{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/3048db34600845ca845e7b5ca020595c\" frameborder=\"0\" width=\"1280\" height=\"960\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":960,"width":1280,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":960,"thumbnail_width":1280,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/3048db34600845ca845e7b5ca020595c-1dec27d8be76a106.gif","duration":217.866667,"title":"Voice agent prompt formatting matters a lot more than you think","description":"This Loom demonstrates how prompt formatting affects the audible performance of a voice agent. The author compares a well formatted prompt (with markdown structure) versus an unformatted version where all formatting markers are stripped, while keeping instructions, model, and voice identical. The formatted prompt handles stacked input cleanly, acknowledges each piece in order, deflects diagnosis appropriately, and follows the flow, while the unformatted prompt drops pieces, engages diagnosis-related content, misorders acknowledgements, and sounds rushed or confused. The author warns that issues become more noticeable with larger prompts exceeding 3500 tokens, emphasizing that prompt structure matters beyond using markdown."}