we design our products so AI agents can use them: MCP servers, hooks and files that AI can read.

AI in our git history

71%

of commits in our two public repositories are co-authored by an AI coding agent (189 of 265).

Commits per week Co-authored with AIOther commits
03060Week of Jul 6: 1 commits, 0 co-authored with AIJulWeek of Jul 13: 59 commits, 29 co-authored with AIWeek of Jul 20: 29 commits, 21 co-authored with AIWeek of Jul 27: 6 commits, 5 co-authored with AIWeek of Aug 3: 0 commits, 0 co-authored with AIAugWeek of Aug 10: 3 commits, 0 co-authored with AIWeek of Aug 17: 2 commits, 2 co-authored with AIWeek of Aug 24: 7 commits, 3 co-authored with AIWeek of Aug 31: 4 commits, 3 co-authored with AIWeek of Sep 7: 40 commits, 26 co-authored with AISepWeek of Sep 14: 58 commits, 48 co-authored with AIWeek of Sep 21: 4 commits, 3 co-authored with AIWeek of Sep 28: 38 commits, 35 co-authored with AIWeek of Oct 5: 14 commits, 14 co-authored with AIOct03060Week of Jul 6: 1 commits, 0 co-authored with AIJulWeek of Jul 13: 59 commits, 29 co-authored with AIWeek of Jul 20: 29 commits, 21 co-authored with AIWeek of Jul 27: 6 commits, 5 co-authored with AIWeek of Aug 3: 0 commits, 0 co-authored with AIAugWeek of Aug 10: 3 commits, 0 co-authored with AIWeek of Aug 17: 2 commits, 2 co-authored with AIWeek of Aug 24: 7 commits, 3 co-authored with AIWeek of Aug 31: 4 commits, 3 co-authored with AIWeek of Sep 7: 40 commits, 26 co-authored with AISepWeek of Sep 14: 58 commits, 48 co-authored with AIWeek of Sep 21: 4 commits, 3 co-authored with AIWeek of Sep 28: 38 commits, 35 co-authored with AIWeek of Oct 5: 14 commits, 14 co-authored with AIOct

Counted from the public git history of claude-pet and llmwiki_json as of Oct 11, 2026. Co-authorship is read from Co-Authored-By commit trailers. Both repositories ship agent instruction files (AGENTS.md).

Moshim speaks MCP

Members edit their own careers by chatting with an AI assistant, and hiring teams search and send inquiries from one too. Three permission levels, hashed tokens and rate-limited public inquiries keep it safe.

  • 16tools
  • 5resources
  • 3prompts

Add the Oct 11 marathon MC gig to my careers and publish.

add_careerpublish

Done. Your page, profile and search index are up to date.

Illustrative flow built from Moshim's MCP tools.

llmwiki_json feeds your AI the right context

Every prompt triggers a lookup. Only the blocks that match the question are injected, never whole pages, and an error never blocks a prompt.

  1. Prompt
  2. UserPromptSubmit hook
  3. SQLite index
  4. Relevant blocks
  5. AI context

Legible to AI by default

Moshim republishes llms.txt, a sitemap and JSON-LD every time a profile changes. Claude Pet and this site publish llms.txt too.

  • llms.txt
  • sitemap.xml
  • JSON-LD
  • MCP

Plans drafted with AI

K-Factory Hub's service plan and phased development plan were worked out in AI conversations.

Progress

What happened, in order

Dates come from release tags, git history and our plan documents.

  1. Moshim

    M0 up and running

    Kicked off on Oct 7. The MCP server, Korean search, member profiles and an inquiry API now run with three pilot members.

  2. Claude Pet

    v1.1.0

    The 30th release since the first public beta.

  3. Claude Pet

    v1.0.0

    Version 1.0 released for macOS and Windows.

  4. llmwiki_json

    Started, MIT-licensed

    A JSON-canonical, AI-maintained knowledge base with automatic context injection.

  5. Claude Pet

    First public beta

    v0.1-beta on GitHub Releases.

  6. Instructor Platform

    Service plan v1.0

    Plans for a student-management platform for instructors: service plan, business plan and a development plan with 16 appendices.

  7. K-Factory Hub

    Supplier-portal wireframes

    Screens for the supplier side of the platform.

  8. K-Factory Hub

    Development plan, P0 to P5

    Six phases, from project setup to AI matching.

  9. Company

    Brilliant Tiger founded

    Registered as a business in Korea.

  10. K-Factory Hub

    Service plan v2.0

    Vision, strategy and a phased roadmap for a Korean manufacturing platform.

About

Building in the open where we can

Brilliant Tiger (빛나는 범) is a registered business in Korea, started in January 2026. We build AI-native platforms and software, and publish what we can as open source. Every number on this page comes from our public repositories or our own project documents.

Company
Brilliant Tiger (빛나는 범)
Founded
January 2, 2026
Based in
Korea
Focus
AI-native platforms and software
Products
Claude Pet · llmwiki_json · Moshim · K-Factory Hub · Instructor Platform

Contact

Let's talk.

Pilot a product, partner with us, or compare notes on building with AI.

Email us Follow us on GitHub (opens in a new tab)

contact@brillianttigercorp.com