Professor builds AI Team Matchmaker: 6,000 swaps in milliseconds vs 30-minute card sort
jonippolito · x · 2026-09-29
UMaine professor Jon Ippolito rebuilt his students' notched-index-card team-sorting system as the free, offline Team Matchmaker app using Claude Fable: where a manual sort tested a dozen swaps in 30 minutes, it now runs 6,000 swaps on 50 students in milliseconds, balancing shared interests, overlapping schedules, and role diversity. Motivated by chatbots isolating students, the post also surveys similar efforts: PAIRR, Sway, and CollaborAITE.
More from AGI Musings
- RLVR Can't Prove What Its Verifier Can't: The ZFC Ceiling Argument — Liu_eroteme · 2026-09-29
- Post-training has a ceiling too: AI will match, not surpass, the best humans — Liu_eroteme · 2026-09-29
- Should the Ousterhout Dichotomy apply to AI models? Fixed tasks favor customized models — josh_wills · 2026-09-29
- Erik Hoel: physical systems have astronomically many description levels, none fundamentally different — erikphoel · 2026-09-29
- Engineering Deans Built Silicon Valley's Ecosystem — Could That Happen Today? — Afinetheorem · 2026-09-29
- Corporate workers treat AI as half the job while Reddit insists AI is useless — yalag · 2026-09-29