01 · Roasts
Launch-Day Larry
Learn-Open-Harness (317 stars) and OpenHarness-notes (7845KB, 114 tests) were both created AND last-pushed on the same day as their parent projects. Shipping fast is a skill; shipping *only* fast is a pattern.
Test Allergist
10 out of 12 repos have HAS_TESTS=no. You wrote 114 tests for OpenHarness-notes in one afternoon but couldn't find time for a single pytest in mind-ai-quest, blog, coding-agent, or components-lib. Selective diligence is still a gap.
CI Ghost
Only 2 repos (minimind-notes, joyehuang) have CI configured. You're deploying to Vercel, ModelScope, and GitHub Pages but running no automated pipelines on 10/12 repos. Hope nothing breaks on merge.
Heatmap Hibernator
Your commit heatmap is a flatline for the first 34 weeks of the year, then a frantic sprint. 690 commits squeezed into ~18 weeks looks less like 'engineer' and more like 'semester project panic mode.'
Star Imbalance
317 of your 431 total stars live in a single repo. Strip Learn-Open-Harness and you have 114 stars across 42 repos — that's 2.7 stars per repo. The portfolio is wide but the impact is a single point of gravity.
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% weight68C
- Consistency20% weight65C
- Quality20% weight69C
- Depth15% weight65C
- Breadth10% weight72B
- Community10% weight55D
03 · Stats
365-day commit heatmap
133 active days
Language distribution
- JavaScript52%
- Python29%
- TypeScript13%
- Java2%
- HTML1%
- CSS1%
- Other2%
04 · Numbers
Owned repos
non-fork
35
Commits
last 12 months
690
Followers
50
Joined GitHub
Sep 2023
05 · Top repos
joyehuang /
Learn-Open-Harness
Interactive educational tutorial for OpenHarness built with Next.js 16 + TypeScript. 12 chapters with animated diagrams, quizzes, and interactive simulators. Launched day-after OpenHarness release, shipped with clean structure and bilingual i18n support. No CI/tests but well-documented with meaningful project architect
joyehuang /
minimind-notes
Educational LLM training tutorial with modular experiments. Typed Python (PyTorch), extensive docs (docs/, design.md, ARCHITECTURE.md), CI/CD (HAS_CI=yes), good code structure. 100 stars, ~30 commits last 6mo, 35MB codebase. Clear educational mission but narrow audience impact.
joyehuang /
mind-ai-quest
Educational interactive web app with Next.js + Three.js 3D scenes teaching ML concepts via gamified rice-farming metaphor. Well-typed TypeScript, structured components, good UX design, but no tests/CI and minimal external adoption signals (1 star, 0 followers visible).
joyehuang /
blog
Personal Astro blog with TypeScript, React interactive terminal, dynamic OG generation. Clean typed architecture, minimal stars/forks; shipped working portfolio site—typical active student project with structured layout and meaningful features.
joyehuang /
pydantic-agent-benchmark
Experimental benchmark tool comparing schema-constrained LLM outputs using Pydantic. Early-stage project with clear structure (typed Python, React frontend, mock tools, two benchmark phases), but incomplete execution (runner.py truncated mid-function, no tests/CI, 0 stars).
joyehuang /
coding-agent
TypeScript template for AI-powered coding agents supporting Claude, OpenAI Codex, GitHub Copilot, Cursor, Gemini, and opencode. Multi-agent SaaS scaffolding with Vercel Sandbox integration, PostgreSQL persistence, OAuth2 auth, and MCP server support. Very young repo (3 days old, 4 commits) with minimal evidence of sust
joyehuang /
components-lib
Early-stage Next.js component library with TypeScript, structured layout, and documented components (blur-highlight). Active recent commits (18/30), but no tests, CI, or license. Minimal adoption (0 stars). Mirrors shadcn/ui pattern with CLI aspirations still in design phase.
joyehuang /
OpenHarness-notes
Early-stage AI agent framework with 43 tools, extensive documentation, and test coverage. Created and pushed within hours (Apr 12, 2026), unproven adoption but structured architectural scope and typed Python codebase with meaningful docs justify mid-tier quality baseline.
joyehuang /
joyehuang
Personal profile config repo with README showcasing student projects and experience. Minimal codebase (30KB), no tests or license, but documents portfolio work in AI/full-stack development with links to featured projects.
joyehuang /
desk-widget
Personal macOS desktop widget for Übersicht displaying WakaTime and GitHub daily stats. Well-documented setup but minimal scope, 0 stars, created 2 days ago with 3 commits total. Meets baseline typed+documented standard.
joyehuang /
sbit-en
Mirror/clone of SBTI test with separated images and HTML. Minimal documentation (3-line README in Chinese), no tests/CI, zero stars/forks, 6 commits over ~5 hours. Appears to be a one-off personal project or tutorial fork.
joyehuang /
skills
Minimal Python project with 1 star, no README, tests, CI, license, or documentation. 5 commits over ~2 months with no typed code or structured architecture visible. Appears to be an early-stage experimental scaffold.
06 · Timeline
- Sep 4, 2023Joined GitHub
- Nov 19, 2024Created joyehuang — Config files for my GitHub profile.
- Oct 23, 2025Created blog — joye's personal blog
- Nov 8, 2025Created minimind-notes — 🚀 [从零构建 LLM] 极简大模型训练原理与实践指南。包含 Transformer, Pretraining, SFT 核心代码与对照实验。 | A minimal, principle-first guide to understanding and building LLMs from scratch.
- Feb 6, 2026Created mind-ai-quest
- Feb 13, 2026Created skills
- Feb 24, 2026Created components-lib
- Mar 11, 2026Created desk-widget
- Mar 22, 2026Created coding-agent
- Apr 6, 2026Created pydantic-agent-benchmark
- Apr 7, 2026Created Learn-Open-Harness — 🤖 Official Interactive Tutorial for OpenHarness – Zero to Hero in 12 Chapters | Learn OpenHarness like Claude Code: Agent Loop, Tools, Memory, Multi-Agent | 面向零基础的 AI Agent 交互式教程
- Apr 10, 2026Created sbit-en
- Apr 12, 2026Created OpenHarness-notes
- Apr 28, 2026Most recent push to blog
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.