Datology webinar: Curation Studio matched top data baselines with under 20% compute

josh_wills · x · 2026-10-27

Datology AI will host a live deep dive of Curation Studio on Oct 27, 10 AM PT, led by CEO Ari Morcos, CTO Bogdan Gaza and the research team, moderated by Josh Wills.

Agenda: behind-the-scenes of 3 years of frontier data research; an end-to-end Curation Studio walkthrough (data registration, compute configuration, synthetic data generation, data mixing); and a deep dive showing data curation that matched the strongest public baselines using less than 20% of training compute, with a check on whether gains held through post-training. Wills previews his skeptical questions: are baselines fair, how do MoE scaling ladders hold up, and what it takes to curate a petabyte-ish of data in your own VPC. Audience questions accepted via replies.

Original post →

More from Apps

Apps channel →