Writing
- Thoughts on AI Safety
A cautious, nuanced case for AI optimism: why safety, interpretability, bias, and alignment matter as much as raw capability.
- Using Codex from Claude: Getting a Second Opinion from a Different Model Family
I wired OpenAI's Codex into Claude Code as a consulting subagent for architecture reviews and tricky debugging.
- Multi-Model Agentic Coding: Letting the Other Model Do the Typing
Delegate implementation from Claude Code to Codex in an isolated worktree, then review the diff through an approval loop that keeps commits under your control.
- Sentinels: The Quiet Power of a Touched File
How I use sentinel files to gate the risky moves my coding agents make.
- Claude Ultraplan: Planning in the Cloud, Executing Wherever
How Ultraplan moves coding-task planning to Claude Code on the web for browser review, remote execution, or local handoff.
- Playwright vs. Chrome DevTools MCP: Driving vs. Debugging
Compare Playwright CLI, Playwright MCP, and Chrome DevTools MCP to choose browser automation and debugging tools for coding agents.
- Entering the Mind of Ralph Wiggum
How the Ralph Loop gives an AI coding agent a fresh context window for every iteration—and why that changes the workflow.
- Memory Systems for AI Agents: What the Research Says and What You Can Actually Build
A three-axis framework—Forms, Functions, and Dynamics—for understanding and building modern agent memory systems.
- Temporal's Developer Skill Is a Promising First Draft
Review Temporal's Developer Skill for coding agents: a sound foundation with gaps in actionable guidance, language awareness, and verification.
- The Anatomy of an Agent Loop
Explore the loop shared by AI agents and the production concerns around it: context compaction, loop detection, cost budgets, and graceful termination.
- Agent Skills, Stripped of Hype
Agent skills are a context-management strategy for routing tasks progressively and reducing cognitive overload.
- Designing a Build System That Runs Untrusted Code
A technical walkthrough of building systems that run untrusted repositories, covering isolation, dependencies, network access, artifacts, and operations.
- Designing an AI Gateway and Durable Workflow System
Design an AI gateway for routing, policies, and cost control alongside durable workflows that survive failures, pause for approval, and replay.
- MCP Apps and the Missing Middle of AI Tooling
MCP servers return data. MCP Apps let them ship a UI alongside that data—so the tool author, not the client, decides how results look.
- My Ridiculous AI-Assisted Development Workflow
A walkthrough of the system I use to ship code with AI agents—from planning in Linear to worktrees, linting gauntlets, and a small army of code review bots.