🚀 I just built a real-time hand-tracking 3D interactive scene in TouchDesigner… in under 10 minutes.
And I did it using nothing but Claude and a few smart prompts.
Here’s exactly how it happened:
Step 1: Hand tracking with MediaPipe
I opened Claude and asked:
“Write a TouchDesigner Python script using MediaPipe to detect hand landmarks from a webcam feed and map hand position to a 3D geometry’s scale and rotation in real time.”
Claude instantly returned a complete working script.
⸻
Step 2: Connecting it in TouchDesigner
Inside TouchDesigner I simply:
• Added a Video Device In TOP for the webcam feed
• Dropped a Script CHOP and pasted the code
• Connected it to a Geo COMP
The result:
✅ Hands tracked instantly
✅ Red landmark points
✅ Green connection lines
Everything worked immediately.
⸻
Step 3: Liquid chrome sphere
Next prompt:
“Add a displacement shader to a sphere geometry that reacts to hand distance. Make it look like liquid chrome.”
I pasted the GLSL code into a GLSL Material and applied it to a sphere.
Suddenly my head turned into liquid mercury.
⸻
Step 4: Interactive glowing cube
Next prompt:
“Noise-based shader for a semi-transparent cube with internal glow, pink & orange palette, reacting to left hand Y position.”
I created a Box SOP, applied the shader, and mapped the left hand Y position to the noise frequency.
The cube started breathing with my hand movements.
⸻
Step 5: Real-time compositing
Final prompt:
“Give me a Render TOP setup in TouchDesigner that composites webcam feed with 3D geometry using alpha blending, no green screen.”
I followed the node graph.
Result:
• Webcam as the background
• 3D objects layered on top
• Zero noticeable latency
Perfect real-time compositing.
⸻
Claude didn’t replace TouchDesigner.
It replaced the three hours I would have spent reading documentation, debugging examples, and searching forums.
This AI-assisted creative workflow feels absolutely insane.
Who else is building like this?
A 7-year-old just shipped his first app… built entirely with Claude Code.
No tutorials. No coding bootcamps. No YouTube videos.
Just a kid describing what he wanted, with his dad sitting next to him showing him how to explain it clearly — and Claude doing the rest.
Most adults spend years learning to code before they ever ship anything. This boy launched his first app before he even learned long division.
One afternoon. One app. Live. People using it. And he’s already earning his first revenue.
By the time his classmates master reading properly, he’ll probably have shipped 10 products.
The old barrier was learning how to code.
The new barrier is simply knowing how to explain what you want.
The future is arriving faster than we thought.
✅ $50K OpenAI API Credits FREE
Use every ChatGPT model daily:
• 250k tokens on GPT-5.5
• 2.5M on minis
• 10M on GPT-3
Tradeoff: Data used for training (disable in settings).
Boris Cherny Anthropic: “I don’t prompt Claude anymore. I have automated workflows that prompt it and decide what to do next. My job now is building those loops.”
While many are still optimizing prompts, the people building AI are already moving to systems that run on their own.
While people debate AI replacing jobs, others are already using AI agents to run ops, sales & management.
A full team is becoming a monthly subscription.
Early adopters win big. Those who ignore it will end up paying for it later.