Blog
Running six coding agents at once is easy. Getting six pieces of code that actually fit together is the hard part. Here's how I think about git worktrees, Conductor, planner and builder agents, feature branches, and building in waves.
I'm speaking at two Boston AI Week events: a hands-on fine-tuning workshop and a practitioner talk on running parallel coding agents with Conductor.
OpenAI just shipped WebMCP support in the ChatGPT desktop app, so I added six tools to this site. Your agent can now search my blog, check my speaking schedule, and even restyle the page while you read it. Here are prompts to try.
Context is a finite resource, and in a web or mobile app you're the one deciding how to spend it. Here's how I think about curating context in apps people actually use.
The world's first Generative UI web browser repaints every page you visit in real time as felt, clay, carved wood, paint, or whatever material you choose.
A start-to-finish walkthrough of fine-tuning an open-weight model on home-lab GPUs: baseline evals, synthetic data, QLoRA, quantization, and serving.
How WGHTLFT pairs a Rails 8 API with a native SwiftUI iOS app for lifting, programs, 3D equipment visualization, Apple Watch, and Siri shortcuts.
A prototype I built went viral. Here's what happened next...
When you’re coding with agents, the most reliable steering wheel is a well-placed comment in the code’s line of sight.
The AI coding era has changed the calculus in ways that are only starting to become clear, and Rails benefits from those changes more than almost any other framework. Convention over configuration turns out to be exactly what a language model needs.
