This startup aims to transform the entire internet into a searchable database

A new startup called Exa is developing an innovative search engine that aims to transform how we find and organize information online by treating the entire internet as a structured database rather than just a collection of web pages.

The technological breakthrough: Exa’s new search engine, Websets, leverages large language models to encode web pages into “embeddings” that capture their underlying meaning rather than just matching keywords.

• The technology creates a semantic understanding of web content, allowing for more precise and relevant search results
• Unlike other AI search engines that simply layer language models over traditional search, Exa has rebuilt the search architecture from the ground up
• The system predicts relevant links by understanding the relationships between different pieces of content across the web

Primary use cases and target audience: Websets is designed specifically for power users seeking structured information that traditional search engines often struggle to provide effectively.

• The platform excels at generating comprehensive lists of companies, people, or other entities that share specific characteristics
• Users can make complex queries that would typically require multiple searches or manual filtering on conventional search engines
• The results are presented in a more organized, database-like format rather than as a simple list of web pages

Current limitations: Despite its innovative approach, Websets faces several significant technical constraints in its current form.

• Search queries can take minutes to process, significantly longer than traditional search engines
• The platform currently indexes only about 1 billion web pages, compared to Google’s trillion-page index
• The system is still being refined and improved as the technology evolves

Market positioning and potential: Exa is positioning itself as a complement to rather than a replacement for existing search engines.

• The company acknowledges its current limitations while emphasizing the long-term potential of its embedding-based approach
• The technology could potentially reshape how specialized searches are conducted in professional and research contexts
• External experts see promise in the approach but note that scaling will be crucial for widespread adoption

Looking ahead: While Exa’s ambitious approach to search represents a significant departure from conventional methods, its success will likely depend on its ability to overcome current performance limitations while maintaining its precision advantage over traditional search engines.