Sponsored
$5/week ↗
Projects

Designer had their Dot turn a floor plan into a 3D Blender model plus client proposal renders

added

模型+跑圖一次滿足的 dots 供參 - 很久沒打開autocad 剛配完平面>讓dots用blender把3D模型建好>做風格對齊>視角截圖>做提案渲染 後來是轉用 three.js 出 html,想說未來傳送給業主比較方便 效果實在蠻好,額度目前看起來還蠻夠我用的

First-hand Threads post (2026-10-11): designer 33.Shan (real handle @33.shan.design; dots stripped per alphanumeric validator rule) reports a concrete delivered result from their Dot. After finishing a floor plan in AutoCAD, they had their Dot build the 3D model in Blender, do style alignment, capture viewpoint screenshots, and render the client proposal — then switched the output to three.js HTML so it can be handed to the client directly. They report the results were genuinely good and the usage quota was plenty. A client-ready deliverable (proposal renders / HTML), not a demo or commentary. Post is in Traditional Chinese; translation above. No #ChatGPT_Partner tag.

See all 780 →
Reddit · Projects

Tried Dot for 1 week and it won me

I know the cool thing has been dunking on it, and I was skeptical too. So I challenged myself to use it starting last Friday. TLDR: I had Dot help build a real, playable Digivice in my downtime. My son ran around the house playing it for two hours. Based on my usage, I have questions about efficiency, but the benefit of having an orchestrator I can just chat with is pretty hard to deny. I had it design and build a real Digivice to play with my son, entirely in my downtime. The idea was simple. Let’s make a Digivice that actually works. Something we can carry around, encounter Digimon, battle, capture, and evolve them. So I started issuing Dot tasks: \- Build a 3D model of a Digivice. \- Build a shopping list to turn that model into an electronic game. \- Build the firmware around my vision for how the game should work. \- ??? \- Profit. Son ran around the house for two hours playing it. Obviously, I still had to print the parts, assemble the electronics, and actually test the thing. And there’s been plenty of iteration. Screw holes needing adjustments, the screen facing the wrong direction, touch controls that sounded good but felt finicky on the device. Real hardware gives you feedback pretty quickly. A lot of my messages were basically: “This doesn’t feel right. Let’s change it.” Then I’d describe what I wanted and have it work through the changes. That’s the part that won me over. I could give it a high-level objective during some downtime, come back, review what it had done, and give it the next set of changes. We went from an idea to something my son and I can actually play with. Playable web preview: [https://goldsziggy.github.io/z-ark-digivice/play/](https://goldsziggy.github.io/z-ark-digivice/play/) Device renderings: https://goldsziggy.github.io/z-ark-digivice/ GitHub, if you want to take a look or build your own: [https://github.com/goldsziggy/z-ark-digivice](https://github.com/goldsziggy/z-ark-digivice) In addition to this, I’ve had it: \- Manage tracking for my son’s mattress order and draft emails asking for updates. \- Work on a Pokopia-like game with Digimon, Pokémon, Palworld, and Pikmin influences. \- Bounce around work ideas, including multi-agent frameworks and MCP proposals like skills over MCP. \- Work on the bedtime book app I use with my son. Based on my usage percentages, I can almost guarantee it isn’t very efficient with its agent use. But I can’t deny the benefit of having essentially an orchestrator in chat that I can delegate high-level objectives to. And in this case, that turned into a real game I can play with my son. No internet required; and can take to the park like it’s Digimon Go