tcheina

Current dev setup

Current dev setup

AI-Augmented Development
tools
development setup
productivity
Mon Jan 19 2026 00:00:00 GMT+0000 (Coordinated Universal Time)

My Current Development Setup

After months of articles and reading about how everyone sets their dev environment up for productivity and personal development (not large scale team coordination, initiatives and management), a lot of testing on my end, I settled on the following:

  • Hardware: MacBook Pro
  • Cloud Platform: AWS for backend services and deployments
  • Quick, small deploys: Netlify/Cloudflare for frontend hosting, Render for serverless, Supabase for databases
  • Editor: VS Code
  • AI Tools: Claude Code and also Ollama Deepseek Coder (local LLM)
  • Terminal: iTerm2 (customized)with Zsh and Oh My Zsh, along with vi, neovim for quick edits
  • Containerization: Docker for isolated environments
  • Version Control: Git with GitHub, using GitHub Copilot for code suggestions
  • Other AI Tools: Bolt with OpenAI and Claude for the random questions not requiring much context
  • IaC: OpenTofu (Hashicorp changed Terraform’s licensing), AWS CLI, Netlify CLI for infrastructure management
  • Project Management: mdx files for documentation and task tracking
  • UI Framework: Astro with React components for building web interfaces, Typescript, Tailwind CSS
  • UI Mocks: Loveable for quick UI mockups requiring beautiful design, Base44 for functional dashboarding design
  • Note Taking: Notion, Obsidian for personal knowledge management and note-taking
  • Diagramming: Miro and Diagram As Code (Diagrams.net) for architecture and flow diagrams
  • Fav Programming Language: Python for backend services, TypeScript for frontend development, Go|Rust for specific performance-critical components

I have reviewed a whole bunch of recommendations from various generous developers who share their setups, and I have tried many of the common ones:

  • Jetbrains IDEs (IntelliJ, PyCharm) - great but too heavy and clunky
  • Warp terminal - interesting but the support is greater for VS Code
  • Cursor - good with limitations
  • Windsurf - good but pricey and really not as good as Claude Code
  • Github Copilot - good for suggestions but not full code generation

I find myself able to farm out work in an agentic manner with ITerm2-Claude Code and Bolt concurrently, which works well enough.

It is mentally draining to switch contexts and also review AI generated code. The quality is improving and once you spent a few hours weighing pros and cons and a systems design that will work for your timeframe, the rest goes easy with littered contexts written to mdx files.

I suspect this will be the way forward and we will have to let go of the past standards of utterly concise, tight code deploys.
Largely due to perception and cost, business will pay for what they feel is quantity since they are not able to assess quality.