RM
Ryan Marten
Member of Technical Staff · Laude Institute
Bio
Ryan Marten is building Harbor at the Laude Institute and works on research-to-production efforts including Harbor, Terminal-Bench, and OpenThoughts-Agent.
Session (1)
Additional Speaker Details
Using publicly available information we constructed an analysis to help you get a feel for this speaker before deciding to attend their session.
Marten co-created the OpenThoughts reasoning-data recipes (OpenThinker models) and is now a core builder of Harbor/Terminal-Bench, the Stanford x Laude Institute agent-evaluation framework. His confirmed public talk delivers concrete, empirically-grounded lessons on building large-scale reasoning-data pipelines — teacher-model choice, answer sampling, question filtering — rather than high-level claims.
Speaking style
Deeply technicalBattle-testedClear teacher
Recent talks (1)
GitHub
@RyanMarten- RyanMarten/distributed_gcp_youtube_download★ 9Download YouTube videos faster using a large number of VMs
- bespokelabsai/curatororg★ 1,696Synthetic data curation for post-training and structured data extraction; Marten is first (equal-contribution) author on the repo's citation — a co-creator/lead
- harbor-framework/harbororg★ 2,848Framework from the creators of Terminal-Bench for evaluating and improving agents; Laude Institute's launch tweet names Marten (@ryanmart3n) among its core builders
Recent writing (3)
- OpenThoughts: Data Recipes for Reasoning Models · paper · Jun 2025
- Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces · paper · Jan 2026
- OpenThoughts-Agent: Data Recipes for Agentic Models · paper · Jun 2026