Beyond RAG — Inside Pinecone Nexus and the Knowledge Engine for Agents

Jeff Zhu, VP of Product at Pinecone, joins Alan Shimel on Techstrong TV to dig into Pinecone Nexus and KnowQL — Pinecone’s biggest bet yet on the shift from human-driven retrieval to agent-driven knowledge infrastructure.

Jeff explains why the access patterns for vector search have fundamentally changed since the early days of ChatGPT, why classic retrieval-augmented generation (RAG) starts to break down once agents are firing 20–30 tool calls per task, and how Pinecone Nexus turns raw data into task-optimized knowledge artifacts that agents can act on directly. The conversation also walks through KnowQL, a new declarative query language Pinecone designed for the agentic era, and unpacks the measured impact Nexus is delivering on customer workloads: up to a 90% reduction in token usage and 30x faster time-to-completion.

Jeff and Alan also dive into the “just because I can doesn’t mean I should” era of agentic AI — the move from experimentation to production, the SaaS budget squeeze, and how Pinecone itself is dogfooding Nexus to dramatically reduce internal dashboards and rethink how the business runs day to day.

Learn more at https://www.pinecone.io