Last updated · May 25, 2026

What I'm focused on

A snapshot of where my attention actually is this month — not a roadmap, not a plan. Inspired by Derek Sivers' /now page convention. Refreshed roughly monthly.

Learning

Deep LangGraph and agentic orchestration patterns — building toward controllable multi-agent workflows with checkpoints and explicit handoffs. ~6 hrs/week. Pairing with the open-world evaluation literature.
Active DSPy — testing whether prompt-as-software actually scales beyond toy examples.
Light vLLM deployment patterns — inference, batching, latency/cost trade-offs at production sizes.

Building

AI Control Plane — scaffolding the v0.1 milestone: agent registry, execution tracing, signed policy verdicts. Open repo coming as soon as the foundations hold up to a self-review.
This site. Astro on Cloudflare Pages. Built in public, every post is a commit.

Writing

Draft in progress: Why Most Enterprise AI Agents Will Fail in Production. Targeting publication within the next two weeks.
Series-in-planning: The orchestration shape that survives contact — four-post arc on identity, latency, audit, and learned mistrust.

Not focused on

Generic chatbot side projects. That market is saturated and uninteresting.
Certificate collection. Every cycle has to produce a derivable — code, architecture, writing.
Multi-agent "swarm" architectures. Distraction at this maturity level; revisit in 2027.
Advisory work that doesn't put me in front of running production systems.