Elsewhere

Talks, podcasts, and writing I’ve done with others.

Talks & Podcasts

The Never Ending Lore of Harness

GroundZero podcast, April 2026. A 90-minute deep dive on what a harness is, deepagents, context rot, opinionated agents, and “Harness as a Service.”

How We Built LangSmith Engine

Interrupt 2026, San Francisco. A behind-the-scenes look at building and evaluating LangSmith Engine.

Context Engineering vs Harness Engineering vs Software Engineering

AI That Works podcast (Boundary). Some lore around these three disciplines and where the industry is heading.

Selected Joint Blogs

Some selected posts co-authored at LangChain.

With the Harvey team, June 2026. Batching verifiers and using open models cuts verification cost by an order of magnitude on Harvey’s LAB legal benchmark.

Building a 100x Cheaper Trace Judge with Fireworks

With Fireworks, June 2026. Fine-tuning a small open model to judge production agent traces at frontier-level accuracy for a fraction of the cost.

Tuning Deep Agents to Work Well with Different Models

With Mason Daugherty, April 2026. Per-model-family harness profiles yielding 10-20 point gains over the default harness.

Better Harness: A Recipe for Harness Hill-Climbing with Evals

April 2026. A recipe for autonomously improving a harness using evals as the learning signal.