{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"The OPTIM Update","title":"Useful Now: The Case for Application-Specific Robots | Arjun Subramaniam of Factory Intelligence","html":"<iframe width=\"100%\" height=\"180\" frameborder=\"no\" scrolling=\"no\" seamless src=\"https://share.transistor.fm/e/a1912def\"></iframe>","width":"100%","height":180,"duration":2535,"description":"Arjun Subramaniam is the founder and CEO of Factory Intelligence - a physical AI company training tactile foundation models for industrial manipulation. He's toured 70+ factories, deployed robots on real shop floors, and is making the contrarian bet that application-specific systems beat humanoids and general-purpose foundation models right now. His first workcell has eight robots building electrical outlets for $3/hour.\nWe cover:\n00:00 - Intro\n00:44 - What 70 factory visits taught him about deployment vs. demos\n02:47 - No SLA in a research paper - why factories are a different game\n04:23 - Why he put a packaging machinery veteran in the COO seat\n06:34 - The \"Useful Now\" thesis and where the robotics narrative is wrong\n08:53 - The Tesla vs. Waymo parallel for robotics\n10:01 - You can't buy your way into a large enough manipulation dataset\n10:27 - Why vision alone isn't enough for industrial tasks\n12:54 - The pen-in-a-bin problem: why vision-only models are too slow\n14:37 - Why robotics is not like LLMs - there is no single scaling law\n16:32 - The application-specific full-stack quadrant: why no one else is here\n17:12 - Best version of the model-first argument - and how he pushes back\n19:50 - What happens to humanoids if \"Useful Now\" works\n21:56 - Inside an electrical prefab shop - what actually happens in there\n23:53 - Prefab-Cell-E1: eight robots, $3/hour, 9x productivity\n24:44 - What \"tailing an outlet\" means - the actual task, step by step\n28:01 - Wire-bending model generalizing to colors it was never trained on\n29:16 - The integration trap: why custom fixtures wreck margins\n31:29 - When do you know deployment economics actually work\n32:08 - The data flywheel: why 50% success rate is the threshold\n33:29 - Touch is filling the gap where vision saturated\n35:14 - Combining neural nets with classical control - and why both matter\n37:44 - The world action model: image, proprioception, tactile, action, all in\n39:39 - You can't buy your way to multimodal data from...","thumbnail_url":"https://img.transistorcdn.com/MD1uw3xsUV8NIC9WNTbAWnAxAqctNuVEgSaMzKwMDRQ/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9jYTFk/YTA0N2ZmOThmMmZl/ZGVjMjA3NWMwZDUw/YmRlMS5qcGVn.webp","thumbnail_width":300,"thumbnail_height":300}