{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"START","title":"START: Vedant Gaur, CEO & Founder, Latent: \"Stop hallucinations before they reach your customers\"","html":"<iframe width=\"100%\" height=\"180\" frameborder=\"no\" scrolling=\"no\" seamless src=\"https://share.transistor.fm/e/74079474\"></iframe>","width":"100%","height":180,"duration":407,"description":"An AI model can be completely wrong and sound just as confident as when it's right.\nThat's a problem when your customers are relying on its answers.\nAnd catching those mistakes isn't as simple as asking another AI to double-check every response.\nRunning a second frontier model on every output is expensive, especially at scale.\nSo how do you catch mistakes before they reach your customers?\nThat's the problem Vedant Gaur is tackling with Latent.\nVedant compares the technology to a lie detector.\nSomeone can tell a convincing story, but a polygraph looks beyond their words for signals that might suggest something is wrong.\nLatent applies a similar idea to AI.\nIts runtime observability technology examines signals from model behavior to identify potentially incorrect or problematic responses.\nRather than relying solely on traditional output monitoring, Latent can flag suspicious answers, hold them for review, or route requests to a stronger model.\nAnd it can do this at a fraction of the cost of conventional AI evaluation.\nIn Latent's own benchmarks, its system labeled answers 176 times faster and 660 times cheaper than using a frontier model as a judge.\nBut detecting errors is only part of the opportunity.\nLatent also turns model outputs into labeled training data, helping teams identify recurring failures and improve their models over time.\nThe idea grew out of a gap Vedant noticed during years of AI interpretability research.\nMajor AI labs have sophisticated internal tools for understanding model behavior.\nBut developers and enterprises building products on top of those models don't necessarily have access to the same capabilities.\nAnd when something goes wrong, simply trusting the underlying model provider isn't enough.\nCompanies need greater visibility and accountability over the AI systems they deploy.\nVedant began researching large language models before ChatGPT launched.\nNow, as a solo founder in Y Combinator, he's bringing that research to developers and...","thumbnail_url":"https://img.transistorcdn.com/q-E1hh7K6IS4AfZiNy2p4MYVGcUOO8lQP92h8QbOEOA/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS80NzVj/MDEzNjkxNjU1N2Uy/NDFhMDQ3M2ZhNWI3/NWY0MS5wbmc.webp","thumbnail_width":300,"thumbnail_height":300}