The Intricacies of Open-Ended Data

​Join us in welcoming Joe Hakim, Head of Operations at Fig, for a live research talk on the intricacies of open-ended data, what makes real work intractable to learn from, and how Fig is developing new approaches to environment creation.

​Computer-use data today is built around predefined tasks on clean desktops, each with a clear objective and a defined end. Real work unfolds as unscripted, multi-objective workflows, with independent goals interleaved, interrupted, and resumed on desktops that keep evolving.

​We recently released RealCUA, a rapidly growing dataset of open-ended computer-use trajectories with unscripted work recorded on real desktops that never reset. It spans over 600 hours across 35 apps, with interwoven workflows that run for thousands of steps and continuous trajectories over 12 hours long.

​We will cover

​How computer-use data has evolved, from better observability to wider distribution

​The hill to climb: learning signals for navigating real environments toward open-ended goals

​Capturing those signals with RealCUA: 600+ hours of real work and what each trajectory contains

​What we observed: long continuous sessions, inherited complex workspaces, and interleaved apps

​How Fig is developing new approaches to environments, turning real, open-ended work into signal agents can learn from.

​The talk will be followed by an open Q&A and discussion.

​Links

​Explore the RealCUA dataset: data.fig.inc/realcua

​Read about release: 600 Hours of Real, Open-Ended Computer Work

​Explore opportunities at Manifold: https://www.manifoldrg.com/opportunities/

​Join the Advancement Network: https://www.manifoldrg.com/advancement-network/

​Subscribe to our Luma calendar for future talks: https://luma.com/manifold