Reading up on OpenHands
2 deep · digging since feb 17
- [2605.09998] Continual Harness: Online Adaptation for Self-Improving Foundation Agents
Continual Harness enables embodied AI agents to self‑improve online without resets, cutting button‑press costs in Pokémon games and nearing expert‑crafted harness performance.
- How System Prompts Define Agent Behavior
System prompts shape coding agent behavior as much as the underlying model, with swaps between agents like Claude and Codex producing distinctly different workflows.