01 · Roasts
2-Hour DB Architect
rust-not-so-sql went from concept to broken parse() signature and unimplemented!() in a heroic 2-hour window. The code won't compile, but the ambition is immaculate.
README > Tests
12 repos analyzed: exactly 2 have tests, exactly 1 has CI. You write ARCHITECTURE.md, STATUS.md, and docs/ folders — but apparently testing whether the code runs is someone else's problem.
Research Dumper
SMART-SSR, _SMART-SSR, and beyond-superni are three separate repos doing roughly the same continual-learning research thing. Consolidation is a skill too.
Sprint-and-Ghost
rust-redis: 4 days. rust-not-so-sql: 2 hours. alacritty-config: 30 minutes. You have a gift for starting things and a talent for never returning to them (staleRepoRatio: 0.48).
9 Stars, 67 Repos
67 public repos, 9 total stars — that's 0.13 stars per repo. The QPU driver and WebRTC demo are genuinely interesting; it's a shame nobody can find them under the avalanche of one-shot experiments.
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% weight62C
- Consistency20% weight65C
- Quality20% weight57D
- Depth15% weight58D
- Breadth10% weight65C
- Community10% weight55D
03 · Stats
365-day commit heatmap
79 active days
Language distribution
- Python65%
- Shell10%
- TypeScript7%
- C++5%
- JavaScript4%
- Rust2%
- Other7%
04 · Numbers
Owned repos
non-fork
46
Commits
last 12 months
162
Followers
20
Joined GitHub
Jun 2019
05 · Top repos
yiannisha /
_SMART-SSR
Research implementation of SSR and SMART-SSR for continual LLM learning on SuperNI tasks. Typed Python with structured config, documented pipeline, and multi-stage training infrastructure. No tests or CI, limited adoption signals.
yiannisha /
personal-blog
Personal Next.js blog/portfolio site built in TypeScript with a feed system, markdown rendering, and admin CMS. Typed, structured, and documented, but zero stars/adoption and experimental scope.
yiannisha /
SMART-SSR
ACL2024 paper implementation on LLaMA-Factory for mitigating catastrophic forgetting in LLMs through self-synthesized rehearsal. Typed Python codebase with structured multi-file layout and README, but no tests, no CI, and minimal external adoption (0 stars/forks).
yiannisha /
beyond-superni
Specialized LLM benchmark harness for SuperNI evaluation across OpenAI and Hugging Face providers, with configurable few-shot learning and metric reporting. Brand new (2 days old), experimental stage with structured Python codebase but no tests.
yiannisha /
ros-intercomms-benchmark
Young ROS 2 pub/sub stream benchmark tool with typed Python, structured layout, tests, and clear README. Single-purpose experimental project with minimal adoption.
yiannisha /
qpu-xla
Early-stage GPGPU framework for Raspberry Pi 5 VideoCore QPU with assembler, driver, and SGEMM reference kernel; 0 stars, ~2 months active with 30 commits, unfinished roadmap.
yiannisha /
libdatachannel-demo
Educational C++ demo of WebRTC producer-consumer architecture using libdatachannel, with GStreamer video pipeline and browser viewer. Functional and documented but experimental, created ~4 hours ago with 7 recent commits.
yiannisha /
microgpt.c
Experimental C implementation of a GPT-like neural network with autograd, incomplete and non-functional (undefined variables in main). No README, tests, CI, or license; lacks documentation and structured architecture.
yiannisha /
rust-redis
Experimental Redis implementation in Rust using Tokio; no README, no tests, minimal commits (6 of last 30), learning-stage code with incomplete features and unpolished error handling.
yiannisha /
process_pid
Learning project exploring PID control based on tutorials; no code files sampled, zero commits in last 30 days, minimal output, no tests/CI/license.
yiannisha /
rust-not-so-sql
Experimental Rust SQL DB with 6 commits in 2 hours, partial parser/executor skeleton, no tests/CI, incomplete implementation with broken parse method signature and unimplemented features.
yiannisha /
alacritty-config
Personal config scaffold with 3 KB of content, 2 commits within 30 minutes, no README, tests, CI, or documentation. A one-shot configuration dump.
06 · Timeline
- Jun 1, 2019Joined GitHub
- Sep 13, 2024Created personal-blog
- Jan 9, 2026Created _SMART-SSR — SMART-SSR: Scoring-Based Synthetic Rehearsal for Continual LLM Learning
- Feb 12, 2026Created rust-not-so-sql — Simple SQL DB implementation in Rust.
- Feb 12, 2026Created microgpt.c
- Mar 3, 2026Created alacritty-config — My alacritty config with a bunch of different themes to choose from.
- Mar 4, 2026Created rust-redis
- Mar 12, 2026Created qpu-xla
- Mar 17, 2026Created beyond-superni — A contamination-aware benchmark suite for rehearsal-based continual learning in modern LLMs.
- Mar 21, 2026Created SMART-SSR
- Mar 27, 2026Created process_pid
- Apr 23, 2026Created ros-intercomms-benchmark
- Apr 26, 2026Created libdatachannel-demo
- Apr 26, 2026Most recent push to personal-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.