The last piece was about uploaders: new listings dried up, authors thinned out.
What about downloaders? Did people suddenly stop needing source packages?
Fourteen years of browse and download data say something simpler:
People are still looking for projects. They just got pickier — wrong package, won’t run, seller gone missing.
So 1.0 share-and-download answered: “Is there a package to grab?”
2.0 curated purchase has to answer: “Can I buy with confidence — and will it run?”
How people used Zuidaima back then
For a lot of long-time users, the path looked like this:
Search a keyword → open a detail page → check screenshots and notes → spend points → unzip and try it locally.
What the site did was blunt: someone uploads, someone downloads, the site keeps score with points.
Search could still match people to inventory, so the loop turned. We call that era share-and-download.
Uploaders left first — that was the last article.
This one stays on the downloader side: how “finding a project” changed over the years.
1. People still come to look — fewer, not gone
Logged-in views only (by year).
| Year | Views | Unique viewers | vs 2018 |
|---|---|---|---|
| 2016 | 469k | 71k | 60% |
| 2018 | 780k | 96k | 100% |
| 2020 | 360k | 57k | 46% |
| 2022 | 154k | 29k | 20% |
| 2024 | 40k | 8k | 5% |
| 2025 | 10k | 2k | 1.3% |
2018 was the peak. After that it slid — a long fade, not an overnight cliff.
By 2025, views are roughly one percent of the peak. And yet, every year, some people still log in and open projects.
What that means:
“I want a system I can learn from or adapt” never fully died.
What changed: fewer people arrive through long-tail search; training and delivery seasons still get busier; the overall pool got thinner.
2. People who download: smaller pool, same habit
Rule: one download count per user per project. No talk of payment. No RMB figures.
Lifetime scale over fourteen years:
- About 695k deduplicated downloads
- About 287k users who downloaded at least once
| Year | Deduped downloads | Downloaders | vs 2018 | Downloads ÷ views |
|---|---|---|---|---|
| 2016 | 88k | 52k | 65% | 18.7% |
| 2018 | 134k | 57k | 100% | 17.3% |
| 2020 | 58k | 27k | 43% | 16.2% |
| 2022 | 27k | 14k | 20% | 17.6% |
| 2024 | 6.9k | 3.4k | 5% | 17.2% |
| 2025 | 1.7k | 1.0k | 1.2% | 16.2% |
Here’s the interesting bit: views fell, downloads fell with them — but the share of “I viewed, then I downloaded” stayed roughly 16%–19% for years.
In plain language:
It isn’t that everyone who arrives only window-shops.
Fewer people arrive, shelves got thinner — and once someone is in, many still download.
For downloaders near the end of 1.0, the lived loop often looked like:
Open the site → inventory feels stale → grab a zip → won’t run or docs are thin → seller unreachable → next time, try GitHub or an AI chat instead.
Local IDEs and AI can help with environments now.
What’s expensive is the wrong package, messy versions, and nobody accountable.
3. What were they actually looking for?
Categories and keywords show a steady taste: full systems you can practice on and change — not another static page.
Categories (live share mix, approximate):
| Category | Live shares |
|---|---|
| Web site development | 1,803 |
| Front-end | 1,469 |
| Syntax / algorithms | 279 |
| Desktop clients | 194 |
| Text / file processing | 192 |
| Mobile clients | 126 |
Deduplicated downloads by keyword hit (title or body; overlaps allowed):
| Keyword | Deduped downloads | Downloaders |
|---|---|---|
| MyBatis | 137k | 55k |
| SSM | 71k | 31k |
| Permissions | 68k | 38k |
| SpringBoot | 37k | 17k |
| Shop / mall | 33k | 20k |
| Blog | 29k | 19k |
| Vue | 7.5k | 4.9k |
| Mini-program | 6.3k | 4.7k |
Shops, blogs, permissions, SSM, SpringBoot — packages you open and touch real business logic.
In the AI era, spinning up another page is cheap.
What downloaders will still spend time on (and later, money on) is only:
Reviewed packages with clear docs, a fixed version, and a platform to talk to when something breaks.
4. 1.0, from a downloader’s point of view
Read together with the sharer piece, 1.0 looks like this on the demand side:
- Share-and-download proved itself. People uploaded, browsed, and downloaded; points made matching work. Nearly seven hundred thousand deduped downloads — not an empty site.
- Demand did not vanish overnight. Views faded on a slope, not a cliff; among people who arrived, roughly one in five still downloaded.
- The pain moved. Before, the scarce thing was “is there a package?” Now it’s “can I trust this package?”
- That was also 1.0’s ceiling. It was good at circulation — see it, take it. It was weak at memory — how many people actually got it running, and who owns the problem when it fails.
So: it’s not “source sites are dead.”
It’s that “grab any points zip” is a dead road.
5. 2.0 for downloaders (buyers): what you get next
We call 2.0 curated purchase. Not another points shelf — a product for “safe to buy, likely to run, fewer disputes.”
The classic library stays: old shares remain browsable and redeemable from history.
New work, and anything you want to buy with confidence, goes Premium. Two tracks. Overlay, not a wipe.
From a downloader’s seat, the change looks like this:
| Before (share-and-download) | After (curated purchase) |
|---|---|
| Lots of zips, priced in points | Listing gates: startup notes, run screenshots — fail the bar, stay off the shelf |
| After download, you’re alone | Checkout locks the version — you get that build |
| “Won’t run” → forum threads | Detail pages show run evidence; limited environment support |
| Private chat with the seller | Official checkout — guarantees and support go through the platform |
| Unclear origin / license | Detail pages show original vs reprint sources and license |
One line for people still hunting for source:
Someone else spends three days prompting from scratch; you buy a curated build and adapt the checklist.
Our tagline stays the same:
Zuidaima: Curated source. Less time, less effort.
Official checkout · Provable runs · Runs locally
Boundary we put on product pages too: materials are for learning, research, and secondary development — not sold as thesis or defense packs, and not meant to be submitted as coursework verbatim.
Closing
Uploaders stopping does not mean downloaders gave up.
The data says people still want full systems — they just fear buying the wrong one.
1.0 made “there’s something to download” work.
2.0 has to make “download with confidence, and get it running” work.
Previous: why authors stopped uploading — publish/en-01-sharer-niubi-to-trust.md (EN draft; Chinese companion linked above).
Method note
- Views: logged-in users. Downloads: user × project dedupe.
- No top-ups, payouts, or RMB revenue in this piece.
- Shares behavioral numbers with the companion article; does not share revenue numbers.
Top comments (0)