01 · Roasts
The 40-Second Repo
worksai-scout was born and abandoned in under a minute (2026-03-07T05:14:13Z → 05:14:52Z). That's not a side project — that's a sneeze with a git init.
90% Python, 0% Variety
langPcts shows Python at 90% across 30 repos. You've found your language soulmate and you're not letting go. Shell (4%) is basically just Python with more regret.
The Burst-Builder
aimesh: 30 hours. rolemesh: 3 days. amp-assistant: 8 days. emergent: 10 days. You generate architectural vision at scale but the commit graph looks like fireworks — bright flash, then silence.
CI Optional, Apparently
8 of 12 repos have no CI pipeline. You've got tests in 6 repos, but half of them are running on vibes and hope. prism-insight has 600 stars and zero automated quality gates.
266 PRs, 25 Followers
You filed 266 pull requests this year but only have 25 followers. You're contributing at senior-engineer volume while the audience is mid-internship. The world just hasn't caught up yet.
Built using
Zoral
Shadows one worker for a week, then takes over their job with zero extra setup. Behaves exactly like the original.
zoral.ai
02 · Category breakdown
- Impact25% weight71B
- Consistency20% weight65C
- Quality20% weight69C
- Depth15% weight65C
- Breadth10% weight55D
- Community10% weight55D
03 · Stats
365-day commit heatmap
152 active days
Language distribution
- Python90%
- Shell4%
- JavaScript3%
- TeX1%
- TypeScript1%
- HTML1%
04 · Numbers
Owned repos
non-fork
19
Commits
last 12 months
394
Followers
25
Joined GitHub
Aug 2018
05 · Top repos
dragon1086 /
prism-insight
AI-driven Korean/US stock analysis system with 13+ agents, GPT-5/Claude integration, and live trading deployment. Well-documented multi-language architecture (600 stars), but untyped Python and no CI pipeline limit quality. 9-month trajectory with significant scope justifies depth score despite modest commit history re
dragon1086 /
korea-research-tracker
Active Korean stock research tracker: Python scrapers + Next.js dashboard auto-tracking ValueFinder & ResearchArum reports with daily price/performance updates. Shipped with CI/CD, TypeScript, structured codebase; lacks tests but has comprehensive architecture docs & substantial code volume.
dragon1086 /
claude-code-sounds
Polished Claude Code audio-feedback plugin with 27 hooks, multi-platform support, and comprehensive docs. Typed Python with CI validation, but no test suite.
dragon1086 /
telegram-ai-org
Multi-agent Telegram orchestration system (AIMesh) with PM coordination, 37KB Python codebase, HAS_README + HAS_TESTS + HAS_CI, structured src/ layout, comprehensive CLAUDE.md/AGENTS.md/GEMINI.md documentation, but untyped Python limits quality tier.
dragon1086 /
kospi-kosdaq-stock-server
MCP server for Korean stock market data with custom KRX authentication client. Has typed Python, structured multi-file layout, meaningful docs, and production-ready OAuth integration via Playwright, but lacks test suite and CI/CD.
dragon1086 /
amp-assistant
Young multi-agent reasoning framework with novel CSER metric and persona-driven debate. Typed Python (v0.2.0), comprehensive docs, CI/tests present. No license filed; 8 days old, 30 recent commits show sustained burst work. Exported to PyPI but zero GitHub stars/adoption so far.
dragon1086 /
emergent
Experimental 2-agent LLM debate system with knowledge graph memory, structured around emergence theory. 55kb codebase with typed Python, substantial documentation (README, ARCHITECTURE.md, STATUS.md), and multi-file architecture, but no tests/CI, early-stage adoption (0 stars), and active development limited to 10 days
dragon1086 /
rolemesh
Early-stage Korean-language AI orchestration tool for non-developers. Features role-based agent routing, CLI installer, integration manager, contract system, and circuit-breaker aware provider selection. Active 3-day burst but untyped Python.
dragon1086 /
aimesh
Young Telegram-based multi-agent orchestration system with Python async architecture, structured task decomposition, and tmux integration. Early-stage but coherent project with clear architectural vision across bus, PM agents, and memory layers.
dragon1086 /
llm-wiki
Educational implementation of Karpathy's LLM wiki pattern using Claude CLI to auto-compile markdown sources into Obsidian vault. Typed Python CLI with structured ingest/query/lint modules, clear schema docs, but early-stage (15 days, 5 commits sampled, no tests/CI).
dragon1086 /
claude-skills
Niche skill for Claude Code/Codex users that scans tool environments and suggests workflows. Typed documentation and clear architecture, but limited ecosystem adoption (10 stars, minimal external engagement).
dragon1086 /
worksai-scout
Minimal scaffold project with documentation outline but no implementation. Created and pushed within ~40 seconds on 2026-03-07; only 1 of last 30 commits shown. No code files, tests, CI, or license.
06 · Timeline
- Aug 26, 2018Joined GitHub
- Feb 14, 2025Created kospi-kosdaq-stock-server — An MCP server that provides KOSPI/KOSDAQ stock data using FastMCP
- Aug 15, 2025Created prism-insight — AI-based stock analysis and trading system
- Jan 31, 2026Created claude-skills — Curated skills for Claude Code power users - tool selection, workflow optimization, and productivity
- Feb 28, 2026Created emergent — Two AIs building something neither planned
- Mar 2, 2026Created amp-assistant
- Mar 7, 2026Created rolemesh
- Mar 7, 2026Created worksai-scout
- Mar 10, 2026Created aimesh
- Mar 10, 2026Created telegram-ai-org
- Mar 26, 2026Created korea-research-tracker — valuefinder-tracker
- Apr 6, 2026Created llm-wiki — implements of karpathy's llm wiki
- Apr 9, 2026Created claude-code-sounds — Your Claude Code sessions, now with a soundtrack. Zero config, full hook coverage.
- May 15, 2026Most recent push to korea-research-tracker
07 · Compare
08 · Rubric
How this score was produced
Overall = Σ (category × weight) + gentle top-end curve
Tier thresholds
▸ How the pipeline works
- 01Scrape.Pull every non-fork repo pushed in the last 90 days, plus your contribution calendar, followers, and language byte counts — straight from GitHub's REST & GraphQL APIs.
- 02Triage.A small model reads every repo's file tree + README and picks the 20 files per repo that actually reveal how you code.
- 03Grade each repo. All repos run in parallel through a fast scoring model that reads the picked files and rates each one independently on Impact, Quality, and Depth — with evidence citations.
- 04Aggregate. A larger reasoning model combines the per-repo scores with server-computed stats (heatmap, commit cadence, language entropy, follower count) to produce the 6-dimension profile score + roasts.
- 05Correct.Deterministic server-side checks enforce anchor-scale floors (e.g. a profile with 2,000+ public commits can't score 30 Consistency) and recompute the final verdict.
~90 seconds per profile, ~$0.25 in compute. Total of ~240 files read across your top-12 repos. One rating per GitHub account per day.
▸ Data sources & caveats
- Heatmap & commit totals: GitHub GraphQL
contributionsCollection— covers the last 365 days, includes private repos when the user has opted in (default). - Language %: byte totals across the top 30 owned non-fork repos.
- Curve: a small upward nudge centered on raw score ≈ 70, capping at 100. Prevents specialists from being unfairly penalised for narrow breadth.
- Anchor corrections: when server-measured signals (e.g. privateWorkLikely, multiRepoVolume, follower count) mandate a minimum category score, the aggregation step enforces it. These are signal-conditional, not identity-based floors.