# A courseware nobody bought, open-sourced, gained 9 stars in 1,600 days: a lesson from Tongjiang Lab for anyone thinking about open source · GitHub open source projects

> Source: https://stringers.io/en/github/gh_smithjohn78390__tongjiang-lab.html (Stringers / 什么值得做, generated 2026-11-08)
> Stringers answers one question every day: what small product can one or two people build in a few months and actually make money from. Numbers come from 15 databases we collect daily, each with source and date.
> For agents: these verdicts and the underlying databases (verified revenue, App Store / Google Play charts, ad libraries, GitHub, Product Hunt…) are queryable live over MCP — tools today / hot / products / directions / direction / data. Setup: https://github.com/komako-workshop/stringers (the user logs in at stringers.io → "配置 agent" to get a personal URL). Site index: https://stringers.io/llms.txt

> Dropping a large batch of finished, free courseware that matches China's textbooks and works offline, with a punchy headline; the author is an old Chinese AI-circle X account, and AIGCLINK, Gorden Sun, and GitHubDaily kept it going with reposts, while the story of "one person used AI to build 54 courses" gave the AI crowd a reason to share.

[GitHub open-source project](https://stringers.io/en/github/index.md) · written 11-6

smithjohn78390/tongjiang-lab · smithjohn78390

# A courseware nobody bought, open-sourced, gained 9 stars in 1,600 days: a lesson from Tongjiang Lab for anyone thinking about open source

**What it is** 54 interactive 3D high school courses aligned to the People's Education Press textbooks: physics, chemistry, biology, geography. Each course is a single HTML file of about 1MB that you can copy to a USB stick and double-click to project offline. Free and open source.

**Why it spread** A big batch of free finished products at once, matched to China's textbooks, works offline, with a punchy headline. The author is an older Chinese AI-circle X account, and AIGCLINK, Gorden Sun and GitHubDaily relayed it. The story "one person built 54 courses with AI" also gave the AI crowd a reason to share.

**B** **How you could follow** Don't redo physics, chemistry, biology or geography. Use the same workflow of "fact table + shared shell + AI mass production + human review" to fill the gaps it left: high school math, junior high science, or other countries' textbooks. Release one textbook's worth at a time, and reach out to Chinese open-source news accounts before launch.

Grade B: clear demand, nobody has followed yet, and issues explicitly ask for math and junior high. But most of the stars rode the author's X position and relays from big accounts. Similar projects from the same period without that distribution got only 100 to 400 stars, and building content for even one subject is expensive. Buzz in small circles

Data (updated daily) · Repo created 2026-10-28 · 1,667 stars · 7 days +485

GitHub · stars

Data 2026-10-29 → 2026-11-08 (11 days, 11 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

On Oct 28 2026, around noon Beijing time, a Chinese AI blogger on X called "Tongjiang · Bedtime at Ten" posted a long thread. The first line: "I've decided to open-source the 3D interactive education course I spent several months building. Only a few scattered subscriptions a month, it's a waste." 45 minutes earlier he had created the GitHub repo and pushed 54 courses, all 3D, aligned to the People's Education Press high school physics, chemistry, biology and geography at once. 9 days later the repo had 1,644 stars and 416 forks (GitHub, 11-06). It never hit trending, ran no ads, and has not been touched since the 5 commits pushed on creation day. Three things are worth a look: why it spread, why the fork count is unusually high, and what gaps it left.

### What it is

Tongjiang Lab (lab.tongjiang.app) is a set of 54 interactive 3D lessons on high-school physics, chemistry, biology and geography. Every topic comes from the textbook: Haber-Bosch ammonia synthesis, the contact process for sulfuric acid, chlor-alkali production, blast furnace ironmaking, magnetic fields from current, the photoelectric effect, mitosis, nerve impulses, the three-cell circulation model, El Niño... Each lesson is tagged with its volume, chapter and section in the People's Education Press (2019) textbook series.

Open a course and on the left is a 3D scene you can drag, rotate, and cut open to see inside a tower; on the right a control panel where sliders and switches can be moved at every step, with readings and curves updating live. Below is a step line of 7 to 9 steps, so a teacher can get through a whole lesson by pressing "next", ending with an assessment that has target values.

The points the author cared most about in his original post are also the ones that got shared most:

- Each course is one HTML file of about 1 MB: "copy it to a USB stick, double-click to open, works offline, on a projector or an electronic whiteboard";

- Every number is traceable: values from the textbook carry page numbers, and values set for teaching are marked separately as "courseware settings";

- They can be downloaded, and an embed code can be copied in one click to drop onto a school website.

The license is CC BY-NC 4.0, no commercial use; the site has no login and no tracking cookies.

### How one person built 54 courses

The site's "how it was made" page and the original post lay out the pipeline clearly: first have the model write a fact sheet from the textbook (column A is textbook facts with page numbers, column B is the numbers and formulas set by the courseware); all courses share one 3D shell (scene, control panel, step bar), and each course only fills in content and scene config. DeepSeek batch-checks textbooks, writes fact sheets and builds courseware in bulk; Claude sets the shell and visual spec and handles review. Each course is run through automated checks at three screen sizes step by step (numbers, occlusion, contrast, whether labels overlap), then a human reviews the screenshots one step at a time.

He doesn't hide the weak spots either: "To save tokens I leaned heavily on DeepSeek, so the results aren't as good as my SaaS site" and "there are definitely still errors in the 55 courses." These courses were originally paid content in his North American AI SaaS version; they didn't sell, so he turned them into an open-source domestic version. In other words, behind these 54 courses is months of work already finished, not something that started the day the repo was created.

### Where the stars came from: one post, three big accounts passing it along

The repo was less than 60 days old and couldn't get onto GitHub trending, so every star came from outside the site. The timeline lines up:

| Date | Stars (GitHub, per day) | What happened |
|---|---|---|
| 10-26 | – | The author teased on X: "I want to build this site out in public" |
| 10-28 | – | Repo created; 45 minutes later the original post went up, ending with "please forward it to teachers you know, or give the repo a ⭐" |
| 10-29 | 570 | The Chinese AI account AIGCLINK relayed it; Sina, AGI Hunt and an English review followed the same day |
| 10-30 | 897 | Single-day +327, peak |
| 10-31 | 1,091 | Retweeted by Gorden Sun, 6,452 views |
| 11-03 | 1,434 | GitHubDaily recommended it the night before; +184 that day |
| 11-06 | 1,609 → 1,644 | Daily gains dropped to around 40 |

Then came the long tail: a Zhihu column "One person used AI to build 55 3D interactive science courses, all free and open source," Rengren Product Manager (11-04), an in-depth review (11-05), CSDN reposts, and a pile of resource and education directory sites picked it up; people also reposted it on YouTube and TikTok.

The author himself was a key link. @CopperForgeAI had posted 4,000 times, writing long-term hands-on AI reviews, so he was already inside the social circle of big Chinese AI accounts, which is why the original post was picked up the next day. News accounts shared it partly for the topic ("54 courses, four subjects, PEP textbooks, free" is a ready-made hard headline) and partly for the story: the Zhihu piece said plainly that "this project is itself a good example of AI-produced content."

### Unusually many forks: teachers are actually using it

The fork-to-star ratio is 25% (411 / 1,609,11-06); a typical new open-source project runs about 11%. Looking at the top 10 forks by stars, they're all verbatim copies with no new commits, which is "copy it and keep it." This matches the form of the courseware: teachers want files, not code.

The Issues section shows real users. A user named bilsemc2 raised 8 science details on the day the repo was created, for example in geography: "0° and 180°E differ by 12 hours; if 0° is 08:00,180°E it should be 20:00." Other issues were requests: "Do some math! Math is harder to learn in high school" (11-05), when the English version comes out, whether it could be made into a language pack so the community can contribute translations (11-03), the "Home" button in the course is 404, index.html is missing, and someone asked for a partnership email.

16 issues, and as of 11-06 not one was closed; the author replied to none (only 2 comments in total, all from other users); the repo hasn't been pushed to since 10-28. The author says he'll finish the four subjects, then do junior high and primary school, and release an English version, but from the repo it currently looks stalled.

### Who else does the same thing

| Name | What it is | How it differs |
|---|---|---|
| PhET (University of Colorado) | 160+ free science simulations, since 2002 | 2D general-concept simulations, not aligned to PEP chapters, no "finish a lesson" step flow |
| NOBOOK virtual labs | Commercial platform; its site claims 30k+ schools use it | Paid, requires account authorization; equipment can be freely assembled |
| Seewo Whiteboard and 101 education PPT 3D resources | Features inside big-brand whiteboards | Locked into their own software, not files you can take away |
| atlas-of-physics | English, 3D interactive textbook aligned to AP Physics, built the same day | Nearly the same form, but physics only, 389 stars (GitHub, 11-06), a quarter of Copper Forge's |
| cn-math-taxonomy | Math/English/physics knowledge graphs and question banks aligned to curriculum standards, built 10-18 | Not 3D courseware, but 6,755 stars shows the "China curriculum standards + free open source" theme is hot right now |
| ChinaStudyFree | Interactive course platform for 148 textbooks, built 10-16 | Broader in scope, no 3D experiments, 104 stars |

Each piece alone has been done, but "lesson-by-lesson PEP alignment + 3D + classroom step flow + free offline single file" together had not been done before. No copycat Chinese 3D courseware repos found yet.

The last three rows are worth a closer look: three projects with similar forms in the same month, one with 6,755 stars, one with 389, one with 104. The difference isn't all about content quality; it's more about whether anyone shared it and whether the title is understandable at a glance.

### If it were you

**Verdict** : worth trying. No one is following this direction yet, and the demand is written out in the issues, but more than half the stars came from the author's position on X and news accounts relaying it, and you may not have that.

**Who it fits** : developers who know a particular subject well and can judge whether it's taught correctly, or who have teachers around who can review for them; it's better if you already have some following in Chinese developer or education circles.

**Who it doesn't fit** : people who want to build a polished general-purpose 3D engine and fill content in later (the stars came from finished content, not the tool); people who can't do content review (teachers will pick out errors line by line from day one, and if there are too many, word of mouth turns).

**How to do it** : pick a piece it hasn't covered, and don't redo physics, chemistry, biology and geography. The most direct are three gaps:

1. **High school math** : someone named it in the issues and the author hasn't built it; function graphs, solid geometry and conic sections are a natural fit for 3D and sliders.

2. **Junior high science** : the author said "later," but the repo has been stalled for over a week; there are more junior high teachers, and courseware for them is scarcer.

3. **Other textbook systems** : A-Level, IB, Taiwan's 108 curriculum, or national textbooks in Vietnam or Indonesia; copy the workflow of "fact sheet → shared shell → cheap model mass-produces, strong model reviews → multi-size screenshot self-checks + manual image review." The English AP physics one only got 389 stars, which shows distribution has to be rethought when you switch languages, so don't expect it to take off on its own.

Copy the form exactly: single file, offline, embeddable, every number traceable. Ship enough finished product at once (at least one book's worth) so there are numbers you can headline on day one. Also do what it hasn't done well: reply to issues, fix errors, and provide a usable table of contents page; these are exactly what people are complaining about now.

**How to get seen** : follow its release rhythm. Tease "building in public" first, then have one original post explain the method and numbers in full, ending with a clear request to "forward it to teachers and star it." Before posting, contact Chinese open-source news accounts like AIGCLINK, GitHubDaily and Gorden Sun, then write a long method piece on Zhihu and Rengren Product Manager. Teacher groups and resource sites will spread it from there.

**When to move, and how early** : this "curriculum-aligned free education content" wave started in mid-October, and Copper Forge's daily gains have already dropped from over 300 to 40. Aim for within a month, meaning release the first batch by early December at the latest; later than that, the original author may add math and junior high himself, and news accounts will get tired of the theme.

**Signs you've misread it** : if your repo hasn't passed 300 stars a week after release, or the accounts you pitched won't repost it; or if Copper Forge updates its repo and ships math and junior high courses before you launch, in which case head-on overlap isn't worth it and you should switch textbook systems.

[GitHub repo ↗](https://github.com/smithjohn78390/tongjiang-lab)

**Research notes: competitor table, acquisition evidence, full research report**

Competitors

| Name | Where | Launched | Scale | How it differs from this one |
|---|---|---|---|---|
| PhET Interactive Simulations | Web | 2002 | 160+ free simulations, the largest in the world | PhET is 2D general-concept simulation, not aligned to PEP, with no classroom step scripts |
| NOBOOK virtual labs | Web | About 2015 | Its site claims 30000+ schools | Commercial paid platform requiring account authorization; Copper Forge is free, a single offline file, and embeddable |
| Physics Lab / Turtle Lab | iOS | 2010 era | Chinese physics simulation app | Only a physics sandbox app, not courseware organized lesson by lesson from the textbook |
| Seewo whiteboard subject tools / 101 education PPT | Big-company features | Years | Covers a large number of public schools | Locked into its own whiteboard ecosystem; Copper Forge is an open browser file |
| nathanliow/atlas-of-physics | Open source | 2026-10-28 | 389 star | English, aligned to AP Physics, physics only |
| xiaoyang111222/cn-math-taxonomy | Open source | 2026-10-18 | 6755 star | Knowledge graph and question bank aligned to curriculum standards, not 3D courseware; evidence of the same theme's heat |
| tanle-mtr/ChinaStudyFree | Open source | 2026-10-16 | 104 star | A broader textbook interactive platform with no 3D experiments, didn't take off |

Where users come from

- **Verified** **The author's own original X post (@CopperForgeAI, an older account, 4,333 posts)** : 10-28 05:14 UTC original post, published about 45 minutes after the repo was created, ending with a request to "forward it to teachers you know and star ⭐"

- **Verified** **Chinese AI media big accounts reposting** : AIGCLINK 10-29, Gorden Sun 10-31 (6,452 views), GitHubDaily 11-01; matches the peak of 10-30 and the +11-03 on 184

- **Verified** **Media, resource sites, content-farm reposts** : Sina 10-29, Zhihu column, Woshipm 11-04, geeyo review 11-05, CSDN, maomp, sucaizy, Zhijiaoyun, YouTube, TikTok @share1006, Facebook dseditorlab

- Inferred **SEO/GEO** : the author's November plan mentions doing SEO and GEO work; no results visible yet

## Tongjiang Lab: why it got 1609 stars in 9 days

> Repo: [github.com/smithjohn78390/tongjiang-lab](https://github.com/smithjohn78390/tongjiang-lab) · Site: [lab.tongjiang.app](https://lab.tongjiang.app) · Channel: GitHub · Research date 2026-11-06

### One line

Chinese AI X creator "Tongjiang · Ten O'Clock Bedtime" (@CopperForgeAI) took the 3D interactive courseware that wasn't selling in their North American AI SaaS, adapted it to the People's Education Press (PEP) textbooks, and open-sourced 54 courses at once (each a single offline HTML file of about 1MB). The traction came from a relay of Chinese AI/open-source news accounts (AIGCLINK → Gorden Sun → GitHubDaily), plus a wave of reposts from content farms and resource sites. What actually makes people star and fork it is the form: free, aligned to specific PEP chapters, and copyable to a USB drive to double-click and project in class, plus the story of "one person + an AI pipeline".

### 1. Data

| Date | star | fork | Stars gained that day |
|---|---|---|---|
| 10-28 repo created (04:29 UTC) | – | – | – |
| 10-29 | 570 | 158 | About 570 on day one |
| 10-30 | 897 | 240 | +327 (peak) |
| 10-31 | 1091 | 296 | +194 |
| 11-01 | 1182 | 313 | +91 |
| 11-02 | 1250 | 332 | +68 |
| 11-03 | 1434 | 367 | +184 (GitHubDaily pushed a wave) |
| 11-04 | 1523 | 388 | +89 |
| 11-05 | 1570 | 398 | +47 |
| 11-06 | 1609 | 411 | +39 |

Source: our GitHub repo_day database (2026-10-29 to 11-06). Over 7 days it gained 712 stars. The fork-to-star ratio is about 25.5%, while the median for new repos with ≥5000 stars is about 11% (see SCHEMA.md). That suggests many people downloaded the files to use them. I sampled the top 10 forks by star count: all have only 1 stars, carry the same name as the original, were created within a few days of the original, and have no new content pushed. They are plain "copy and keep" forks, not reworked copycat repos. The repo's last push was on 2026-10-28 11:35; it pushed 5 commits within 7 hours on creation day and then stopped (geeyo's review says the same). 16 issues, none closed (GitHub API, 11-06). The author's GitHub account has only 2 repos; the other, curated-translations-zh, has 152 stars.

### 2. Growth tactics: where users come from

Timeline (add 8 hours for Beijing time):

1. **10-26 teaser** : the author said on X, "I want to build this site publicly... none of my earlier sites really sold, so I might as well use them for content" (https://x.com/CopperForgeAI/status/2114583434911961267).

2. **10-28 05:14 UTC original post** , published about 45 minutes after the repo was created (https://x.com/CopperForgeAI/status/2115311202876453351). The post explains the origin: "Just a few scattered subscriptions each month felt like a waste. Since my earlier AI SaaS was the North American version... I can now open-source 55 courses". It lays out the method (DeepSeek builds the courses, Claude handles the shell and review, automated checks on three sizes, human review of screenshots), and ends by asking people to "share it with teachers you know, or give the repo a ⭐". The account has posted 4,333 times, mostly AI hands-on tests and a 3D factory series, so it was already an active Chinese AI-community account (profile [x.com/CopperForgeAI](https://x.com/CopperForgeAI)).

3. **10-29 04:21 UTC, relayed by AIGCLINK (@aigclink)** (https://x.com/aigclink/status/2115660284358775293). This is a major Chinese AI-community account. The same day, Sina (https://www.sina.cn/news/detail/5359280251210869.html, 5359331977463123), AGI Hunt, the jamecling blog, and the English-language SI Build Lab review (https://sibuildlab.com/articles/tongjiang-lab-tested/) all followed. About 570 stars on day one; it peaked at +327 after 10-30.

4. **10-31, Gorden Sun (@Gorden_Sun) reposted** , 6,452 views (https://x.com/Gorden_Sun/status/2116476410105929954).

5. **11-01 23:30 UTC, GitHubDaily (@GitHub_Daily) recommended it** (https://x.com/GitHub_Daily/status/2117036490987651256). The +184 on 11-03 lines up with this wave.

6. **Long-tail reposts** : Zhihu column "One person used AI to build 55 3D science interactive lessons, all free and open source" (https://zhuanlan.zhihu.com/p/2100185926089564847), Woshipm (11-04, [www.woshipm.com/ai/6488772.html](https://www.woshipm.com/ai/6488772.html)), the geeyo in-depth review (11-05), CSDN/gitblog SEO articles, resource sites like maomp and sucaizy, the Zhijiaoyun navigation site, a YouTube intro video (QyNVZfdxY50), TikTok @share1006, and the Taiwanese Facebook page dseditorlab.

Conclusion: the core traffic came from a **relay of Chinese AI/open-source news accounts** , plus **automatic reposts by resource sites and content farms** . The subject itself is "education + free + built by one person with AI", which news accounts love to share, and teacher groups pass it on further. I saw no paid ads; the author's November plan mentions "SEO and GEO practice" (from an X post summary, seen via Jina search), but SEO has not yet become a source. Trending: the repo was under 60 days old, so it could not enter trending, and all stars came from off-platform.

### 3. Competitors

| Name | Where | Launched | Scale | How it differs |
|---|---|---|---|---|
| PhET Interactive Simulations (University of Colorado Boulder) | Web / open source | 2002 | 160 simulations, the world's largest free science simulation library (public information) | PhET is a general conceptual 2D simulation set. It is not aligned to PEP chapters and has no step-by-step scripts for a full lesson. Tongjiang is 3D, follows PEP lesson by lesson, and includes step lines and assessments. |
| NOBOOK virtual labs | Web/client, paid | About 2015 | Official site says "30000+ schools, 2000+ institutions" (nobook.com) | A commercial platform schools buy. Equipment can be assembled freely, but it requires an account and license. Tongjiang is free, a single offline file, and can be embedded in school websites. |
| Physics Lab / Turtle Universe | iOS/Android | 2010 era | Chinese physics simulation app with a community | Physics-only (electricity, astronomy) sandbox, and an app rather than courseware. Tongjiang covers four subjects in courseware form. |
| Seewo whiteboard subject tools, 101 education PPT 3D resources | Big-company features | Has existed for years | Interactive whiteboards used across many public schools | Tied to their own whiteboard and software ecosystem. Tongjiang is an open file that opens in any browser. |
| nathanliow/atlas-of-physics | Open source | 2026-10-28 (same day) | 389 stars (GitHub repo, 11-06) | English, aligned to AP Physics, 3D interactive textbook. The form is nearly identical, but it targets the US exam and covers only physics. Its stars are a quarter of Tongjiang's. |
| xiaoyang111222/cn-math-taxonomy | Open source | 2026-10-18 | 6,755 stars, 1,758 forks | Math/English/physics knowledge graphs and question banks aligned to China's curriculum standards, not 3D courseware. Still, it shows the same "China curriculum standards + free and open source" theme is hot in the Chinese GitHub community. |
| tanle-mtr/ChinaStudyFree (Congcong Academy) | Open source | 2026-10-16 | 104 star | The interactive course platform for 148 textbooks. It is broader, has no 3D experiments, and did not take off. |
| OpenLab Science | Web | Not verified | 8 labs, 138 tutorials | English, aligned to AP/IB |
| Papermorph, codepawl/viz3d | Open source | 2026-10 | 1,230 / 143 star | Tools that let AI turn a book or course into an interactive web page. Tongjiang ships finished content directly, not a tool. |

### 4. Why it took off

1. **Lots of finished content released at once, and free** : 54 courses loaded in one go, each with 7–9 steps, labeled by textbook volume/chapter/section (catalog.json). Outlets can write hard headlines like "54 courses, four subjects, PEP". Off-the-shelf tools like PhET are not tied to Chinese textbooks, and NOBOOK charges money, so for Chinese science teachers this was an empty spot.

2. **The form suits the classroom** : a single file of about 1MB, copied to a USB drive and opened by double-clicking. It works offline, projected, and on electronic whiteboards, and can be embedded in a school site via iframe. This is the point repeatedly singled out in reposts (a comment under Gorden Sun's post: "works offline, that's practical"; GitHubDaily: "convenient to bring into the classroom"). SI Build Lab ran all 8 courses on a machine without a GPU, and they all opened.

3. **The "one person + AI pipeline" story** : first write a fact sheet with page references, use one shared shell for all lessons, DeepSeek builds the lessons and Claude reviews them, automatic checks on three sizes plus human review of images (https://lab.tongjiang.app/how/). AI-community accounts reposted it as a sample of "AI-made content", and the Zhihu article explicitly calls the project "a good sample of AI-made content".

4. **The author already had a presence on X** : an established Chinese AI-community account. Within the big accounts' social circles, AIGCLINK picked it up the same day it was posted.

5. **The timing is right** : in late October there was a wave of "free education content aligned with China's curriculum standards" on GitHub Chinese circles (cn-math-taxonomy 10-18 onward, gaining hundreds of stars a day), and the same week atlas-of-physics, an AI-built 3D textbook, showed up in the English circles. Readers are primed for this kind of content.

6. **Forks are mostly real use** : fork/star is about 25%, far above typical new repos. Issues show teachers checking the science details line by line (bilsemc2 raised 8 items), and people asking for an English version, math, and a collaboration email, which points to real teacher users.

### 5. Differentiation

Compared with PhET, NOBOOK, and whiteboard tools, its difference is that it combines **alignment to PEP lesson by lesson × 3D scenarios × step-by-step classroom scripts × free and offline use** , delivered as an open file. This combination is what drove it: each element has been done before, but none had been put together for Chinese high school teachers, for free. Another difference is "traceable numbers" (each figure maps to a textbook page, or is labeled as a "courseware setting"), which gives teachers confidence to use it; after checking, the reviewer said it was "basically accurate" (geeyo). Note, though, that the stars were mainly driven by the story and the news accounts. The product differences shape forks and retention, not reach.

### 6. How many people have seen it so far

It is spreading in small circles and becoming common: three big Chinese AI-community accounts (AIGCLINK, Gorden Sun, GitHubDaily) shared it, along with a Zhihu long-form article, Woshipm, Sina reposts, and many resource sites. The English community had only one article, from SI Build Lab. No revenue has been made public: the courseware is permanently free, sponsorship is not open, and the monetization path is X DMs for custom courseware and factory digital twins (https://lab.tongjiang.app/about/).

### 7. Copycats

No direct copycat repo of Chinese 3D courseware has been found; the forks are straight copies. A similar form appeared the same period in atlas-of-physics (English AP physics, created the same day). These arose in parallel, not as imitations.

### 8. What to borrow

- The pipeline of "fact sheet → shared shell → each lesson fills only content → cheap models for mass production, strong models for review → multi-size screenshot self-checks + human image review" can be moved to A-Level, IB, NGSS, vocational education, medicine, and textbooks in other languages.

- Open-sourcing paid content that wasn't selling, in exchange for stars and trust, then funneling custom demand to DMs: this is a classic open-source acquisition funnel.

- Release rhythm: first tease "building publicly", then explain the whole method in one original post, and explicitly ask people to "share with teachers and star it". Chinese news accounts will then relay it.

- Choosing a single offline HTML file matched the real constraints of schools: poor networks and the need to project.

### 9. Still unclear

- How many teachers actually use it in class: no site traffic data is available; only forks and issues hint at it.

- Engagement counts for the original post and the AIGCLINK post were not captured (X pages did not show them); the publication date of the Zhihu article is unverified.

- The stargazers API returns 404, so star timing by the hour is unavailable, and each repost cannot be matched precisely to the growth at that time.

- The author, November, says they'll keep open-sourcing the "remaining courseware," but the repo hasn't moved since 10-28, so whether it gets maintained long-term is an open question. The CC BY-NC license also blocks commercial redistribution.

This article was written by AI from public data and the data we collect daily. Numbers are sourced, and inferences are labeled.