video1.0<iframe src="https://www.loom.com/embed/50501cf31bad4cb48116404f4bdb7bda" frameborder="0" width="1920" height="1440" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>14401920Loomhttps://www.loom.com14401920https://cdn.loom.com/sessions/thumbnails/50501cf31bad4cb48116404f4bdb7bda-a45091704c218a4f.gif83.051This Loom introduces Omnicon, a system for automatically rewriting and benchmarking summarizers across models, communities, and use cases. It explains that summaries currently lack transparency and shared improvement, with people making and fixing the same mistakes manually. Omnicon Gold, created two weeks ago, helps create “golden records” for summaries as they should be, and Omnicon RSI at the Scribble Valley Hackathon uses Flash for summary generation and Probe for judging to recursively improve instructions by testing candidate changes and climbing to the best performers. It concludes with a coming soon plan to pull community data and benchmark summarizers and judges together, similar to Hugging Face.