LeCun Now Says Building Agentic AI on LLMs Is 'Recipe for Disaster'

LeCun Now Says Building Agentic AI on LLMs Is 'Recipe for Disaster'

Yann LeCun, Meta's Chief AI Scientist, has publicly stated that "building agentic systems on LLMs is a recipe for disaster" — a significant position reversal. Gary Marcus noted the shift: "God, nowadays he sounds exactly like me." LeCun has not publicly acknowledged abandoning his prior positions, which broadly championed large-scale LLM development as the path forward.

Why It Matters

LeCun's reversal carries institutional weight: he oversees AI research at one of the largest AI labs globally. A prominent industry voice now aligning with LLM-architecture critics adds credibility to alternative architectural approaches — and flags further turbulence ahead for LLM-centric product roadmaps.