# One floor plan, one hour, 3,600 stars: floorplan-3d isn't a hit because of its features · GitHub open-source project

> Source: https://stringers.io/en/github/gh_wy51ai__floorplan-3d.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

> It rode the launch buzz of Opus 5.5, which was marketed on understanding 3D space, plus the author's demo video with 370k views. Then Chinese recommendation accounts with 136k followers repackaged it as a free renovation tool, and recommendation networks, remixes and reposts in Hong Kong and Taiwan kept pushing it for several more rounds.

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

wy51ai/floorplan-3d · wy51ai

# One floor plan, one hour, 3,600 stars: floorplan-3d isn't a hit because of its features

**What it is** A home-design mini-tool that is just one web page: place furniture and remove walls on a floor plan, estimate flooring costs, and switch to 3D with one click to walk through it. Open it and use it, free and open-source.

**Why it took off** It rode the hype around the Opus 5.5 launch, which pitched "understands 3D space," plus the author's video with 370k views. Then a Chinese recommendation account with 136k followers reframed it as a free home-design gadget, and a matrix of recommendation accounts, remixes and reposts from Hong Kong and Taiwan kept pushing it for several rounds.

**B** **How you could build on it** Don't make yet another identical demo. Fill the gap it leaves: automatically turn a sales-office floor-plan photo into an editable 2D/3D layout of your own home, keep it playable online, MIT-licensed. For developers, make a video; for ordinary people, hand it to Chinese recommendation accounts and Xiaohongshu.

Grade B: the gap is clear (the most common issue and reply is "how do I swap in my own home"), but several projects like ZWD11 are already filling it, floor-plan recognition itself is hard, and this wave of stars owes much to the new-model hype plus the author's and recommendation accounts' traffic, so late entrants may not get the same reach. Word of mouth

Data (updated daily) · Repo created 2026-09-29 · 3,623 stars · 7 days +344

GitHub · stars

Data 2026-10-09 → 2026-11-08 (31 days, 30 observations). Points before 10-28 come from deeply / archive backfill, not every day has one

On 2026 Sep 28, WY (@akokoi1), an AI creator on X with 20k followers, posted a video: a plain floor plan fed to the newly released Claude Opus 5.5 turned into a web page within an hour where you could drag furniture, remove walls, and switch to 3D to walk through. He only created the GitHub repo wy51ai/floorplan-3d the next day. As of 11-06, this project, with just one `index.html`, had 3,600 stars and 750 forks, and the author hasn't changed the code since Oct 2. It gathered that many stars because it landed just as a new model launched, a Chinese recommendation account framed it again, and it was a single file anyone could fork. Its biggest gap, "can't import your own floor plan," is still open, and you can fill it.

### One video, and a second post with a new angle

The author's original post said: "Opus 5.5's 3D modeling and comprehension are scary good. I built this small project in an hour, and I only supplied one floor plan image..." It has 373k views, 644 likes, and 920 bookmarks so far. He also posted the original floor-plan image and the full prompt below the post ("Based on this floor plan, write a single-file HTML home-design tool: 1, 2D plan: restore the walls using the dimensions marked on the drawing (mm)..."), and as a result the post was collected into several "Opus 5.5 prompt example" compilations.

The timing was lucky. In Sep 25, Andon Labs posted that Opus 5.5 ranked first on Blueprint-Bench 2 (where AI turns photos of an interior into a floor plan). "Understanding 3D space" was the pitch everyone was talking about that week. Turning a floor plan into a house you can walk through is the most direct demo there is.

What really pushed it out of the developer circle was the second post. A Chinese recommendation account with 29,136k followers, "Namei Knowledge Base", posted: "Recommending a pretty addictive open-source project: Floorplan 3D. I haven't even started renovating and I'm already playing it like a game... In real life I don't dare buy furniture on a whim, so I just drag things around my virtual house first. No install needed, just open the webpage and play." It got 113k views and 1,951 saves. Same thing, different pitch: the author said "AI wrote 2000 lines in an hour", while Namei said "move your furniture around a virtual house before you renovate". The audience shifted from programmers to ordinary people about to renovate or remodel a rental.

### What it actually does

Open wy51ai.github.io/floorplan-3d and use it right away. No sign-up, data is stored locally in the browser:

- **2D** : rebuild a three-bedroom apartment at the original dimensions (mm), drag in pieces from 60 furniture types, rotate and resize them, snap to walls. You can measure, delete non-load-bearing walls, and load-bearing walls are marked separately.

- **3D** : one click switches to a Three.js scene: bird's-eye, top-down, or WASD first-person walkthrough, doors open and close, with a sunlight slider and night lighting. It looks great as a video.

- **Cost estimate** : swap the flooring for each room (hardwood, tile, marble…) and add 5% waste by area for a rough cost.

The weakness is obvious too: the floor plan is hardcoded. The README says "Room data lives in `index.html`: `ROOMS`, `WALLS`…" To use your own home, you have to edit the code yourself or have an AI do it. The top reply under Namei's post was "If it supported importing floor plans, it would be unbeatable." The repo's #2 issue was titled "Hey, how do I change the floor plan?" and was followed by "Please add importing a floor plan from an image" (#3) and "It would be even better if I could edit my own floor plan" (#13).

### Where the stars came from

| Date | star | fork | What happened |
|---|---|---|---|
| 09-29 | +472 (day one) | — | Author's post + Namei's post (Trendshift) |
| 10-09 | 1,842 | 340 | Reposts on Weibo, Threads, and Taiwan/HK Instagram; a Traditional Chinese PR was merged |
| 10-21 | 2,659(+167) | 540(+64) | Source not found; over 64 forks in a day, looks like people learning how to adapt it |
| 10-29 | 3,183(+153) | 649 | Source not found |
| 11-05 | 3,552(+168) | 740 | Source not found |
| 11-06 | 3,582 | 748 | GitHub daily snapshot; live count at 3,605 that afternoon |

Channels I could confirm:

1. **The author's own posts** , plus the prompt got picked up in roundup collections. A main driver on day one.

2. **A matrix of Chinese AI recommendation accounts.** After Namei, Denzii on X (10-03) and Weibo AI creator Zhang Daiyue (10-13) posted nearly identical lines: "Haven't started renovating, so I'm playing with the house as a game first" and "No install, just open the web page and play." I can't tell whether it's one team or people copying each other.

3. **Remixes bring traffic back.** On 10-17, @jinglian recommended someone else's version built on it, "CAD import + Blender rendering." It got 8,779 views, and daily stars picked up again.

4. **Taiwan, HK, and long tail** : Taiwanese users submitted a Traditional Chinese PR, and Cantonese creators recommended it on Threads and Instagram. A CSDN article, "Free open source vs. Kujiale and friends," claimed the search term "free alternative to Kujiale."

Where it did not go: Hacker News, Reddit, and English tech press never covered it, and there was no paid promotion. There are three small spikes of over 150 extra daily stars that I couldn't match to any post on X or the web. My guess is Xiaohongshu, WeChat Official Accounts, or Telegram channels (inference).

The fork ratio is worth a separate note: 748 / 3,582 is about 21%, versus a median of about 11% for new repos in the same period. The reason is the weakness above: to use your own floor plan, you have to fork and modify it. In Sep 30, a small renovation company in Hanoi, Vietnam also opened an issue saying they wanted to turn it into an internal tool for sales staff (import a floor plan image, trace walls, place furniture, export PNG for clients) and asked whether it could get an open-source license. The repo had no license at the time; it's MIT now.

### Who's competing for this spot

| Project | Form | Scale | How it differs |
|---|---|---|---|
| Kujiale | Commercial web / app | Claims 65M+ registered users | Requires sign-up, aimed at professional drawings; floorplan-3d is free with no sign-up, but only one floor plan |
| Planner 5D, RoomSketcher, etc. | Commercial, freemium / subscription | Million-level | Can draw any floor plan, paid |
| Sweet Home 3D | Open-source desktop software, 2006 | Long-established | Needs installation, dated UI |
| blueprint3d, react-planner | Open-source library | Accumulated 2,000 / 1,500 stars over ten years | Need to build it yourself; floorplan-3d passed both in 38 days |
| openPlan3D, house-planner | Open source | 272 / 372 stars | Freely drawable floor plans and more features, but it didn't catch the wave |
| ZWD11/house-plan-designer | Agent Skill | 371 stars, rising from 10-20 | "Give an agent a floor plan image plus one sentence, and it generates a designer for your own home." That covers the floor-plan-swapping gap. |
| oxomkbk/Furnish | Open source | 74 stars | Nearly the same form, with 12 built-in floor plans |
| RobertJBye simple-floor-planner | English web pages | X post with 75 likes | Also an AI-written floor plan tool, released half a month earlier, but nobody amplified it |

On features, it loses to all of them. It wins because three things came together: the hype around a new model, a finished web page you can open and use right away, and a Chinese renovation story. The English equivalent, RobertJBye, had been out for half a month and got only 75 likes. What it lacked was someone to reframe it and push it out.

### If it were you

**One-line take:** Don't build another "AI-wrote-it-in-an-hour floor plan tool." This demo has already been done. Build the part it hasn't: upload a floor plan image (the kind a sales office hands you), auto-detect walls, doors, windows, and rooms, and generate your own home that you can furnish and walk through in 3D. This is the most requested feature in the issues and replies, and no hit has emerged in the open-source world yet.

**Who it suits:** developers who can write front-end code with AI and are willing to tackle the hard part (turning floor plan images into walls); ideally with some followers on X's Chinese-language scene, Xiaohongshu, or a renovation community, or with connections to recommendation accounts.

**Who it doesn't suit:** people who just want to copy a single-file demo. There are already over 700 forks, and plenty of small similar repos; another one would go unnoticed. It's also a bad fit for anyone without a distribution channel: more than half of this project's stars came from the author and recommendation accounts with 20k and 136k followers.

**How to do it:**

1. Keep its best form: a single file or pure static page, playable on GitHub Pages, MIT, no sign-up.

2. Do one thing at the core: upload a floor plan → detect walls and dimensions → output an editable 2D/3D. If recognition is wrong, users must be able to drag walls to fix it manually. If this step isn't done well, users won't stay. Someone has already posted a fork with floor plan recognition in issue #14. Try it first and see where it breaks.

3. You can also take floorplan-3d's furniture library and 3D effects (MIT allows it) rather than building from scratch.

**How to get users:** copy its two pitches. For developers, post a video of "floor plan image → your own home in 3D," and publish the prompts or the recognition pipeline. For ordinary people, pitch through Chinese recommendation accounts and Xiaohongshu: "Drag the floor plan from the sales office in, and move furniture around your home first." Leave a comment on floorplan-3d's issues #2, #3, and #13 (the people asking questions there are your first users). It hasn't been to English-speaking communities or Show HN, so you could post an English version separately.

**When to move:** The wave of "AI makes 3D floor plans" that Opus 5.5 kicked off probably has a month or two of heat left, and ZWD11 and issue #14 are already doing the same thing. If you ship and launch before year-end, you should still be in time.

**Signs you've misread it:** If something like ZWD11, which has an agent generate your house designer on the spot, gets to thousands of stars first, it means people want agent skills, not finished web pages, so switch to building a skill. If recognition accuracy stays low after you ship and the comments are full of "my floor plan won't recognize," then this step is harder than it looks, and the single-file demo approach no longer holds up.

[GitHub repo ↗](https://github.com/wy51ai/floorplan-3d)

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

Competitors

| Name | Where | Launched | Scale | How it differs from this one |
|---|---|---|---|---|
| Kujiale | Web | Around 2013 | Claims 65M+ registered users | Requires sign-up, aimed at commercial design and rendering; it is free, needs no sign-up, and runs locally, but the floor plans are hardcoded and the features are shallow |
| Planner 5D / Floorplanner / RoomSketcher / Space Designer 3D | Web | 2010 era | Millions (Space Designer 3D claims 6M) | Freemium or subscription; it is fully free and open source, but you can't freely draw floor plans |
| Sweet Home 3D | Open source | 2006 | An established free software | Needs installation, dated UI; it runs in the web page and looks modern |
| pascalorg/editor | Open source | 2025-10 | 25,363 stars | A full 3D architectural editor (BIM, MCP); it is a demo-level single file aimed at homeowners renovating |
| furnishup/blueprint3d, cvdlab/react-planner | Open source | 2015 / 2016 | 1,999 / 1,490 stars | Libraries or frameworks you have to build on yourself; it is a finished web page and passed both in 38 days |
| laanlabs/openPlan3D, egmalt/house-planner | Open source | 2026-02 / 2026-10 | 272 / 372 stars | Freely drawable floor plans and more features, but it didn't catch the wave |
| RobertJBye simple-floor-planner | Web | 2026-09-10 | X post with 75 likes | Also an AI-written English floor plan tool, with no Chinese recommendation accounts to amplify it |

Where users come from

- **Verified** **Author's X video post (Opus 5.5 capability demo)** : @akokoi1 posted 2026-09-28, with 644 likes, 920 bookmarks, 373k views. The reply thread had the original image and full prompt, and the post was picked up by prompt collections.

- **Verified** **Big Chinese AI-tools recommendation account** : NaMi Knowledge Base @fhwofjow51260 (136k followers) posted 09-29, with 1,565 likes, 1,951 bookmarks, 113k views

- **Verified** **Copy-matrix spread (X and Weibo)** : Denzii 10-03 and Weibo account Zhang Daiyue 10-13 used copy nearly identical, line by line, to NaMi's post

- **Verified** **Fork-developer shout-out** : @jinglian 10-17 recommended @Summerz3002's CAD-plus-Blender fork, 8,779 views

- **Verified** **Hong Kong and Taiwan traditional-Chinese circles** : 10-01 merged a traditional-Chinese PR; @cego.seeatgo 10-06 recommended it in Cantonese on Threads and IG

- **Verified** **Aggregators and content-farm SEO** : CSDN 10-25 "Free open source vs Kujiale and co.", jamecling 10-29, Trendshift, GitHub Signals

- Inferred **Sources of the 10-21, 10-28 and 11-05 spikes (likely Xiaohongshu, WeChat Official Accounts, Telegram tutorials)** : about 160,10-21 forks added that day, +64 in one day, but no matching posts were found

## wy51ai/floorplan-3d: why it took off

**In one line** : a pure front-end floor-plan and interior design tool in a single `index.html` (about 185KB, 2000 lines). In the 2D plan you drag furniture, remove non-load-bearing walls, measure dimensions and swap floor materials, with an automatic cost estimate. One click switches to a 3D Three.js view, either bird's-eye or first-person walkthrough. MIT license, hosted online via GitHub Pages (wy51ai.github.io/floorplan-3d).

**Conclusion** : it did not take off on product depth. It was an "Opus 5.5 capability demo" that ran into a network of Chinese AI recommendation accounts. Around 2026-09-25, Claude Opus 5.5 was getting attention for understanding 3D physical space (Andon Labs posted that it ranked first on Blueprint-Bench 2 for drawing floor plans). The author posted a video on X around 9-28 saying "one floor plan, one hour," with 370k views. The next day, 娜美知识库, a Chinese recommendation account with 136k followers, packaged it as a "renovation tool, open the web page and play," and it spread again. Weibo, Threads, Hong Kong Instagram, CSDN and assorted aggregators followed, carrying it from developer circles to ordinary people planning a renovation. Its real differentiator is "opens in the browser, no sign-up, free, and it's an AI template you can modify," not the features themselves. Feature-wise it is far behind 酷家乐 and Planner 5D, and it can't even handle custom floor plans.

---

### 1. Data

| Timeline | star | fork | Source |
|---|---|---|---|
| Repo created 2026-09-29 (earliest commit 09-29 01:50 UTC) | — | — | GitHub API commit history |
| Day one | +472 | — | Trendshift chart (trendshift.io/repositories/268336) |
| 10-09 | 1,842 | 340 | our GitHub repo_day |
| 10-16 | 2,300 | 436 | Same as above |
| 10-21 | 2,659(+167) | 540(+64) | Same as above |
| 10-28 / 10-29 | 3,030(+162)/ 3,183(+153) | 618 / 649 | Same as above |
| 11-05 | 3,552(+168) | 740 | Same as above |
| 11-06 | 3,582 | 748 | Same source; gain_7d 358, prev_7d 472 |

- The fork-to-star ratio is about 21% (748/3582), roughly double the median for comparably new repos (about 11%, from the github repo SCHEMA notes). That fits its form: it's a single file, and you can only swap in your own layout by editing `ROOMS/WALLS`, so people who want to use it have to fork or download.

- After Sep 29 there was almost no new code: the last push was 10-02, the author made 9 commits in total, and only one external PR was merged, a traditional-Chinese translation (#7), itself written with Claude Opus 5.5. Since then stars have kept climbing from external shares, in a pattern of a long tail plus occasional spikes.

- Author WY (GitHub wy51ai, X @akokoi1, 20k followers on X). The same account's edulab (education Agent skills) has 1,762 stars. The author is an AI-tutorial blogger who handles their own distribution.

### 2. Competitors

| Product | Where | Launched | Scale | How it differs from this one |
|---|---|---|---|---|
| Kujiale (Qunhe Technology) | Web / App | Around 2013 | Official site claims 65M+ registered users and 40k+ corporate partners | Kujiale requires sign-up and leans commercial design, producing rendered visuals; floorplan-3d needs no sign-up and runs entirely locally, but has only one hard-coded layout |
| Planner 5D / Floorplanner / RoomSketcher / Space Designer 3D (claiming 6M users) | Web / iOS / Android | 2010 era | Million-level | All subscription or freemium; it is free, open source and account-free, but the feature gap is large |
| Sweet Home 3D | Open-source desktop / web tools | 2006 | An established free software | They need software installed and have dated interfaces; it opens and runs immediately with a modern look |
| pascalorg/editor | Open source | 2025-10 | 25,363 stars | Proper 3D architecture editors (BIM, MCP); it's a demo-grade single file |
| furnishup/blueprint3d, cvdlab/react-planner | Open source | 2015 / 2016 | 1,999 stars / 1,490 stars (both built up over a decade) | Both are libraries or frameworks you must assemble yourself; it's a finished web app, and it passed both in star count within 38 days |
| laanlabs/openPlan3D, egmalt/house-planner | Open source | 2026-02 / 2026-10 | 272 stars / 372 stars | They allow free-form layouts and more complete features, but didn't catch a wave of spread |
| ZWD11/house-plan-designer | Open source (Agent Skill) | 2026-10-06 | 371 stars (10-20 rising) | Followers: "one floor plan plus one sentence, and an Agent generates a similar single-file designer", which fills the gap that floorplan-3d can't swap layouts |
| Home Assistant 3D floor-plan cards (neonplan3d 1,056 stars, NextFloor, ha-diorama, etc.) | Open source | 2026-09–10 | Hundreds to thousands of stars | The scenario is a smart-home dashboard, not renovation; same wave of "Three.js + AI floor plans" hype |
| RobertJBye's simple-floor-planner | Web | 2026-09-10 | X post with 75 likes | An English-language equivalent of "write your own floor-plan tool with AI and skip the $29/month fee"; it didn't get the amplification from Chinese recommendation accounts |

Conclusion: there are many similar tools. Commercial ones include Kujiale and Planner 5D; open-source ones include decade-old libraries and AI-generated projects from the same period. It has no feature moat. It wins on timing, format and Chinese-language spread.

### 3. Where users came from (growth tactics)

1. **Author's own X video post (confirmed)** : @akokoi1 posted on 2026-09-28: "Here's another practical use case. Opus 5.5's 3D modeling and comprehension are scarily good. I built this small project in one hour. I only provided one floor plan..." Figures: 644 likes, 69 reposts, 920 bookmarks, 373k views (fxtwitter API). He put the original image and the full prompt in the reply thread, so the post was also collected as an "Opus 5.5 prompt example", for instance on jasonzhu.ai's prompt page and in awesome-claude-opus-5.5 (110 stars). Note: this post went up before the repo was created, which happened on 09-29 02:57 UTC. The viral demo came first, and the open-source repo was added after.

2. **Chinese AI-tools recommendation account (confirmed, main driver on day one)** : NaMi Knowledge Base @fhwofjow51260 (136k followers) posted at 09-29 12:01 UTC: "Recommending a rather addictive open-source project: Floorplan 3D. Before I've even started renovating, I'm already playing with it like a game". 1,565 likes, 314 reposts, 1,951 bookmarks, 113k views. It repackaged the repo from an "AI capability demo" into a consumer story: move furniture around a virtual house before you renovate.

3. **Copy-matrix spread (existence confirmed; whether it's one team is unverified)** : Denzii @denziideng (10-313k followers, 5,060 views), Weibo AI blogger Zhang Daiyue (10-13, location listed as "USA"), and NaMi's copy are nearly identical, line by line: "Buying a house and renovating hasn't started, but I'm already playing with my home like a game", "Big furniture items you don't dare buy randomly in real life can be moved around in a virtual house first", "No installation, just open the webpage and play". This is a common rewrite or matrix distribution pattern among Chinese AI recommendation accounts.

4. **Second wave from fork developers (confirmed)** : @jinglian (10k followers) posted on 10-17, recommending @Summerz3002's "CAD import plus Blender rendering" fork built on it (8,779 views). The post also responded to complaints that "importing your own floor plan is unstable", positioning it as "a template plus an idea". This roughly lines up with the daily-star recovery after 10-16.

5. **Hong Kong and Taiwan traditional-Chinese circles (confirmed)** : 10-01 merged a traditional-Chinese PR (submitted by a Taiwanese user). @cego.seeatgo 10-06 recommended it in Cantonese on Threads and Instagram (973 views on Threads). The Trendshift page description is also in Cantonese.

6. **Long-tail SEO from aggregators and content farms (confirmed)** : CSDN gitblog 10-25 "Free open source vs Kujiale and co.", jamecling.com 10-29, Trendshift, GitHub Signals, githubtree, plus an auto-narrated YouTube video (153 views). Each is small, but together they claim the search term "free alternative to Kujiale".

7. **Three small daily-star spikes of 10-21, 10-28/29 and 11-05 (150–170 per day): sources not found** . On 10-21 there were over 64 more forks in a single day than usual, which looks like a "teach you how to use it" tutorial or community share. Neither our X corpus nor Jina search turned up matching posts. Xiaohongshu and Douyin were not checked (the domestic-platform TikHub endpoints were not used).

8. No coverage on HN, Reddit or English-language tech media, and no paid promotion.

### 4. Why it took off

- **时机** :Opus 5.5 9 月下旬发布,「3D 和空间理解」是它的主打卖点之一(Andon Labs 9-25 说它在 Blueprint-Bench 2 画户型图排第一)。大家都在找能直观展示这一点的演示,「一张户型图、一小时、能走进去的 3D 房子」正好命中。

- **Crossing audiences** : developers saw "AI wrote 2000 lines in an hour", while everyday people saw "try furniture placement before renovating, no software to install, free". Recommendation accounts amplified the second framing, widening the audience from programmers to people preparing to renovate or remodel rentals. Renovation anxiety on Chinese internet is a large pool.

- **Zero-friction trial** : the GitHub Pages version runs as soon as you open it. First-person walkthroughs, opening doors, sunlight and night views make for striking short videos, which suits social-platform spread.

- **A "modifiable template" drives forks and remixes** : single file, MIT, data at the top of the file, so anyone can have AI change it to their own layout. That produced the CAD-plus-Blender fork, automatic floor-plan recognition (#14), and an Agent Skill version (ZWD11), and the fork ratio is unusually high. Each remixer in turn sends a new wave of traffic back to the original repo.

- **The author has distribution capacity** : a 20k-follower X account, which earlier produced a repo with 1.7k stars. They know to publish the prompt alongside the post, which turns it into a "copyable example" that prompt collections pick up.

### 5. Differentiation, and whether it's the reason it took off

Compared with Kujiale and Planner 5D: free, no sign-up, data stays local, open source and modifiable. Compared with older open-source libraries: it's a finished web app rather than a framework you assemble yourself, and visually far more polished. Compared with AI-generated floor-plan tools from the same period (RobertJBye, the Home Assistant cards): a Chinese interface, Chinese home conventions (load-bearing walls, 1:60/1:100, mm, a 5% waste allowance on each floor material when estimating cost), plus amplification from Chinese recommendation accounts.

Among these differences, what actually drove growth was the combination of "open and play, a Chinese renovation story, and an AI capability demo", not the features. The most common user feedback was "how do I change it to my own layout?" (#2, #3, #13, #14), which suggests it isn't very usable as a tool yet. Stars mostly reflect bookmarks and likes, not actual use.

### 6. Visibility and bandwagoning

- Visibility: **circulating in small circles** . The Chinese AI-tools recommendation scene and the Chinese-language X community have already passed it around (the author's post at 370k views, NaMi's post at 110k views), and it has also appeared on Weibo, CSDN and Hong Kong/Taiwan social platforms. But there's no English media or HN coverage and no major account has done an in-depth piece. It is open source and free, so no revenue is involved. The card's claim that "using it for homeowners' renovation changes earns ¥300–800 per job" has no original source I could find in this pass (unverified).

- Copycats: ZWD11/house-plan-designer (371 stars, Agent Skill version, up 10-20 since it launched), oxomkbk/Furnish (74 stars, 10-03 built, "Web interior design tool, 2D with three.js walkthrough, 12 built-in floor plans", nearly the same format), Falix23/floorplan-3d-agent-skill, hzdlive/floorplan-3d-wy51ai, @Summerz3002's CAD + Blender version, and hazhirrashidi/floorplan-3d (same name, English version, 1 stars).

### 7. Takeaways

- **Publish a capability demo first, open-source later** : ride the buzz of the new model launch, pick a scenario ordinary people instantly understand (home renovation), turn it into a video, and publish the prompt publicly.

- **Single file, playable online, MIT** , so recommendation accounts can offer "try it in one click" and remix creators can "modify it in one click." Forks and remixes feed traffic back.

- **Change the pitch** : same product, but for developers say "an AI wrote it in an hour," and for the general public say "move the furniture around a virtual house before you renovate." Hand the second version to Chinese recommendation account networks to spread.

- **The gap it left open is the opportunity** : importing your own floor plan, recognizing floor plan images. ZWD11 and others are already building this; an open-source version of Kujiale-style "upload a floor plan, get a 3D model" has no hit yet.

### 8. Uncertain

- The specific sources of the three peaks 10-21, 10-28/29, 11-05 (suspected to come from RedNote, Douyin, WeChat Official Accounts or Telegram channels; not verified).

- Whether the identically-worded accounts Namei, Denzii and Zhang Dayue belong to one team's network or are copying each other, and whether they have any partnership with the author.

- Real usage: GitHub Pages has no public access data, and the claim of "300–800 yuan per order" has no source.

- Star history breakdown: the stargazer API can't return data with our token, so the first-day 472 stars can only be cited from Trendshift.

### Sources

- Repo, README, commit history, issues, forks: [github.com/wy51ai/floorplan-3d](https://github.com/wy51ai/floorplan-3d) (GitHub API, 2026-11-06)

- Daily stars and forks: agent-im GitHub repo, repo_day and repo_profile (2026-10-09 to 11-06)

- Trendshift:[trendshift.io/repositories/268336](https://trendshift.io/repositories/268336)

- Author's tweet: [x.com/akokoi1/status/2104520072014508316](https://x.com/akokoi1/status/2104520072014508316) (data from api.fxtwitter.com)

- Namei knowledge base: [x.com/fhwofjow51260/status/2104904378822095203](https://x.com/fhwofjow51260/status/2104904378822095203)

- Denzii:[x.com/denziideng/status/2106174629790519797](https://x.com/denziideng/status/2106174629790519797)

- jinglian:[x.com/jinglian/status/2111293280466133029](https://x.com/jinglian/status/2111293280466133029);Summerz3002:[x.com/Summerz3002/status/2110981167193403432](https://x.com/Summerz3002/status/2110981167193403432)

- Weibo Zhang Dayue (Sina repost): [www.sina.cn/news/detail/5353633941294339.html](https://www.sina.cn/news/detail/5353633941294339.html)

- Threads:[www.threads.com/@cego.seeatgo/post/DeJtCQ6n-iT](https://www.threads.com/@cego.seeatgo/post/DeJtCQ6n-iT)

- CSDN:[blog.csdn.net/gitblog_00169/article/details/169051511](https://blog.csdn.net/gitblog_00169/article/details/169051511);jamecling:[www.jamecling.com/archives/8631](https://www.jamecling.com/archives/8631)

- Andon Labs Blueprint-Bench 2:[x.com/andonlabs/status/2103537821680304282](https://x.com/andonlabs/status/2103537821680304282)

- RobertJBye:[x.com/RobertJBye/status/2098149345573105731](https://x.com/RobertJBye/status/2098149345573105731)

- Competitor star counts: GitHub API (2026-11-06); user numbers for Kujiale and Space Designer 3D come from their official sites' self-reported claims.

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