Founder / Lead Creative Engineer @lusionltd and building wearable AI https://nitter.cf/t.co/A7QGyvoMhy
United Kingdom
Joined October 2010
- Tweets8.7K
- Following788
- Followers8.3K
- Likes7.3K
Pinned Tweet
If you missed my Particle-Love at #3dfestweb, here is the live link: edankwan.com/experiments/par… Motion Blur Everywhere!
It is insane that I can vibe coded a Synergy alternative with Claude that I can control my X11 Ubuntu and Mac from my Windows machine. It has already supporting clipboard text, image and even file/folder transfer. Synergy file transfer support is on the wishlist for like over 10 years. And clipboard is hit or miss and it is mostly miss.... Next step is drag and drop file transfer and modifier keys support.
Btw, it only used 9% usage from a 20x sub at this point.
Telling Claude not to stop until it finished every task on the to-do list was a terrible mistake. As it turns out, Claude has been "discovering" new tasks as it goes, apparently inventing problems just to solve them.
It started with 91 tasks. Claude has somehow completed 409, and there are still 46 left. 90% of these "new features" are completely useless nonsense that nobody asked for.
Claude Mods is a good example that I believe in the future all software should have a protocol to allow AI to add customizable features.
Testing out H3 rendering 360 to generate Gaussian Splats workflow with Codex. Then it mentioned $12,000 budget... What the hell was it doing?
My job is safe for now
JEV + SONNET 5.5 is the most dangerous combo in design right now...
That's not a screen recording you're watching - It's a live website - and here is how it came together
Step 1: I went through oryzo frame by frame and took its grammar (the pacing, the type, the way the camera moves)
Step 2: The whole scroll hangs on one hero object, so I rebuilt Claude's pixel mascot in code from 500 voxels without downloading a single 3D file
Step 3: Each frame depends only on the scroll position, with no timers and nothing random, so the same spot in the scroll always gives you the same frame and the site renders straight into a film
Step 4: Jev handles the routing, so scenes, 3D and copy go to Sonnet 5.5 while the frame by frame taste review goes to Opus 5.5
Step 5: Four Sonnet 5.5 agents worked in parallel and had four scenes built in roughly 24 minute
Step 6: Nothing ships until every still has been looked at, and whatever looks cheap gets rebuilt
You and I are using the same model. What changes the result is the harness you build around it
Prompt's in the comments
Gaslighting your AI agent to optimize your app further seems like a good way to push things further.
Okay, everyone has been sharing what they’ve built with Opus 5.5, so I finally bit the bullet and gave it a try.
This is my first project: My Mini Houdini.
I started building it last Friday, and this is where it’s at so far.
Shading:
Basic unlit, basic realtime, and realtime+ with contact shadows, post, etc.
There’s also a basic path tracer with OIDN denoising.
Node Graphics:
For geometry creation, I currently have model import and export, curves, grids, boxes, circles, spheres, lines, tubes, torus, test geometry and more.
There are quite a few modifiers in there already that I haven’t even properly tried yet.
Copy to Point, Packing and Sweep are also working in a basic form.
There’s a Solver too, although it’s still a bit funky. I need to figure out a better caching structure for it.
I’ve also added a Compute node and a JavaScript node.
I decided to keep pretty much everything on the GPU, so this isn’t really designed for massive geometry or heavy simulation. Cache budget can become a problem pretty quickly.
All geometry data stays inside compute buffers, with a hash table handling string attributes.
Material / Shading:
MaterialX for now.
Still very early, it is an electron app. Just a fun pet project. I am impressed.
