← back
Building a Smarter AI Agent with Neural RAG - Will Bryk, Exa.ai
Takeaway
AI agents need a web-search engine designed for them — high-throughput, query-rich, embedding-native — not Google retrofitted onto an LLM.
Summary
- Bryk recounts founding Exa in 2021 (YC S21) to redesign search by embedding every web document with transformers — the 'bitter lesson' applied to information retrieval, replacing inverted-index keyword search.
- Demo query 'shirts without stripes' fails on Google (returns striped shirts) but works on Exa because embeddings capture meaning, not bag-of-words.
- Argues GPT-4 has ~10TB of weights vs the web's exabytes, so LLMs must search; but legacy search engines were built for slow humans, not AIs that can ingest huge result sets in parallel.
- Positions Exa as 'one API to get any information from the web' purpose-built for agents, not humans.
neural-searchexaretrieval
Original description
RAG quality for AI agents is critical, and traditional keyword-based search engines consistently underperform in agentic or multi-step tasks, where semantic grounding and contextual nuance matter most. In this talk, Will Bryk, CEO of Exa will live code two AI agent applications–one using traditional keyword search RAG and one using neural network RAG via vector search. He’ll then evaluate both applications based on task performance, relevance, and latency. With a live demo (no theory or pre-baked applications), the audience will get a firsthand look at the practical differences between keyword and semantic systems in production, and learn embedding strategies, indexing trade-offs, hybrid retrieval techniques, prompt tuning, and more. About Will Bryk A year before ChatGPT launched, Will was already spending his time building Exa’s API to crawl the web intelligently, focusing on finding quality sources over SEO spam. Backed by NVIDIA and Lightspeed, Exa now powers products for customers like Databricks, Cursor, and LlamaIndex. Recorded at the AI Engineer World's Fair in San Francisco. Stay up to date on our upcoming events and content by joining our newsletter here: https://www.ai.engineer/newsletter