Three.js & WebGL
Complex scenes — car configurators, product viewers, shader-heavy interfaces — render and compile without stuttering. Hot reload stays fast even with 50k+ polygon models loaded.
Developer Setup · AI · 3D Web · 2026
Mac Studio M3 Ultra
My primary machine for AI automation, Three.js scenes, WebGL shaders, 3D product configurators, local LLM inference and full-stack web development. Here is what I actually use it for — and what you do not need it for.
M3 Ultra
Chip
96 GB
Unified Memory
1 TB
SSD Storage
60-core
GPU
32-core
CPU
32-core
Neural Engine
Why I chose it
Built for heavy creative work
The M3 Ultra is not the only machine that can do this work — but it is the one where nothing ever slows me down. When you build interactive 3D web experiences, run local AI models and manage automation pipelines simultaneously, unified memory matters more than raw clock speed.
Local LLM testing
96 GB unified memory allows testing selected local models with less aggressive quantization than smaller machines require. Useful for prototyping AI features offline before connecting to production APIs. Very large models may still need cloud compute.
Parallel workloads
VS Code, a dev server, a local n8n instance, a browser with heavy DevTools and a Python script running simultaneously — no fan noise, no slowdown.
Longevity
A machine at this level stays relevant for 6–8 years of serious professional work. The cost per year is much lower than it looks at purchase time.
My stack
Tools I use every day
The machine is only as useful as the workflow around it. These are the tools I actually open daily for web development, AI projects and automation.
Dev
VS Code
Primary editor. Extensions: Prettier, ESLint, GitLens, GLSL support for shader work, Copilot for autocomplete on repetitive patterns.
3D
Three.js + Vite
Every interactive 3D project — car configurators, shoe simulators, product viewers — built with Three.js. Vite handles the dev server and build pipeline.
AI
Claude · Ollama
Claude for complex reasoning and code generation. Ollama for running local models (Llama, Mistral, Qwen) when testing AI-powered features that need to work offline.
Automation
n8n
Self-hosted locally during development. Connects APIs, handles webhooks, runs AI nodes and data pipelines alongside other workloads without breaking a sweat.
Dev
Python + Node.js
Python for data scripts, AI API integrations, local servers and automation tools. Node.js for the frontend build ecosystem. Both managed with pyenv and nvm.
3D
Blender
Asset preparation for Three.js scenes — converting formats, baking textures, optimizing polygon count before export to glTF. The M3 GPU makes previews fast.
Dev
Terminal · iTerm2
Multiple concurrent terminal sessions with zsh. Git, npm scripts, Python servers, Docker containers and SSH connections all running in parallel split panes.
Design
Figma
UI and layout design before translating to HTML + CSS. Also used for quick mockups and client presentations. Runs perfectly even with heavy component libraries.
Perfect for
When it makes sense
- Interactive 3D web experiences (Three.js, WebGL, R3F)
- Building and testing AI-powered applications
- Running local LLMs for offline AI feature testing
- Self-hosted automation workflows (n8n, custom scripts)
- Heavy build processes with large codebases
- Parallel dev environments without slowdown
- Video and animation preview for web motion projects
- Long sessions — no battery throttling, desktop never slows down
Not necessary for
When it's overkill
- Standard website development (HTML, CSS, simple JS)
- Brochure sites and landing pages
- Email, spreadsheets, document editing
- Basic React or Vue applications without heavy assets
- Occasional freelance work without demanding technical requirements
- Beginner projects and learning environments
Honest note
Hardware helps, but it does not replace the workflow. The biggest advantage is not the Mac Studio — it is the habit of building, testing, publishing and improving projects every week. A MacBook Pro M2 or M3 Pro is completely sufficient for most web development.
How I work
My daily workflow
The setup is designed to keep context switching to a minimum and keep the machine doing heavy lifting in the background while I focus on the work in front.
01
Idea and structure
I start with a clear use case: AI assistant, product configurator, interactive tool, landing page, game or automation workflow. No tool replaces this step.
02
Local build
I build with HTML, CSS, JavaScript, Three.js or simple backend tools and test everything locally — multiple servers, browser dev tools and asset iterations in parallel.
03
AI-assisted development
Claude for architecture decisions and complex code problems. Local Ollama model for testing AI features offline before connecting to production APIs.
04
Automation testing
For AI workflows, I test n8n logic, webhook flows, lead scoring and email drafts locally before deploying. The machine handles it alongside everything else.
05
Deploy and publish
Git push triggers CI. While that runs, I move to the next feature. Metadata, structured data and internal links go in before publishing.
06
Iterate from GSC data
I watch impressions, queries and clicks in Google Search Console, then improve title, content and project direction based on real search data — not assumptions.
Built on this machine
Projects from this setup
Interactive web projects built with Three.js, WebGL and AI tooling — all developed and tested on the Mac Studio M3 Ultra.
AI
AI Assistants & Automation
Lead capture, scoring, language detection, email flows and small business workflow demos.
→Three.js
3D Car Configurator
Interactive browser-based car color and material configurator built with Three.js.
→Three.js
Shoe Color Simulator
3D product customization concept for fashion, ecommerce and premium visual demos.
→Browser Game
Snake Game
A browser game used as a practical SEO and interactive frontend experiment.
→Browser Game
Rock Paper Scissors
A small game page built to test search intent, engagement and browser logic.
→Browser Game
Browser Games Hub
A small hub connecting games and interactive tools for better internal linking.
→FAQ
Mac Studio M3 Ultra questions
Is Mac Studio M3 Ultra worth it for web development?
Yes, if you work with Three.js, WebGL, large build processes or local AI models. The unified memory and M3 Ultra chip eliminate most bottlenecks in heavy frontend development. For basic web work, the M2 or M3 Pro is more than enough.
Do you need Mac Studio to build AI automation or 3D websites?
No. Many AI workflows, n8n prototypes and frontend projects run fine on a normal laptop. Mac Studio becomes useful when the workflow grows: more tools, more tabs, heavier testing and parallel local processes. The workflow matters more than the hardware.
Can Mac Studio M3 Ultra run local AI models?
Depending on model size and configuration, yes. With 96 GB unified memory, mid-size models like Llama 3 or Mistral can run with less aggressive quantization than smaller machines require. Very large models may still need cloud compute. It is useful for prototyping and testing AI features offline, not for production inference at scale.
Is Mac Studio good for n8n automation workflows?
Yes. Running n8n locally on Mac Studio is very reliable. You can run complex automation workflows, local AI nodes, webhooks and integrations without cloud dependency during development.
What is the difference between Mac Studio M3 Ultra and Mac Studio M3 Max?
The M3 Ultra is essentially two M3 Max chips connected via Apple's UltraFusion technology. You get double the CPU cores, GPU cores, neural engine and memory bandwidth. For most developers, M3 Max is sufficient. Ultra makes sense for very large AI models, complex 3D rendering and heavy parallel workloads.
Need a custom 3D or AI-powered web experience?
I build interactive 3D web tools, AI-powered interfaces, automation systems and modern frontend experiences.
View Services