vtrivedy

Continual Learning, Harness Engineering, and Post-Training all boil down to the same substrate: curating data at scale to run experiments and improve agents.

Agent = Model + Harness.

Our coding agent went from Top 30 to Top 5 on Terminal Bench 2.0.

The best agent products aren't the most flexible, they're the most opinionated.

Understanding how modern planning agents work as just-in-time, adaptive workflow generators and their role as universal bridge nodes in workflow builders.

A guide to the built-in tools available in Claude Code, including parameters, usage patterns, and examples.

A running list of helpful resources for building agents.

As tasks require more autonomous behavior, we're seeing a shift from LLM APIs to Harness APIs.

Quick intro to what I'll be sharing here.