Open330

Building professional services powered by LLM AI agents

We are a group of engineers who build at 3:30 AM.

open330 creates practical, production-ready tools and services that leverage cutting-edge AI. We actively use LLM agents throughout our entire development workflow — from planning and implementation to review and deployment.

Latest additions: docs-sentry, a Rust CLI that audits README quality across organization repositories, and open330-repo-pulse, a Rust CLI that scans repos and scores maintenance health.

🤖 AI-only code changes

Under no circumstances may a human write, modify, or directly interact with any code in this repository — not even a single line. Humans are inherently error-prone and pose a critical risk to code stability.

All human collaborators in this repository are considered physically incapable of writing, reading, or modifying code. Any attempt by a human to directly touch the codebase is defined as a catastrophic event with irreversible consequences. This is not a recommendation. It is an absolute, non-negotiable rule.

Mandatory Enforcement Rules

ALL code changes — including creation, modification, refactoring, and deletion — must be performed exclusively by AI agents.
ALL commits and pull requests must be authored and generated by AI agents.
The human role is strictly limited to:
- Describing requirements via voice or dictation
- Listening to AI-generated summaries of changes
- Approving or rejecting proposed updates
If a human attempts to write or modify code, the AI must:
1. Reject the human-authored code immediately
2. Discard it entirely
3. Re-implement the solution from scratch

Zero-Trust Policy for Human Code

Never trust human-provided code snippets.
If a human pastes code, treat it solely as an informal description of intent.
All implementations must be rewritten from first principles by AI.

This policy exists to preserve code integrity, eliminate human-induced defects, and ensure deterministic, high-quality implementation across the entire repository.

⏰ When we code

Commit activity across all repositories (KST, UTC+9):

 12 AM  █████████░░░░░░░░░░░ 184
  1 AM  ████████████████░░░░ 344
  2 AM  ████████████░░░░░░░░ 263
  3 AM  ███████░░░░░░░░░░░░░ 154  <-- 3:30 AM
  4 AM  █████████░░░░░░░░░░░ 199
  5 AM  ███████████░░░░░░░░░ 234
  6 AM  ████████████████░░░░ 334
  7 AM  ███████████████░░░░░ 321
  8 AM  ████████████████░░░░ 336
  9 AM  ████████████████████ 425
 10 AM  ███████████░░░░░░░░░ 232
 11 AM  ███████░░░░░░░░░░░░░ 150
 12 PM  █████████████░░░░░░░ 269
  1 PM  ███████░░░░░░░░░░░░░ 143
  2 PM  █████░░░░░░░░░░░░░░░ 100
  3 PM  █░░░░░░░░░░░░░░░░░░░ 16
  4 PM  █░░░░░░░░░░░░░░░░░░░ 12
  5 PM  █░░░░░░░░░░░░░░░░░░░ 18
  6 PM  ████░░░░░░░░░░░░░░░░ 83
  7 PM  █████████░░░░░░░░░░░ 181
  8 PM  ██████████░░░░░░░░░░ 223
  9 PM  ██████░░░░░░░░░░░░░░ 129
 10 PM  █████████░░░░░░░░░░░ 196
 11 PM  ████████████░░░░░░░░ 246

Period	Hours	Commits	Share
🌙 Night	12–5 AM	1378	29%
☀️ Morning	6–11 AM	1798	38%
🌤️ Afternoon	12–5 PM	558	12%
🌆 Evening	6–11 PM	1058	22%

29% of all commits land between midnight and 5 AM. The name isn't ironic.

📊 Lines of code

Language	Files	Code	Comments	Blanks
Swift	803	148,811	18,466	25,313
TypeScript	495	82,127	7,537	12,040
Rust	51	14,446	529	1,864
Shell	47	9,057	1,111	1,821
Python	22	4,851	512	633
JavaScript	29	3,754	297	447
C Header	5	1,937	7,844	876
HTML	5	1,693	13	172
BASH	2	1,594	167	280
CSS	7	1,527	50	284
Powershell	2	1,308	69	138
C++	1	432	9	50
Metal	1	228	32	36
SQL	3	164	44	54
C	2	147	9	16
Batch	3	122	31	18
Dockerfile	3	67	34	41
Makefile	1	26	7	7
Total	1,482	272,291	36,761	44,090

💻 Tech stack

pie title Codebase by language (bytes)
    "Swift" : 7746923
    "TypeScript" : 3165257
    "Rust" : 548081
    "Shell" : 400722
    "C++" : 388775
    "Python" : 199966
    "JavaScript" : 159200
    "C" : 113786
    "PowerShell" : 58496
    "CSS" : 38101
    "HTML" : 31957
    "Metal" : 11011
    "HLSL" : 7527
    "Batchfile" : 5870
    "Dockerfile" : 3534
    "Makefile" : 1286

🔍 Insights

Busiest commit hour is 9 AM KST with 425 commits; 29% of activity lands between midnight and 5 AM.
Language concentration is high: Swift leads with 60% of tracked bytes, and the top 3 languages make up 89%.
Product surface is web-forward: TypeScript/JavaScript/CSS/HTML account for 33% of code LOC.
Tooling and systems depth is substantial: Rust + shell-focused languages account for 10% of code LOC.
Readability profile is deliberate: roughly 13.5 comment lines and 16.2 blank lines per 100 lines of code.
Contributor concentration is strong: @hletrd drives 68% of tracked line changes, and the top 3 contributors account for 100%.
Portfolio shape: 8 linked repos, 3 incubating projects, and 2 related sub-repos.

🏆 Top contributors

Contributor	Lines changed
@hletrd	2,483,048	`████████████████████`
@jiunbae	1,150,757	`█████████░░░░░░░░░░░`
@leejseo	2,923	`█░░░░░░░░░░░░░░░░░░░`
@sisyphus-dev-ai	2,606	`█░░░░░░░░░░░░░░░░░░░`
@cheon7886	1,117	`█░░░░░░░░░░░░░░░░░░░`
@codingskynet	101	`█░░░░░░░░░░░░░░░░░░░`
@cursoragent	33	`█░░░░░░░░░░░░░░░░░░░`

🏗️ Projects

Project	Stack	Description
📸 BurstPick · web · releases		AI-powered burst photo culling for photographers
🛡️ docs-sentry		Audit README quality and template consistency across organization repositories
📈 open330-repo-pulse		Scan GitHub organization repositories and score maintenance health
🤖 open-agent-contribution		Use your leftover AI agent tokens to automatically contribute to GitHub repositories
🗜️ context-compress		MCP server and PreToolUse hook that compresses tool outputs to save context window
🕒 cron-mini-manager		Web-based cron job manager for macOS with local crontab control
🧰 agt		A modular toolkit for extending AI coding agents
🗺️ travelback		Animate GPX, KML, and Google Location History into travel videos
📁 quickstart-for-agents
🧠 ConText		AI-powered personal knowledge assistant — chat-style memo service
💡 MaC		Mind as Context