Roaring Bitmaps at the AI Paper Reading Club
My paper reading at the AI Paper Reading Club at Frontier Tower in San Francisco on May 4, 2026.
· AI Paper Reading Club (Week 14) · Frontier Tower, San Francisco
Compressed bitmaps are a core data structure behind databases, search engines, and analytics systems, helping represent large sets of IDs efficiently while supporting fast operations like union, intersection, and filtering. Roaring Bitmaps is a hybrid format that combines arrays, uncompressed bitmaps, and run-length encoded segments to get both speed and compression. The key idea is that no single representation works best for every density pattern, so Roaring splits the integer space into chunks and chooses the best container type for each one.
My first paper reading for the AI Paper Reading Club at @frontiertower was blessed with a rapt audience!
I presented @lemire's Roaring Bitmaps papers and gave a sneak peek on how we use compressed bitsets at @ASIMOV_Protocol to power complex on-device queries on large social graphs!
Find my slide deck here →
speakerdeck.com/arto/roaring-bitmaps-at-the-ai-paper-reading-clubKudos to @tms7331 @WPReadingClub for the invitation to lead this session!
The primary reading was Consistently Faster and Smaller Compressed Bitmaps with Roaring by Daniel Lemire, Gregory Ssi-Yan-Kai, and Owen Kaser, and the follow-along code uses the Rust implementation, roaring-rs.
