jiezi-ai/grant

Jiezi Grant, meaning roughly Decode Characters, is not a piece of software. It is a public grant program that gives Chinese university students access to paid AI coding tools, run openly through this GitHub repository. The name references an ancient Chinese book from about 100 years ago that broke down thousands of Chinese characters into their smallest building blocks, so that anyone who learned those basic parts could understand any character. The project applies the same idea to AI: today the smallest unit that AI works with is called a token, and the founder argues that access to enough tokens has become a new kind of gatekeeping, much like literacy once was. The founder, a former engineer at two AI companies, observed that university students are often the people most eager to learn and build with AI, yet a monthly AI coding subscription can cost as much as half a student's monthly living expenses, pricing many of them out. He personally funded the program's first year with 150,000 yuan to sponsor 600 students with API token access. The program runs in three stages: a trial stage open to 600 students in batches of 100, each given a modest amount of API credit, a deeper support stage for about 60 students who show real projects, offering more credit plus a cloud server, and a long term support stage for 3 to 5 standout students with the largest grant, chosen from stage two participants. Students apply by filling out a form on the project's website, then submitting a GitHub Issue with their application code, verifying their university email address, and receiving their API token details by email. The whole process happens through this repository. Applications are handled as GitHub Issues, an intentional choice meant to introduce students to using GitHub and open source itself, not just verify their identity. All policies, budgets, spending records, and changes are published openly in the repo for anyone to review, and the underlying financial ledger uses double entry bookkeeping. The repository's content is released under the Creative Commons Attribution 4.0 license.