Philosophy

I build with AI systems. Not demos — systems. The kind that run in production, process real data, and solve actual problems.

The “AI Alchemist” label isn't metaphor for mysticism. Alchemy was the predecessor to chemistry — people experimenting with combinations of materials, trying to understand transformation at a fundamental level. Some of it worked. Most of it didn't. The ones who succeeded were the ones who documented their experiments honestly.

That's what this site is. Honest documentation of what works in AI tooling right now. The tools that are actually useful, the patterns that actually scale, the workflows that actually save time. No hype. No vaporware.

The substrate is mathematics. The medium is code. The goal is tools that work.

I work across the full AI stack — from model selection and prompt engineering to agent architecture, MCP integrations, and production deployment. I'm particularly interested in the intersection of AI agents and developer tooling: how we can make AI systems that are genuinely useful for building software, not just generating it.

Video Essay

From prompts to operating systems

A James-voiced explanation of the shift from one-off prompting to real AI systems with continuity, skills, routing, and momentum.

Focus Areas

What I Build

AI Agent Architecture

Designing multi-agent systems that coordinate, reason, and execute with reliability at production scale.

MCP Integrations

Building bridges between AI models and real-world tools through the Model Context Protocol.

Production Systems

Shipping AI that runs in production — not demos that break on real data.

Developer Tooling

Making AI genuinely useful for building software, not just generating it.

The Proof

Built With Its Own Tools

This site itself was built with AI agents, using the tools documented in The Manuscript. It's the proof of the pudding — a production system assembled by the very workflows it describes.

Read The Manuscript

Now

AI Alchemist — building at the frontier

Ongoing

Open-source MCP tooling and agent frameworks

Always

Documenting what actually works

Approach

Systems

Not demos

Production

Not prototypes

Honest

Not hype

Stack

TypeScriptNext.jsPythonLangChainMCPOpenAIAnthropicVercelSupabasePostgreSQL

Let's build something real.

If you're working on AI systems that need to actually ship, I'd like to hear about it.