# yuxiaopeng/Github-Ranking-AI: fifteen topic tables ordered by stars, seven other columns nobody ranks on > This repository regenerates a daily table of the most-starred AI repositories for fifteen topics, and each row carries forks, language, open issues and last commit. Stars decide the order, which is why the same table holds projects updated that morning beside projects untouched for seven months. **yuxiaopeng/Github-Ranking-AI** — A list of the most popular AI Topic repositories on GitHub based on the number of stars they have received, updated automatically every day. - Repository: https://github.com/yuxiaopeng/Github-Ranking-AI - Website: https://yuxiaopeng.com/Github-Ranking-AI/ - Stars: 578 · Forks: 53 - Language: Python - License: MIT - Published: 2026-08-23 · Updated: 2026-08-23 · Language: en - Canonical page: https://hysenlabs.com/projects/yuxiaopeng-github-ranking-ai ## Fifteen topics, ten rows each, and a Top100 behind every table The structure is uniform. The table of contents names fifteen topics: LLM, ChatGPT, OpenAI, Deepseek, LLaMA, Chatbot, AI Agents, Claude, RAG, Mistral, Transformer, MoE, AGI, Generative_AI and AI. Each topic gets a top ten table, and each of those is a truncated view. Under every heading sits the same line pointing at a longer file, Top 100 Stars in LLM at Top100/LLM.md and so on for the other fourteen. So the README is the short form and the Top100/ directory holds the long form, one file per topic, and a project absent from a topic's top ten can still be in its top hundred. Every table uses the same eight columns: Ranking, Project Name, Stars, Forks, Language, Open Issues, Description and Last Commit. The first column is positional within that topic only, so the number 1 means first in that list and says nothing about the project's position in any other. The repository's own description states the basis plainly, a list of the most popular AI topic repositories on GitHub based on the number of stars they have received, updated daily. ## The snapshot runs minutes behind the newest commit it records Freshness is measurable from the file itself. The README carries a Last Automatic Update Time line reading 2026-10-02T09:41:26Z, and the newest Last Commit value anywhere in the visible tables is 2026-10-02T09:29:45Z, belonging to hermes-agent at rank 1 in the ChatGPT table and rank 1 in the OpenAI table. That gap is about twelve minutes. The pipeline is therefore not working from a cache or a nightly export; it is querying GitHub and rewriting the file close to when you read it. Most rows carry a timestamp from 2026-10-01 or 2026-10-02, which is the same pattern. The repository state agrees. The last push to the default branch main is dated 1 October 2026, the repository is not archived, and there are no GitHub releases, so nothing is published as a tagged artifact. The commit itself is the delivery mechanism, and the README is the product. ## Stars do not track activity, and the table proves it in the same row Because Last Commit is a column rather than a sort key, the ranking's blind spot is visible without leaving the page. Two entries make the point. At rank 3 in the LLM table, andrej-karpathy-skills carries 216,429 stars, which is more than enough to place it third overall in that topic, and its Last Commit value is 2026-04-20, about 165 days before the snapshot. At rank 10 in the ChatGPT table, Prompt-Engineering-Guide has 78,778 stars and a Last Commit of 2026-03-11, about 205 days before the snapshot. Only three distinct older dates appear in the visible tables at all, 2026-03-11, 2026-04-20 and 2026-08-11 for NextChat. So a reader has three possibilities in front of them: a project updated the same morning, a project whose last commit is about five months old, and one about seven months old. All three are ranked by the same column, and the ordering says nothing about which is which. This is not a flaw in the data collection. It is what ranking by stars means, and it is why the other seven columns are on the page at all. ## Open issues separate the entries more sharply than stars do The Open Issues column ranges across three orders of magnitude in the visible tables, and that spread is the most useful number after stars. hermes-agent, at 250,674 stars, shows 14,447 open issues. ollama shows 2,531. transformers shows 730, langchain 497, dify 433, OpenHands 455. At the other end, generative-ai-for-beginners shows 1 open issue, LLMs-from-scratch shows 1, and andrej-karpathy-skills shows 0 alongside 216,429 stars and a Language column reading None. Read carefully, this is a workload signal rather than a quality one. A project with fourteen thousand open issues is one where a large user base is filing work faster than it is being closed, and a project with zero is one where nobody is filing, which for a single CLAUDE.md file is exactly what you would expect and not a sign of neglect. The None in the Language column is the same kind of artifact. The entry's description says it is a single CLAUDE.md file, so there is no dominant language for GitHub to report. ## Fork ratios split the list into two different kinds of repository Forks are the column that hints at how a project is consumed, and dividing forks by stars sorts the visible entries into two populations. One group sits around a fifth or less. ECC has 40,500 forks on 270,884 stars, about 15 percent. AutoGPT has 45,974 on 187,656, about 25 percent. transformers has 34,742 on 166,905, about 21 percent. hermes-agent has 53,686 on 250,674, about 21 percent. The other group is far above that. NextChat has 58,949 forks on 88,830 stars, roughly two thirds. generative-ai-for-beginners has 63,601 forks on 120,931 stars, over half, and it is a Jupyter Notebook course rather than a library. LLMs-from-scratch, also notebooks, sits at about 15 percent. Both of the highest-ratio entries ship lessons people copy into their own environment, which is what a fork of a course repository is for. A library like transformers is forked to patch, extend or vendor. Neither ratio is better; they tell you what kind of thing you are looking at before you read a line of its documentation. ## One repository can rank first in three topics, which flatters it Topics overlap, and the tables are generated independently, so a project strong in one area appears in several lists at once. hermes-agent is the clearest case. It is rank 2 in LLM, rank 1 in ChatGPT and rank 1 in OpenAI, from one set of numbers. prompts.chat appears in all three as well, at rank 7, rank 2 and rank 4. AutoGPT appears twice, at rank 5 in LLM and rank 3 in OpenAI. The consequence is about how you read the lists. Counting how many topics a repository tops is not measuring how many independent communities chose it, because the categories are not independent: a chat product is also an LLM product and also an OpenAI product. The sixteen of the twenty-four distinct visible entries appear in one table, four appear in two, and two appear in three. That also means the fifteen tables are not fifteen views of a partitioned set. They are fifteen overlapping filters over the same pool, and a project missing from a topic is a statement about the tagging, not about the project. ## Python fetches the numbers, Jekyll renders the site, a shell script schedules it The repository is small and says so in its dependencies. requirements.txt has two entries and no pins: pandas and requests. The primary language on the repository is Python. The presentation layer is separate and older technology. The top level carries a Gemfile and a _config.yml, which is a Jekyll site, and the project's homepage is served from yuxiaopeng.com rather than from GitHub. So Python collects and rewrites, and a Ruby static site generator publishes what Python produced. The rest of the tree is data and automation. Data/ and source/ hold the working material, Top100/ holds the fifteen long-form files, auto_run.sh is the scheduler behind the daily claim, and ranking.log is a log file committed alongside the output rather than kept outside the repository. There is a .github/ directory, a LICENSE, MIT, and a .gitignore. No release tags exist, so the way to consume this is to clone it, or to read the README and the homepage, rather than to install anything. ## Conclusion The Github Ranking AI list works as a daily snapshot of what the AI open-source ecosystem is drawing attention to, and it does not work as a measure of whether a project is worth adopting today. Every table is ordered by stars, while the same row shows a last commit 205 days old next to one made the morning of the snapshot. Read it as a candidate generator, then check the Last Commit column before anything else, treat Open Issues as workload rather than quality, and remember that one repository can sit at the top of three topics at once, so appearing repeatedly is not independent corroboration. ## FAQ ### How often is the Github-Ranking-AI list updated? The project description says the ranking updates daily, and the README carries a Last Automatic Update Time line, which read 2026-10-02T09:41:26Z in the current version. The newest Last Commit value in the tables was about twelve minutes earlier than that timestamp. ### What columns does the Github-Ranking-AI ranking table use? Eight: Ranking, Project Name, Stars, Forks, Language, Open Issues, Description and Last Commit. Every topic heading uses the same eight columns, so the tables are directly comparable row by row. ### How many topics does Github-Ranking-AI cover? Fifteen, named in the table of contents: LLM, ChatGPT, OpenAI, Deepseek, LLaMA, Chatbot, AI Agents, Claude, RAG, Mistral, Transformer, MoE, AGI, Generative_AI and AI. Each shows a top ten and links to a Top 100 file for that topic. ### Does Github-Ranking-AI rank only by stars? The ordering is by stars, which is what the project description states. Forks, language, open issues and last commit are carried as columns but are not part of the ranking, so they are available to the reader without affecting position. ### Why does a top entry in Github-Ranking-AI show Language as None? andrej-karpathy-skills is listed at rank 3 in the LLM table with 216,429 stars, Language None and 0 open issues. Its description says it is a single CLAUDE.md file, so there is no dominant language for GitHub to report. ## Sources - [Official documentation](https://yuxiaopeng.com/Github-Ranking-AI/) - [Official README](https://github.com/yuxiaopeng/Github-Ranking-AI#readme) - [Project repository](https://github.com/yuxiaopeng/Github-Ranking-AI) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/yuxiaopeng-github-ranking-ai