💛💛💛

For years, people said “learn to code first.” Now you can have the idea first and figure out the rest later. Asked Claude Opus 5.5 to build this little artistic world in pure JavaScript where a tree grows over time, you can grab and shake its branches, and the leaves fall naturally. And it just DID it. Without specifying what has to be used, how to do it and anything. It just DID it. Best time in history to be a creative.

Claudnardo da Vinci 🎃

Seriously, this is INSANE 😱 I asked Claude to draw a complex watercolor painting, and it did the WHOLE process, matching it perfectly pixel by pixel. This is tooo good. Maybe even useful for learning how to draw artworks like this step by step?

Thank you :))

A short film for kids created entirely in javascript using Claude Opus 5.5 👀

Thank you so much :)

Thank you so much :))

This city has no end 😱 Claude Opus 5.5 built a tram ride through the hills of Lisbon that keeps growing as you travel. White plaster houses, terracotta roofs, iron balconies filled with plants, blue tiles and cobbled streets, all generated entirely in code. No models. No Blender. Travel through winter, autumn, spring and summer with rain, fog, clear skies and different times of day from dawn to night. Enjoy your ride :) Live: endless-train.wasmer.app/ Code: github.com/thebuggeddev/endl…

Thank you so much :)

A house through the seasons, painted in watercolor with pure JavaScript. Built with Claude Opus 5.5

What if you could see what the world’s biggest inventions looked like on day one? 😲 The wheel, phone, airplane. Everything we use today started as something much simpler, so I built a website where you can explore famous innovations as interactive 3D models in @threejs and see how they evolved over time. The problem was this needed a stupid amount of historical research and I was way too lazy to do all of that manually 😭 So I used @Apodex_AI Deep Research. I asked it to find top 11 innovations of humanity and it basically went off, scouted the internet, gathered sources, figured out how each one evolved, organized everything into proper research, and even created prompts for the visuals. Its research is so goated I even use it to find coupons before buying stuff 😉 The coolest part though is how automated the actual build became afterwards. Before, I used to generate images separately, download them, move them into another tool, turn them into 3D models, clean everything up, and then bring those models back into the project. Now I just gave the research PDF and my prompt to Codex (GPT-6 Astra). It generated the visuals with Imagen, passed them through the @tripoai Codex plugin, turned them into 3D models, placed them into the experience, and kept building. I still had to manually iterate on a few models, but most of the major innovations went through that pipeline with barely any manual asset work. You really do not need 10 or 20 autonomous agents fighting and spending millions of tokens to build cool things anymore. Just use the right tool for the right job. Live: praxis-human-innovation.theb…

Fluffy social buttons in Three.js with TSL, built with Claude Opus 5.5. And no, there's no fur texture here. It's shell texturing where you render the same button 48 times, each copy pushed slightly farther out than the last. On every layer most pixels get discarded, and the ones that survive line up across layers into tiny strands or flur. That's how it's done. Pure Fur Math.

Does Claude Opus 5.5 actually perform more like Claude Fable 5 or 5.1? ☠️ I had to test it, so I asked Opus to recreate a V8 engine,.. but but entirely out of wood. I’d built a MechLab before, and even that wasn’t as detailed as what Opus managed to create with this wooden V8. Maybe it’s because this was just a single engine task, or maybe these models have seen plenty of V8 engine examples during training. Either way, the result turned out so incredibly well, isn't it? This Opus 5.5 V8 engine is noticeably more detailed than what I’ve been able to get from any of the other models I’ve tried Also less tokens and time.

Yep, Opus 5.5 feels good so far. Thank you :)

Claude Opus 5.5 built an endless 3D sketchy Santorini Island of Greece

Grok 4.6 after 208k tokens and 17m41s producing almost same result as Grok 4.7 🙄

Thank you so so much 🥹♥️

Grok 4.7 after 183k tokens and 24m 51s replicating a dashboard design 👀

How good is GPT-6 Astra at mobile app development? 😈 So gooood that it build a 13 screen mobile app in ~30 minutes from basically half of a single design reference image. And the prompt was just a few lines. Nothing extraordinary loop plan and such. Super basic instructions and Astra figured out the rest on its own. It really nailed whole design system, animations, micro interactions, transitions and overall smoothness way better than I expected. The entire app is built in React Native using @expo’s latest SDK with everything fully up to date. 13 screens. Half a reference image. A few lines of prompting. Around 30 minutes. READ THAT AGAIN! Isn’t that a little too Astraordinary? Design: dribbble.com/shots/26787072-…

Do you actually need to know 3D modeling to build websites like this? 🤔 Probably not. I don't know 3D modeling much in detail, and I still managed to build this animal inspired drone website using GPT-6 Astra and @tripoai. And no, it's not anything complex to build. I found a design on Pinterest, then used GPT Image 2.5 to generate a few animal inspired drone concepts. I passed those images into Tripo, which turned them into 3D models with smart UVs. You also don’t need to write some huge prompt for Astra, btw. I just explained what I wanted the website to look like, how the transitions should work, and how the 3D models should behave... a couple of lines were enough for it to get the whole thing working. A lot of people don’t like AI generated 3D models because the meshes can be messy, there can be extra vertices or triangles, and the topology might not be perfect. That’s fair, but for websites like this, I don’t think it matters that much. If the model looks good, the site runs smoothly at 60 fps, and it does what you need it to do, that’s enough, no? You don’t need a perfect 3D asset just to have a cool 3D element on your portfolio type projects. Am I wrong? Live: av8-r-drones.thebuggeddev.ch… Design: pin.it/ise0Id6yi #TripoAstra #Tripo3D #TripoAmbassador

Thank you so much :D

If a maple tree could grant your wishes,.. built with Claude Fable & Opus 🍀 Place your wish anywhere on the tree, explore wishes left by others, and watch the tree grow with every new wish. wishwood.wasmer.app/

Claudingg alongside a bookie and plushie 🧡 It's so interesting to read that everyone has a different way of prompting. Mine is as simple as describing what I want in plain words. What's yours? Btw, thank you @PacktPublishing and @Prajwalyes for sending this book. It's def a good read a.co/d/03RuIokq

Thank you so much :D

⚠️ Warning: Not for the weak hearted. Enter at your own risk 😂 Everyone on X has been creating beautiful 3D environments lately so I decided to make something a little different... weirdly different. Introducing Haunted Hollow District, an endless 3D horror district built using GPT-6 Astra and @tripoai where you genuinely have no idea what you are going to run into next. Creepy creatures, a terrifying NUN, broken cars, strange encounters and all kinds of things messing with you around every corner. All of those 3D assets were created directly through the Tripo Codex plugin. Astra planned the whole experience, figured out the assets, movements and environment, used Tripo to model everything and then brought the entire district together. See how long you can survive wandering around :D hollow-district.thebuggeddev… #TripoAstra #Tripo3D #TripoAmbassador P.S., Doing my hashtag duties for a chance at AirPods 5. You should participate too 🫣

Built a 3D Robosnake combat arena without learning opening touching Blender ❤️‍🔥 It has some really cool animations and combat moves like fire breath, claw combos, lunges, and it even roars. All using @threejs. The whole thing started with a SINGLE image in @tripoai. It generated the 3D model, then I used Edit Mesh to fix a few parts, generated Smart UVs, and auto rigged it too. Then handed generated model over to Claude Opus 5 and it built the rest from there. Everything right within browser. Isn't it super cool? It's crazy how 3D creation is getting ridiculously accessible now. You can go from one image to a rigged character and build games, 3D experiences, or interactive applications around it in minutes. Really exciting to see how fast this space is improving Live: robosnake.wasmer.app/ Code: github.com/thebuggeddev/robo…

No, initial try missed some details. When I used Edit Mesh, it added those missing parts too. You can see those details in this image, which I had originally provided to create that model :)

I got you in the first part, but not in the last part 🤔

You said it would take another 1–2 years before AI could properly edit meshes, handle rigging, and do all of this… and Tripo just did it already 🤯 It looks like they’ve casually launched P 2.0 alongside Edit Mesh, Smart UV and Auto Rigging. Just look at how detailed and clean these edits are while preserving the rest of the model almost perfectly. The original model had four fingers. After editing, it now has five,.. while everything else stays consistent. And somehow, it even managed to silently the number of faces and vertices in the process. You may not realize it but we’re getting incredibly close to a real revolution in 3D modeling. studio.tripo3d.ai/3d-model/3…

Tried it and this is arena result 🫣

It's output is even bad compared to the one above from AI Studio 😂

Are you sure? 🤔

So this is the result of Gemini 4 Pro hiding under the name “Gemini 3.8 Flash” 🙄 I tried this so called rumored “Gemini 4 Pro” against famous "V8 Engine" demo from X and no, this is definitely not what all those posts about Gemini 4 Pro supposedly topping Arena benchmarks made it sound like. It does feel better than any Gemini model I’ve used so far, but it’s still not there. I wouldn’t even put it at Opus 5 level yet. Not sure what this hype is about, tbh. ai.studio/apps/14a298c1-7df5…

Thank you :)

Thank you so much :)

Thank you :D

Thank you :)

Thank you :D

Thank you so much :)

I genuinely never thought this would be possible 😱 AI just built a full embroidery editor that runs entirely in the browser. Freaking crazy. And no, it’s not GPT 6 Astra. It’s Claude Fable 5.1. Also, No Three.js. No WebGL. No shaders. Just Canvas 2D and math. You literally draw using a needle cursor and the app converts your path into actual embroidery stitches like running, back, chain, cross, satin, French knots and lazy daisy. Even the fabric is generated pixel by pixel. Two overlapping sine waves create a height field for the horizontal and vertical threads. Then a fake normal is calculated and lit from one direction. Basically a bump map with zero GPU rendering. The embroidery thread itself is rendered in 6 passes. Shadow, core color, twisted ply lines, dark edge and two highlight layers based on the light angle. That’s why it actually looks like embroidery floss instead of a simple marker stroke. And trust me, I've zero sense or knowledge of any of this. ZERO. All built. All working. All by AI. And somehow it’s all inside one HTML file. Nothing bigg. Absolutely insane what’s possible now. You can really genuinely take a random idea and bring it to life. And that too in < 1 hour. And who knows, this could actually be pretty useful for fashion designers and embroidery artists maybe? Live: stitchcraft.wasmer.app/

You’d probably call it slop, but I like this AI slop way too much lol 😂 modak-festival-catch.thebugg…

Same prompt. Same dashboard,.. but a very different execution 🥶 I tested Kimi K3 against GPT 6 Astra on the exact same dashboard prompt and their output difference is so obvious. Astra replicated the reference almost flawlessly. The font family, charts, layout, spacing and overall design system were all extremely close to the original. Kimi K3 also performed well, especially considering this is just a single prompt result. But it missed the exact font family, the charts were not as accurate and it took almost an hour to complete the dashboard. Astra built the same thing in < 10 minutes, if I remember correctly. For usage, I am on the Max plan for both. I am not 100% sure how directly comparable their usage metrics are, but Astra seemed to consume around 1 to 2% of a session while Kimi only used roughly .40% percent of a session to build this. One thing I really liked about Kimi K3 was how it handled swarm agents. I had enabled swarm mode, but it recognized that this was ultimately just a single dashboard UI task. Instead of unnecessarily invoking multiple parallel agents, it automatically switched to a single agent build. That felt genuinely intelligent. Of course, with a couple more iterations and a stricter prompt, you could probably get Kimi K3 much closer to Astra's result. But, as I said before, difference is that it takes more time and more hands on iteration compared to Astra, while potentially using fewer tokens overall. Live: moonrow-dashboard.kimi.page/