Latest Insights
Practical lessons from building and scaling technical systems.
Beyond Prompt Engineering: Why Modern AI Workflows Require Orchestration Platforms on Top of Tools Like Claude Code
An interactive AI coding agent that works brilliantly for one developer breaks down the moment you try to run it as an engineering platform. The fix isn't a better prompt or a longer CLAUDE.md — it's a deterministic orchestration layer wrapped around a probabilistic model.
The Execution Layer Is the Trust Boundary: MicroVM Sandboxing for Claude Code
An LLM is just the brain. The shell underneath it is where the actual damage happens. Here's the architecture for running Claude Code with full autonomy — --dangerously-skip-permissions included — without ever exposing the host.
I Built the Agentic Protocol Stack From Scratch. Here's What's Actually Going On.
795 lines of vanilla Python, zero frameworks. I built MCP, AG-UI, A2A, and A2UI from scratch to find out what these protocols actually are under the hood. The answer is less impressive — and more useful — than the marketing suggests.
Want to work together?
I'm always open to discussing technical leadership, mentorship opportunities, or interesting engineering challenges.
Get in Touch