{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"Travel Tech Podcast","title":"Above China, Your Starlink Doesn't Exist","html":"","width":"100%","height":180,"duration":3422,"description":"The most reliable AI on your next flight might be the one that assumes it has no internet connection at all.\nAjith Balakumar is the CEO and founder of Forthcode, an aviation technology company that grew out of a point-of-sale system built for a 2,500-outlet coffee chain in India. The conversation covers how an offline-first architecture built for retail became the foundation for in-flight commerce, catering, loyalty, and an AI ground-staff agent called Davix, and how Ajith thinks about AI maturity, regulation, and building software at founder speed.\nWhat You'll Learn\nOffline-first architecture: In-flight software can't assume connectivity, because satellite access can be blocked by entire governments regardless of available bandwidth.\nAI trust: Ground staff trust an AI's answer more when it cites the exact SOP page and paragraph behind it, not just when it gives a confident response.\nAI maturity: Organizations move through three stages: co-pilots where the human decides everything, agents where the human approves a subset of decisions, and operational intelligence where multi-agent systems feed executive decision-making.\nRegulatory readiness: GDPR already applies to any European passenger's data, and standardized AI compliance layers, similar to an API gateway, are emerging from major cloud providers.\nData sovereignty: Forthcode avoids routing airline data through frontier LLMs, preferring locally hosted models so information never leaves the airline's environment.\nAdoption pattern: Most airlines are still running isolated AI pilots rather than operational AI, because executives want quantifiable ROI before scaling.\nBuilder speed: AI-assisted prototyping now lets a founder build and test a fully functional application in about an hour, compared with days of spec-writing and revision cycles.\nPrompting as a diagnostic: How someone prompts a general AI tool reveals their own problem-solving process, which is part of why teams with AI tools still underuse them....","thumbnail_url":"https://img.transistorcdn.com/LxpvuNpWwfSGFL1KA1WhoZf9L55ykAqb5rgjXNFqi3c/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9mY2Yz/ZjA5OGE1ZmEyMTk4/ODJkYmU1YjhlYjRk/YTMzNC5wbmc.webp","thumbnail_width":300,"thumbnail_height":300}