The success of LLMs, driven by the "bitter lesson" that scale is paramount, isn't unique to language. The same principles—pre-training, post-training, and reinforcement learning—can be applied to search models to achieve breakthrough performance in information retrieval.
AI agents, unlike humans, need complete and exhaustive information (thousands of results) and use complex, controllable queries. A search engine built for human keyword simplicity and limited results will fail to serve them effectively.
With model intelligence advancing, the next hurdles for perfect search are operational. First, building infrastructure to handle a 1000x increase in agent-driven queries. Second, the "data bottleneck" of capturing and indexing vast information that exists only offline.
The most efficient AI architecture separates reasoning from knowledge. Models will shrink, focusing parameters on intelligent processing, like an "Einstein who never saw the world." They will rely on cheap, efficient tools like retrieval for information, solving compute shortages.
Instead of using massive, expensive LLMs for every task, companies can solve the "tokenpocalypse" (runaway token costs) by pairing smaller models with high-quality retrieval systems. This allows cheap models to act like large ones, saving significant costs.
Exa's CEO reframes complex societal issues as search failures. Political polarization stems from a failure to find accurate, comprehensive information, while loneliness is a failure to find and connect with compatible people. Perfecting search could address these core human challenges.
Inspired by Elon Musk’s "make humanity interplanetary," Exa's CEO emphasizes creating powerful, memorable names for internal projects. A strong name acts as a memetic core, ensuring everyone understands a project's mission, which is critical for alignment as a company grows.
Google's moats (human click data, large re-ranking teams) are less relevant for AI agents. LLMs allow small, agile teams to build superior search products by training their own models without needing decades of user signal data.
While consumer search focuses on common queries, serving thousands of diverse businesses creates constant, varied pressure for higher quality and deeper capabilities. This intense, specific demand from niche use cases is a more powerful driver for cutting-edge research.
