Memory and self-learning in agents face major unsolved challenges despite attempts using RAG, vector DBs, and long context windows—agents must maintain consistency where it matters while only differing in areas where variation is appropriate.

factualpending

Speaker

Constantine

Evidence Quote

“we've been trying everything from rag and vector DBs to really long context windows, but everyone in this room knows that there are still major challenges in true memory and self-learning on that true memory and actually getting the agents to be able to be consistent where it matters”

Source

AI's Trillion-Dollar Opportunity: Sequoia AI Ascent 2025 Keynote— Sequoia Capital
Created: 8/11/2026, 7:13:14 AM

Your notes

Loading notes...