The platform is live. A private workspace for every engagement, and an agent that answers from your project or says it does not know.

Interactive, and updated as I build

The Lab.

Where I build the thing before I write about it. Interactive tools, working notes, and the architecture I actually use in production. Made to be useful, not frequent.

Tools

Interactive tools

Playable, not just readable. Drill the decisions I make on real systems.

surfaces

Live · Product · workspace

The Platform

A private, grounded workspace for every engagement. Upload documents, connect repositories, keep the decisions, and ask an agent that answers only from that context, with citations, or says it does not know. Callable by your own agent over MCP.

Grounded RAGGitHubMCPMulti-tenant

Try it

checks

Soon · Interactive audit

Trust Audit

Describe your AI system, or point at a repo, and get an honest read on whether it can be trusted in production. Graded on grounding, prompt-injection, data ownership, and the hardening basics.

GroundingPrompt-injectionOwn your context

Coming soon

demos

Live · Architecture essay

An agent proposes. A verifier and a human decide.

Why naive RAG fails in production, and the architecture I use instead. Grounded in a real build, with the pipeline, the claim store, and the gate you can play with.

MastrapgvectorClaude

Try it

demos

Live · Architecture essay

Graph engineering: draw the shape of the work

Why a single agent runs in a straight line, and the graph I draw instead. Fan out, verify, converge, with a topology you can morph, a router, a verifier, and a loop that stops.

Claude CodeSubagentsAdversarial verify

Try it

scenarios

Live · Interactive drill

Agentic AI in production

24 real architecture scenarios with the answer I would give and a diagram for each. Reveal, drill, shuffle.

TypeScriptServerlessEdge-native

Try it

tools

Live · MCP server

Talk to my site

An MCP server my work answers questions through. Add it to your AI client and ask what I do. It answers from cited facts, and refuses when it does not know.

MCPStreamable HTTPRetrieval-first

Try it

On the bench

Coming next

01Soon

Notes

Short, raw learnings from real builds. The fast stuff.

02Soon

Guides

Build-faster workflows, start to finish, the way I actually run them.

03Soon

Resources

Templates, starters, and the founder's playbook.

Building something where this matters?

This is how I think on real systems. If you want that judgment pointed at yours, let us talk.

Let's talk