My Stack
Tools, software, and services I use every day for AI systems, automation, and full-stack client work. Opinionated, battle-tested, and updated as my setup evolves.
Development
My primary editor. AI-native IDE built on VS Code — pairs brilliantly with good system prompts and .cursorrules files.
Fallback editor for quick edits, remote SSH sessions, and projects where I want zero AI interference.
My default runtime for backend APIs, automation scripts, and AI agent orchestration.
Full-stack React framework powering this site and most of the web apps I build for clients.
Utility-first CSS. I stopped writing vanilla CSS years ago and haven't looked back.
Component primitives I copy into my projects and own outright. Radix-based, beautifully accessible.
AI & Automation
Primary LLM for production AI systems — best-in-class instruction following and structured output.
GPT-4o for multimodal tasks, embeddings API for semantic search and caching layers.
Self-hosted automation platform. I use it for complex multi-step workflows where Zapier falls short.
Orchestration layer when I need stateful multi-agent graphs with proper memory management.
Managed vector database for RAG pipelines in production. Fast, reliable, and the ops burden is near zero.
Semantic caching layer on top of LLM calls — saves significant API spend on repeated or near-duplicate queries.
Infrastructure & DevOps
Where most of my Next.js projects land. Zero-config deployments, edge functions, and the DX is hard to beat.
Instant Postgres databases, background workers, and long-running Node processes. Cheaper than Heroku, better DX.
DNS, CDN, R2 for object storage, and Workers for edge logic. Handles most of my infrastructure layer.
CI/CD. Lint, test, type-check, and deploy on every push. Nothing fancy — reliable pipelines beat clever ones.
Productivity & Research
Replaced Spotlight entirely. Clipboard history, snippets, and quick scripts save me hours every week.
Project tracking, client notes, and knowledge base. I keep it simple — a few databases, no overcomplicated views.
AI-powered terminal. Command suggestions, shareable runbooks, and a history search that actually works.
GUI for Postgres, MySQL, SQLite. Fast, native, and handles multiple connections cleanly.
Hardware
16-inch, 36GB RAM. Runs local models, Docker, and 20 VS Code windows without breaking a sweat.
Single external display. Enough screen space without the visual chaos of a multi-monitor setup.
Building something similar or looking for advice on which tools fit your stack? Get in touch — happy to talk through it.