Uses

My primary coding tool. Agentic CLI that lives in the terminal — I use it for everything from scaffolding to debugging to git workflows.

VS Code fork with AI built in. I use it when I need a visual editor alongside Claude Code, especially for frontend work.

OpenAI's agentic coding CLI. I keep it in rotation to compare approaches and use it for tasks where GPT-5-CODEX shines.

My default for everything web. Full-stack with Next.js, API routes, AI SDK integrations — strict mode always on.

For ML pipelines, scripting, and tools like DreamPixel Forge. Also the backbone of most MCP server prototyping.

App Router, server components, API routes — my go-to for every web project. This site runs on Next.js 16.

React 19 with server components. I prefer composition over complexity — context for state, RSC for data.

Utility-first CSS. Custom design tokens for every project, no component libraries. Tailwind v4 on this site.

Unified interface for streaming LLM responses, tool calling, and structured output. Powers the terminal chat on this site.

Auth, Firestore, and hosting for most of my SaaS projects. Fast to ship, scales well enough for the 0-to-1 phase.

Deploy target for all Next.js projects. Preview deployments and edge functions make iteration fast.

API for running open-source ML models. I use Flux Schnell for image generation in several projects.

My main LLM for coding, writing, and thinking. Opus for hard problems, Sonnet for speed. Claude Code is my daily driver.

GPT-5 powers the chat on this site. I also use the embeddings API for RAG pipelines.

Google's model — I use it in MCP servers for multimodal tasks like YouTube video analysis and image generation.