Create and share 3D architectural projects. GitHub: https://xitter.cf/t.co/zVdeBKum3l

USA
Joined April 2024
Awesome work, good job Section Sweep feels the most premium Now imagine it on real blueprints, real building codes, architect-grade AI That's what we're building at @pascal, open source
7 creative ideas to turn a flat drawing into 3D as you scroll. Which one feels the most premium to you? 1- Section Sweep: a scan turns the ink into real materials and you walk in through the cut 2- Card Model: scroll down to the tiny figure and the white card model becomes the real house 3- Build Time-Lapse: walk the site while the house goes up week by week 4- Open Section: the cut opens into depth and the camera visits each floor 5- Sunlit Elevation: the hatching on the drawing is the real shadow, so when it turns to brick not one shadow moves 6- Exploded House: the layers land one by one in building order, then you walk in 7- Schedule Build: scroll the construction schedule and a crane builds the house I’m sharing everything (starter prompts and repo). Stay tuned. All built with Opus 5.5.
1
3
43
2,329
"What turned out to be needed wasn't architectural knowledge, it was the ability to translate my tastes into words" This is exactly why we open-sourced @pascal Editor Describe the house, Claude builds it, you walk through it Thanks for the write-up
집 구조를 클로드가 같이 만져준다는 별 2.5만 개 물건 인테리어 업체 부르기 전에 클로드한테 방 구조부터 맡겨봤다는 사람이 나올 법한 물건을 발견했다. 별이 약 2.5만 개(25.1k, 지금 시점 기준) 붙은 저장소다. 이름은 Pascal Editor. 먼저 불편한 점부터 말하면, 집 구조를 머릿속으로만 그리다가 막히는 순간이 있잖아. 벽을 어디까지 세울지, 방을 어떻게 나눌지. 종이에 그리면 지우개 가루만 쌓이고, 전문 도면 프로그램은 켜는 순간부터 포기하게 된다. 이걸 브라우저에서 3D로 집을 짓고 만지게 해주는 무료 공개 프로그램이 이 물건이다. 다만 시공용 전문 도면을 뽑아주는 건 아니고, 벽이나 바닥, 지붕 같은 3D 장면을 만지는 쪽이다. 한 줄 정의는 이거다. 오픈소스(누구나 갖다 쓰라고 코드를 공개한 것) 3D 건축 편집기. 내 컴퓨터 안에서 돌아가고, 프로젝트도 내 컴퓨터의 로컬 데이터베이스(SQLite, ~/.pascal/data/pascal.db)에 저장된다고 적혀 있다. 여기서 내 눈이 멈춘 부분. 이 편집기가 MCP를 지원한다. MCP는 클로드에 꽂는 확장 커넥터, 콘센트 멀티탭 같은 거다. 꽂으면 클로드가 이 편집기의 장면을 직접 조작할 수 있다는 뜻이다. README에는 AI 에이전트(알아서 일해주는 AI)를 MCP로 연결한다고 되어 있고, 설치하면 인증이 걸린 MCP 서비스가 뒤에서 같이 켜진다고 한다. 그래서 상상해봤다. 예전엔 이랬다. 마음에 드는 집 구조가 떠올라도 도면 툴 사용법부터 유튜브로 배우고, 마우스로 벽 하나하나 끌어다 놓다가 지쳐서 접었다. 이제는 클로드한테 방 세 개짜리 구조 잡아달라고 말 한 번 던지면, 편집기 화면에 집이 올라오고 나는 그걸 돌려보며 여기 벽 옮겨달라고 말만 하면 되는 그림이다. 직접 해본 건 아니고, README가 말하는 구조대로 그려본 장면이다. 데모 페이지(Pascal Next)도 구경했는데, 실제 집 하나를 방부터 마감재, 벽 속 구조까지 재현해뒀다. 건물을 분해해서 펼쳐보기, 단면으로 자르기, 눈높이로 걸어 들어가기, 문 열어보기가 된다. 단 이 화면은 다음 버전 방향을 보여주는 미리보기고, 아직 오픈소스로 공개된 편집기에는 들어있지 않다고 본인들이 밝혀놨다. 이건 구분해서 봐야 한다. 쓰는 길은 두 가지다. 브라우저로 열거나, 터미널 같은 명령 입력창으로 돌리거나. Node.js만 깔려 있으면 저장소를 통째로 받지 않아도 npx @pascal-app/cli editor 한 줄로 내 컴퓨터에 설치된다고 한다. 건축은 전문가 영역이라 AI가 못 끼는 줄 알았다. 근데 도면을 AI가 말로 만져줄 수 있게 만든 순간, 필요했던 건 건축 지식이 아니라 내 취향을 말로 옮기는 능력이더라. 도면은 못 그려도 이사 갈 집에 가구 어떻게 놓을지 밤마다 머릿속으로 돌려본 사람 있을 텐데, 그걸 말로만 해서 눈으로 볼 수 있다면 써볼 건가? 난 일단 거실 소파 위치부터 옮겨볼 것 같다. 출처: GitHub — pascalorg/editor
2
1
6
343
Most detailed third-party write-up of @pascal Editor so far Covers what's open source, what's hosted, what MCP does and doesn't do, data flow, licensing If you're evaluating it for client work, start here
装修方案别只在脑子里摆家具,先把房间搭出来看看 🏠 Pascal Editor 是一个基于 React Three Fiber 和 WebGPU 的开源 3D 建筑编辑器。按照官方入门文档,可以画出封闭房间、放置门窗和家具,再切换 2D 平面、3D 或分屏视图,查看空间与布局。想讨论室内方案、调整家具位置,或者把建筑想法做成可视化模型,它提供的是一块可以继续编辑的画布。GitHub 它还支持通过 MCP 连接 AI 客户端,让助手参与模型的创建、检查和修改。官方提供了 pascal-3d 与 furniture-fit 两个技能,分别用于场景工作流和家具占地尺寸判断。不过,安装技能不等于已经连上 MCP,也不等于授权 AI 修改项目;本地同一个服务应只连接一个活跃 Agent,不能默认多个助手同时操作互不影响。Pascal Editor 家具判断尤其要看清范围:furniture-fit 主要检查给定尺寸、位置和旋转角度下的平面占地,不能凭一句“放得下”,就认定高度足够、柜门能打开、家具能搬进屋,或真实模型之间完全没有碰撞。这些都需要额外的尺寸和证据。Pascal Editor README 里的 Pascal Next 演示也要和当前开源版区分。爆炸图、剖切、X-ray、漫游等展示属于托管预览,完整体验尚不属于开源 Editor 的发行内容。它可以辅助空间方案探索,但不能据此宣称替代建筑师、AutoCAD 或设计阶段的 BIM;真实施工、工程与合规用途,仍需人工审查和验证。GitHub 使用方式有托管与本地两条路。托管版创建、编辑和保存项目需要通过 Google 或邮箱登录,修改会自动保存到云端。本地 CLI 则需要 Node.js 22.13+ 和 npm,目前官方明确支持 macOS,Linux、Windows 的 CLI 尚未验证,不能把“浏览器编辑器”理解成所有电脑和手机都能直接使用。Pascal Editor 本地入口是 npx @pascal-app/cli editor:首次启动会从 GitHub 下载运行文件并校验 SHA-256,交互式启动成功后还会尝试全局安装 CLI。项目默认保存在本机的 ~/.pascal/data/pascal.db,服务监听本机回环地址;关闭终端后,后台服务仍可能继续运行。这不是只打开一个网页、完全不改本机环境。Pascal Editor 本地 MCP 连接不需要 Pascal 账号或 API Key,也不会自动上传项目,但外部 AI 客户端仍有自己的费用和数据处理规则,因此不能直接称为完全离线。连接托管项目时,则需要按客户端配置账号、API Key 或 OAuth,并分别确认项目访问权限与积分消费授权。Pascal Editor 费用方面,官方目前说明手动建模、编辑和分享免费;托管 AI 聊天、图片视频生成、物品和材质生成会消耗 Credits。订阅及外接 AI 服务的费用另行计算,不能把源码开放理解成所有 AI 功能无限免费。源码采用 MIT 许可,复用时需要保留版权和许可声明,托管服务另有使用条款。Pascal Editor 涉及真实住宅、客户方案或其他敏感资料时,也要先看数据流向。托管服务会收集账号、项目、上传内容和 AI 活动;使用 AI 功能时,相关提示词、图片和项目内容可能交给第三方模型处理,项目关联的 AI 聊天记录会保留到删除项目或账号。Pascal Editor 还有一个不能忽略的分享边界:官方文档说明,已有的素材和 Studio 存储链接,即使所属项目设为私有,仍可能公开访问。项目目录的访问权限不会自动让这些链接变成私密链接,删除文件记录也不保证底层文件已被彻底擦除。Pascal Editor AI 接入和导出也不是一回事:当前文档说明 MCP 的 export_glb 尚未实现,不能把 JSON 导出或预览画面当作已经生成的 GLB 文件,更不能保证任意 AI 客户端都能完成完整的渲染与导出。实际能力应以连接的服务和安装版本为准。Pascal Editor 参考来源:Nico Garcia(@NicoGarcia_IA)关于浏览器 3D 建筑编辑器的介绍: xitter.cf/NicoGarcia_IA/status/2… 🔗 github.com/pascalorg/editor
2
1
5
676
We ❤️ architects The tools get obsolete, not the people Free the architect from the software and watch what they create AI is here for them today, at @pascal
Arquitectos luego de que CAD y BIM se vuelvan obsoletos en 2028
1
1
24
1,743
🔥🔥🔥 @pascal for lift-off 🚀
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model. 🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal. 🔹 Full-project coding: games in Unreal, Unity & Godot. 🔹 Excels at parametric 3D modeling in Blender & CAD. 🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%. 🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks. Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API. Open weights by Nov 4, 2026. 🔗 API: platform.kimi.ai 🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
1
9
364
Not a demo, not a pitch. A construction company running @pascal across 10 teams for scanning and sales. Free. Better than Bosch 😍
Using Pascal in our Construction Business across 10 Teams for scanning - The best part its free, better than available apps like bosch, and you can use the scan even in your sales process and project materials etc in the room, flat, houses. 10/10
1
2
7
957
Lines don't know they're walls. Objects do. The shift from 2D to BIM took architects 20 years. We're compressing it to one scan at @pascal: LiDAR in, scene graph out, every wall an object that knows what it is.
In 2D CAD, a wall is just lines. The software doesn't know what they are. In BIM, a wall is an object that knows its material, size and quantity. Move it once and every plan, section and elevation updates. That's the shift I'm learning to make.
4
3
1
48
3,677
Easy job with Pascal Capture on iOS. It uses LiDAR scans, processes it with GPT 6 Astra, and drop it into an editable environment apps.apple.com/us/app/pascal…
I have discovered the the BEST way to get 3D modeling results from LLMs The app LiDAR scans the room while taking hundreds of high quality photos. Then a workflow + skill I've created models out every item recursively checking the model against the photo at the exact spot + angle where the photo was captured The result is a a dimensionally accurate and high quality model of your room All in my latest video on how to do it youtube.com/watch?v=JGQCcCzu…
1
3
16
760
We use @bot extensively now. 15 agents and a chief of staff handling the entire stack. Am I running a company or just being the idea guy? btw @elonmusk I want to completely replace Slack. When group threads?
2
2
9
455
We named the company @pascal after the man who built the first calculator at 19, invented probability theory, and wrote the Pensées before dying at 39. Blaise Pascal is the greatest man who ever lived. If Thiel is right, we’ll be fine.
Peter Thiel has a theory most founders ignore: your company name predicts whether you'll win or fail. "I believe very strongly that the names of companies are often very predictive of future failure or success." His framework, in his own words: PayPal worked because it felt human. "It was the friend that helps you pay." Friendly, transactional, trustworthy — the name did sales before the product did. Napster was doomed from the start. "You nap" — the connotation was sleep, theft, something illicit. The name telegraphed exactly what regulators and labels would eventually come after it for. The deeper point isn't aesthetics. A name is a compressed signal — about your market positioning, your customer relationship, and how the world will narrativize you when things go wrong. Names expose founder psychology. A name that sounds aggressive or antisocial is usually a tell that the founder hasn't thought hard enough about how their company will be perceived at scale. Most founders treat naming as a branding exercise. Thiel treats it as a predictive test.
2
1
8
789
One prompt. One building. @pascal does it 10x faster and 10x cheaper than GPT, Claude and Gemini, and the 3D is better. Frontier model for the physical world. Meet the AI architect. Big release and big announcements coming.
6
8
72
3,980
Dropping cars into a garage is awesome. The same loop with real data is a product: @pascal Capture scans your garage, Studio places furniture and fixtures snapped to actual walls and slabs, and an agent can check if it fits through MCP. Try it in the browser.
build out your garage and fill it out with your dream cars Mint let's you create your own 3D assets inside your environment Try it below!
1
1
6
498
Love this. AI-generated gym equipment is exactly the kind of asset we want dropping into real buildings. @mintdotgg as a Pascal Studio plugin: generate the treadmill in Mint, place it in your actual scanned home gym in @pascal. Who wants to build it?
Everyone’s using AI to build games. We built a Room building game where you use AI to build. In this tutorial, I generated my dream home gym and gym equipment in Mint. Here’s how 👇
1
1
28
2,191
Moving Atoms is doing for industrial plants what @pascal is doing for buildings: turning scans into scene graphs, not meshes. Agents and simulations need objects they can select, label and reason about. Great work! 💪💪
A scan gives you a picture. A simulation needs objects. We can output Revit files, USD, IFC. Anything needed for BIM/Simulation workflows. This is a real process plant in Rotterdam. On the left is the raw LiDAR, about 5 million NavVis points. On the right is our reconstruction, filmed from the same camera. Most scan-to-3D pipelines stop at one fused mesh, which looks right but behaves like a single rock. A robot can't open, avoid or reason about something it can't tell apart from the floor. So after we rebuild the geometry, we separate the scene into objects. In this plant that's 89 of them: pumps, vessels, heat exchangers, stairs, guardrails, pipe networks, foundations. Each one: → is its own mesh with its own pivot point, so you can select it, move it or swap it → is labelled by type, like "pump assembly" or "vertical heat exchanger" → traces back to the scan panoramas and points it came from → is flagged where we had to infer hidden geometry rather than measure it
1
8
109
6,533
Tell your agent about Pascal. Same idea, but for real buildings: describe a space, or scan it with your phone, and get a scene graph you can light, measure and edit. Street lights included. editor.pascal.app
Probably the most impressive output I received from this skill: A street light with adjustable pole height and light distribution. You're able to see the light move between the road, the houses and the sky. Things get interesting when the LUX map gets turned on.
2
52
4,651
Congrats Jake. Fun timing: github.com/pascalorg/editor just passed 24.8k stars doing the same thing one level up. Text-to-CAD gives you a part. Pascal gives you the building the part goes in. Open source, MIT, MCP-native, runs locally with one npx command.
text-to-cad just crossed 20,000 stars on github! here's a lightning talk i gave about it at frontiers, enjoy
2
4
29
2,537
Love this. Manga artists need backgrounds, so someone built a school generator. Insurers, contractors and homeowners need the same thing for real houses, and nobody has it. That’s what we’re building at @pascal.
ボタン1つで学校を作成し、細かく編集できるツールを作りました!!全部屋、内装付きです!! 部屋ごとに作り直す機能も用意しています。あなただけの理想の学校を作り、漫画で使ってみてください!! tools.oyasumi-gamedev.com/ha…
1
3
31
2,399
“Every object is separated and editable” is the game. A point cloud is a photo, a scene graph is a database. Moving Atoms is building that for factories and robots. @pascal is building it for buildings and the people who insure, sell and repair them.
Moving Atoms (YC S26): Scan to Sim is here. Send us a LiDAR or SLAM scan of your factory floor, warehouse or yard, from any scanner. Upload multi-gigabyte scans straight from your browser (up to 200 GB per file). Within 48 hours you get back a simulation-ready digital twin of your real site. Every object is separated and editable, and accurate robot embodiments are placed right on your floor: autonomous forklifts, pallet AMRs, tuggers, tote robots and palletizing arms. Test the workflow before you build it. Try layouts, fleet sizes and routes, then watch the whole operation run sped up, before a single robot arrives on site.
4
1
43
2,221
Adaptive point size is underrated. We get asked constantly why @pascal keeps the raw scan around instead of meshing everything. This is why: every mesh is an opinion about the data. The points are the data.
Nira:メッシュ化なしで点群を面のように表示 ☁️ 🔹点群をメッシュ変換せずに描画 🔹ソリッドな表面のような見た目を実現 🔹メッシュ化によるデータ表現の不一致を回避 情報元はリプ欄。
1
3
20
1,228
Graphs are the right primitive for agents. We went one step further at @pascal: the scene graph is the agent's world model of a building. Every wall, pipe and fixture is a node. Information in, information out.
Andrew Ng just dropped the best 2-hour course on Graph Engineering: from single agent to full automation 9:14 - your first agent 33:11 - loop engineering 1:02:46 - graph engineering 1:30:15 - agents that rewrite themselves 1:49:05 - full graph system two hours, and it replaces every agent tutorial you bookmarked this year Prompts → Agents → Loops → Graphs most people will stop after the first agent and call it automation he saves the last forty minutes for the graph that runs it without him same model, same tokens, completely different week watch it today
4
1
9
924