Chinese Independent Developer: a chronological feed of small web apps, capped by GitHub

One README, five entries, and a host-imposed size limit

The repository is five top-level entries: `.agents/`, `.claude/`, `.github/`, `.gitignore` and `README.md`. There is no code, no build, no site and no license file, and the project is reported as Python only incidentally. The list is a document, maintained by pull request and issue.

The most telling detail is the archive. Alongside the main board there are three separate documents kept under `.github/pages/`: a Programmer Edition for anything that needs a command line or code, a Game edition, and an archive of 2018 to 2024 projects that the README says was split off because of GitHub rendering limits. That last clause is the whole structural story in one line. A single-file list grew until the host could no longer render it, and the fix was to move history into a second file rather than to add structure.

The consequence is that anything looking for this list long term should expect the seam to move. Newer entries live in the main README, older ones sit behind the archive link, and a project that matters to you could be on either side with nothing in the file telling you which.

The main board takes websites and apps, and refuses developer tools

The selection rule is stated in one line and it is restrictive. Entries must be a website or an app, and must not be a developer tool or a forum-type site. There is a further note that what you are reading is the main board, that the products collected there are ready to use, and that they are a different type of product from those on the sub-boards.

So the main board is a directory of finished, browser-openable things, and the Programmer Edition exists precisely to hold what the main board refuses. That is a coherent editorial position and it is stated clearly, which is more than most lists manage.

The consequence is for readers arriving with the wrong expectation. Someone looking for an open-source command line tool, a library or a developer community will find nothing on the main board and has to know that the Programmer Edition document exists to look there instead. The boundary between an app and a developer tool is also a judgement call made by whoever merges the pull request, so a borderline submission may land on either board depending on who read it.

Entries are grouped by the day they were added, not by what they do

The list is organised chronologically. Visible headings run 2026-09-30, then 09-29, 09-27, 09-26, 09-25, 09-24 and 09-23, and under each date sit per-author headings linking to a GitHub profile. Some author headings carry a city, and one entry notes that the author added it themselves rather than through the usual route.

There is no category taxonomy, no tag system and no index beyond the document's own anchors. Each product gets a status from a fixed set of three: in development, launched, and closed or unmaintained.

This is a deliberate trade and it is worth judging on its own terms. A chronological feed is excellent for the question the list exists to answer, which is what is people shipping this week, and useless for the other question, which is what exists in a category. On the visible page, one author has five entries across Morse code tools, a visual timer, printable classroom worksheets and a materials calculator, and nothing connects them except the fact that they arrived together. If you want a catalogue, this is the wrong shape.

The one-line description is the quality bar, and it is self-reported

The list says that writing a good introduction matters, and gives the standard as a before and after. A bare `AI 视频生成工具`, an AI video generation tool, is the rejected form. The accepted form is `无需剪辑经验，3 步生成竖屏短视频`, generate vertical short videos in three steps with no editing experience required. A contributing guide covers how to write it.

That example is a good one, because the difference between the two is not polish, it is information. The first says what category the product claims, the second says what the user gets and what they do not need to know already. Whoever reviews submissions is effectively judging one sentence per product.

The consequence is that the list's usefulness is bounded by its submitters, and it has no way to check them. A person who writes a precise sentence about a product that does not deliver has produced a better entry than a person who writes a vague sentence about something excellent, and the review process described here cannot tell the difference. Every directory built on user submissions has this property; it is worth knowing before you treat the list as a recommendation.

Free is not a consistent signal across entries

Read the commercial terms across a single day's entries and the spread is wide. One tone generator states that it is free, runs entirely in the browser with no upload and no install, and exports WAV. A video-to-prompt tool is free for five videos a day after signing in with Google. A headshot generator gives one free watermarked preview without registration, with watermark-free download and repeated edits sold as a one-time payment. A subtitle translation tool is free and open source but requires you to bring your own cloud service API credentials. A gas station discount finder is a WeChat mini program. One browser MIDI converter is linked with referral parameters.

So the list mixes free utilities, freemium products, paid tools, self-hosted tools and a platform-native mini program, and it does not standardise how any of them are described.

The consequence is that you cannot filter this list by cost. Reading a description is the only way to know what a given entry costs, and the words free and open source appear alongside both genuinely free tools and products with a paid tier. The privacy claims have the same property, appearing per entry as statements that a name list is not uploaded or that audio is processed locally, which are claims the list repeats rather than verifies.

No license file, for a document made of other people's work

The licence field for this repository comes back unknown, and there is no LICENSE file among the five top-level entries. The MIT licence that covers most of the projects I have written about in this batch is simply absent here.

That matters more than it would in a repository containing a program, because this document is not the author's work. Every entry describes a product built by somebody else, and those descriptions were written by the submitters, not by the maintainer. The maintainer's own contribution is the selection, the grouping and the rules.

The consequence is that reading the list and linking to it is unremarkable, while copying its text into your own site, feeding it into a dataset or translating it wholesale sits on an unstated footing, because there is no grant in the repository to point at and no stated exception for the aggregated third-party descriptions. The copyright on each description belongs to whoever wrote it, and the repository does not say how that was handled.

Product names collide with incumbents, so entries carry disclaimers

Several entries in the list use names that belong to better-known products. There is an AI image and video tool called Claude Imagine whose description ends by stating that it is an independent product and has no relationship to Anthropic. There are entries named after established open-source projects, including a tool that shares a name with a known upscaling application, and one that shares a name with a well-known browser photo editor.

The disclaimer is the tell. An independent developer picking a name is optimising for a keyword, and the keyword space is already occupied by companies and established projects, so collisions are structural rather than unlucky. The list's response is to let the submitter note the distinction in the description.

The consequence is that a reader's first assumption about any entry will often be wrong, and the only correction available is a parenthetical in prose. There is no field in the format for declaring a relationship, or its absence, so that information lives in the same sentence as the marketing copy and is easy to skim. The three-status scheme helps here, since closed or unmaintained is recorded, but nothing records whether a product is related to an incumbent.

Editorial conclusion

This list is a good way to find what individual Chinese developers are shipping right now, and a poor way to research a category. Its design optimises for recency and for the author's curiosity about side projects, which is exactly what it says the list is for, and the resulting artefact is a stream rather than an index. Three things to know before you rely on it. Developer tools and forum sites are excluded by an explicit rule, so a CLI project will only appear on the separate Programmer Edition board. Free is not a consistent signal, since entries range from no-registration free tools to one-time paid watermark removal and to products that need your own API key. And there is no LICENSE file, so the licence under which you may reuse the text is unstated. Before you point a team at it, check three things: whether the Programmer Edition board is where your interest lies, whether an individual entry is still alive given the list tracks a closed or unmaintained status rather than a date, and what you intend to do with the descriptions, which were written by the submitters.

Frequently asked questions

Official sources

1. Official README
2. Project repository