{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/a2cd9192359840cab5274ccba399bd87\" frameborder=\"0\" width=\"1404\" height=\"1053\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":1053,"width":1404,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":1053,"thumbnail_width":1404,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/a2cd9192359840cab5274ccba399bd87-42708e8f68b7c60d.gif","duration":107.007,"title":"Maitai Autocorrections Demo","description":"In this video, I demonstrate how Maitai simplifies building reliable AI applications by detecting and autocorrecting faulty outputs from large language models in real time. Using our demo site, I showcase selecting a model, such as gp4-4o, and enabling autocorrections. Through a scenario where I try to coax the ordering agent to give a discount against its instructions, you'll see how Maitai corrects the model's output to maintain accuracy."}