The Project
Building AI systems requires a development environment that’s not just powerful, but intelligently designed to leverage AI at every step of the workflow. My setup is carefully curated to maximize productivity while maintaining the flexibility to work with cutting-edge AI tools and frameworks.
Hardware Foundation
The foundation of any serious AI development setup starts with capable hardware that can handle everything from local LLM inference to complex multi-container deployments.
Primary Workstation:
- 14” MacBook Pro M3 Max with 128GB RAM: The M3 Max provides excellent performance for both development and local AI model inference, while 128GB ensures I can run multiple large models simultaneously without memory constraints.
- LG IPS 4K Ultra-wide Monitors: Dual ultra-wide displays provide massive screen real estate for managing multiple terminals, code editors, documentation, and monitoring dashboards simultaneously.
Input Devices:
- Custom Ferris Sweep Split Keyboard: This ultra-minimalist 34-key split keyboard eliminates unnecessary keys and promotes better ergonomics during long coding sessions. The custom layout is optimized for programming symbols and Vim motions.
- Ploopy Adept Trackball: Precise pointing device that reduces wrist movement and provides consistent accuracy for detailed work.
Workspace:
- Secretlab TITAN Evo 2022: Ergonomic chair designed for extended work sessions, crucial when deep-diving into complex AI architectures.
Development Environment
My development setup is built around efficiency, AI integration, and the principle that the best tools get out of your way.
Primary Editors:
- Zed Editor: My main GUI editor, chosen specifically for its exceptional AI integration. I extensively use its agent mode capabilities, Model Context Protocol (MCP) servers, and custom rules. The first-class Vim support makes it feel native while providing modern AI-assisted development features.
- Neovim: My terminal-based editor for quick edits, remote work, and when I want the pure efficiency of a keyboard-driven workflow. Heavily customized with plugins for AI integration.
- Doom Emacs: Used primarily for Orgmode, which remains unmatched for knowledge management, planning, and documentation. I also use it occasionally for Windsurf and Cursor when specific AI features are needed.
Terminal Environment:
- Ghostty: Fast, GPU-accelerated terminal with excellent font rendering and performance
- WezTerm: Alternative terminal with powerful configuration options and multiplexing capabilities
- AeroSpace Tiling Window Manager: Automatically organizes windows for maximum efficiency, essential when managing multiple terminal sessions, editors, and monitoring tools
AI Integration at Every Level
What sets this environment apart is the deep integration of AI throughout the entire development workflow.
Command Center - Raycast: Raycast serves as the central nervous system of my entire workflow:
- AI Integration: Direct access to multiple LLMs for quick queries and problem-solving
- MCP Servers: Custom integrations that connect my knowledge bases and tools
- Custom Scripts: Automated workflows for common development tasks
- Snippet Management: Intelligent code snippet system with AI-powered suggestions
- Clipboard Manager: Advanced clipboard history with AI-powered search and organization
AI Development Tools:
- OpenAI Codex & Claude CLI: Command-line access to powerful coding models for automated code generation and review
- Google Jules: Integrated development assistance for complex problem-solving
- Insomnia: API testing tool that’s more intuitive than Postman, with excellent support for AI-related API testing
Local AI Infrastructure:
- Ollama & LM Studio: Running local LLMs including Qwen 3, Gemma, and PHI 4 for privacy-sensitive work and offline development
- Open WebUI & Librechat: Self-hosted interfaces for managing local and remote AI models
- Msty: Lightweight AI client for quick model interactions
Knowledge Management & Productivity
Effective AI development requires excellent knowledge management and organizational systems.
Knowledge Systems:
- Doom Emacs and Orgmode: The gold standard for hierarchical note-taking, project planning, and documentation. Nothing else comes close to Orgmode’s flexibility and power.
- Anytype: For collaborative notes and knowledge sharing with team members who don’t use Emacs
- Highlight.ai: Real-time context capture and retrieval during development sessions
Monitoring & Observability:
- LangSmith: Primary platform for LLM application monitoring and debugging
- Langfuse: Open-source LLM observability for detailed analytics
- Langwatch: Real-time monitoring of AI agent performance
- Comet Opik: Experiment tracking and model performance analysis
- Latitude.so: Collaborative AI development and deployment monitoring
Creative & Design Tools
AI development often requires creating content, diagrams, and visual materials.
Design & Graphics:
- Figma: Interface design and collaboration, especially for AI application UIs
- Photoshop: Advanced image editing and manipulation
- Recraft.ai: AI-powered design asset generation
- Replicate.com: Running custom AI models for specialized creative tasks
Video & Content Creation:
- iMovie: Video editing for YouTube content and documentation
- Google Veo 3: AI video generation for creative projects
- OpenAI Sora: Experimental video generation capabilities
AI Image Generation:
- Google Imagen 4: High-quality image generation integrated into workflows
- Replicate.com: Creative image generation for presentations and content
AI Model Access & Management
Working with AI requires access to the latest models and the ability to compare their capabilities.
Cloud AI Services:
- ChatGPT Plus: Access to GPT-4 and latest OpenAI models
- Claude Pro: Anthropic’s most capable models for complex reasoning tasks
- Gemini Pro: Google’s advanced models with excellent multimodal capabilities
Local Model Management:
- Qwen 3: Exceptional performance for coding and technical tasks
- Gemma: Google’s open-source model family, great for experimentation
- PHI 4: Microsoft’s efficient small model for quick inference
Workflow Philosophy
The entire setup is designed around several core principles:
AI-First Development: Every tool in the environment either natively supports AI integration or has been configured to work seamlessly with AI assistants. This enables a development flow where AI is a natural extension of my capabilities rather than a separate tool.
Keyboard-Driven Efficiency: From the custom split keyboard to Vim keybindings throughout the environment, everything is optimized for keyboard efficiency. This reduces context switching and maintains flow states during complex development work.
Obsessive Organization: With Raycast as the command center, Orgmode for planning, and comprehensive monitoring tools, every aspect of the development process is tracked and optimized for continuous improvement.
Privacy & Control: Local LLMs and self-hosted tools ensure that sensitive code and data never leave my control, while still providing access to the latest AI capabilities when appropriate.
Wrap Up
This AI-powered development environment represents the convergence of powerful hardware, intelligent software choices, and deep AI integration. Every component has been selected and configured to support the complex, iterative process of building production-ready AI systems.
The key insight is that AI development requires an environment that’s not just capable of running AI models, but one that actively leverages AI to enhance every aspect of the development process - from initial ideation through deployment and monitoring.
This setup enables me to move fluidly between designing complex agent architectures, implementing robust automation systems, and creating content that makes AI accessible to others. It’s a development environment that grows more powerful as the AI ecosystem evolves.