Nick Trenkler
for the third week in a row, andrej-karpathy-skills by forrest chang keeps its position among the most-starred github repositories with over 24k stars gained in the last 7 days
also, two repos have shown amazing growth this week: free-claude-code with 85.3% and hugging face's ml-intern with 83.80%
here are the highest-starred repositories of the week:
andrej-karpathy-skills +24,129
free-claude-code +16,154
Open Generative AI by anil matcha +3,823
ml-intern +6,388
FinceptTerminal by fincept corporation +5,117
the five repos and what they do
three of the five repos help developers work with ai coding agents. here is what each one does, in ranking order:
- andrej-karpathy-skills – a single
CLAUDE.mdfile for claude code, built on andrej karpathy's post about llm coding pitfalls - free-claude-code – a proxy that sends claude code's requests to other model providers: nvidia nim, openrouter, deepseek, lm studio, llama.cpp or ollama
- open generative ai – an open-source studio for ai image and video generation with 200+ models, pitched as an alternative to higgsfield, freepik and krea
- ml-intern – hugging face's agent that researches, writes and ships machine learning code with the hugging face hub's docs, papers and datasets. it launched in late april
- finceptterminal – a desktop finance app with market analytics, investment research and economic data
open generative ai and finceptterminal are the only two that are not tools for coding agents.
the fastest climbers
karpathy-skills slowed down but kept first place. it gained 40,732 stars the week before and 24,129 this week, reaching 101,911 in total.
free-claude-code grew fastest, by 85.3% to 18,940 stars. it lets developers keep claude code as the interface and pay less for the model behind it, or nothing at all.
ml-intern grew 83.8% to 7,623 stars in its first full week. it was the only repo in the top 5 from a large ai lab.
update, october 16, 2026: where the five repos are now
five and a half months later, four of the five repos are still active. the table compares their stars on april 30 with today.
| repo | stars on apr 30 | stars on oct 16 | growth | status today |
|---|---|---|---|---|
| andrej-karpathy-skills | 101,911 | 220,215 | +116% | no commits since apr 20 |
| free-claude-code | 18,940 | 60,884 | +221% | 74 releases, v6.16.9 on oct 16 |
| finceptterminal | 18,127 | 32,559 | +80% | v4.5.1, oct 13 |
| open generative ai | 10,350 | 30,515 | +195% | v2.0.0, may 23 |
| ml-intern | 7,623 | 10,809 | +42% | archived |
free-claude-code grew the most. it more than tripled to about 61k stars and shipped 74 releases since april.
ml-intern did not last. on september 14 hugging face retired its cli and its hosted web app and prepared the repo for archiving. the readme now points users to huggingchat.
karpathy-skills more than doubled to 220k stars without a single commit after april 20.
the community tools that make claude code cheaper or better kept growing. the one tool from a large lab was shut down within five months.
sources
- andrej karpathy's post on llm coding pitfalls – january 2026, the source of the rules in karpathy-skills
- free-claude-code releases – v6.16.9, oct 16
- retire ml intern cli and prepare repository archival – hugging face, september 14, 2026
- open generative ai releases – v2.0.0, may 23
- finceptterminal releases – v4.5.1, oct 13
