Pinned Loading
-
chess-bench
chess-bench PublicLLM chess arena: pluggable players (random, scripted, any LLM provider) play full games; Elo leaderboard, CLI, full test suite
Python
-
text-style-transfer
text-style-transfer PublicActivation-steering style transfer for open-weight LLMs: residual-stream style vectors, forward-hook injection, bridge-style composition, CLI
Python
-
yt-performance
yt-performance PublicPredicting channel-relative YouTube video performance from title embeddings, thumbnails, and transcripts (CatBoost / AutoGluon / NN), with documented negative results
Jupyter Notebook
-
ashby
ashby PublicJob-search copilot for ML roles: polls Ashby/Greenhouse boards, scores postings against your resume three ways (LLM rubric, coverage, embeddings), pre-fills forms - you click submit
JavaScript
-
mana
mana PublicMagic: The Gathering collection and deck manager - React 19 + TypeScript PWA, Node/Prisma backend, tested service layer
TypeScript
-
ML4FF
ML4FF PublicNFL player decline prediction, walk-forward validated over 15 seasons: AUC 0.878, 3,389 out-of-sample predictions, calibrated risk tiers
Python
If the problem persists, check the GitHub status page or contact support.



