TLDR AI · DEEP DIVES & ANALYSIS
A TAXONOMY OF SELF-EVOLVING AGENTS
Summary from the TLDR AI newsletter
Shilong Liu proposes a taxonomy that classifies self-evolving agents into three categories: artifact optimization, harness self-improvement, and model learning. The framework distinguishes whether evolution occurs in outputs, agent infrastructure, or model weights, providing a common language for emerging agent research.