Definition
Plain language
An early framework that uses AI agents to design other AI agents.
As stated in the literature
Automated Design of Agentic Systems, an LLM-driven search method for evolving multi-agent architectures, one of the prior systems that optimize-anything subsumes under a single declarative loop.
Why it matters: It points toward a future where the architectures of AI systems are themselves found by AI search rather than hand-engineered.
For example, ADAS might evolve a multi-agent debate system by having one agent propose new agent designs, test them, and keep the best.
Heard on the show
“… authors quietly line up five different LLM-based optimization systems — AlphaEvolve, FunSearch, GEPA, ADAS, OpenEvolve — and point out something almost embarrassing. …”Episode 065 — One Loop to Optimize Them All: A Universal API for LLM-Driven Discovery