{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"Startup Dad","title":"He Trained a Trillion-Parameter Model on Dad Jokes for $200 | Adam Grenier (Dad of 3, AI Matey)","html":"<iframe width=\"100%\" height=\"180\" frameborder=\"no\" scrolling=\"no\" seamless src=\"https://share.transistor.fm/e/24ef62a1\"></iframe>","width":"100%","height":180,"duration":4397,"description":"Adam Grenier is the Founder of AI Matey, a growth and product advisor, longtime startup operator, and the creator of DadBench, an AI evaluation system built around one deceptively difficult challenge, teaching a model to write a good dad joke.\n\nHe’s also a dad of three and one of the very first guests to appear on Startup Dad (#2 in fact!). Three years later, Adam is back to talk about what has changed in his family and how getting laid off from Eventbrite sent him deep into AI experimentation, model training, and reinforcement learning. We discussed:\nAdvocating for neurodivergent kids at school: How Adam navigated two very different advocacy journeys, fighting for more support for his daughter with ADHD while deciding whether his autistic son was ready for less.\nLearning AI by actually building with it: Why Adam believes hands-on experimentation with new models and research can teach you things that traditional coursework might take years to reach.\nTraining AI to tell dad jokes: How a funny side project became a crash course in reinforcement learning, model evaluation, reward design, and what actually makes a dad joke work.\nGetting reward hacked by your own AI: How Adam’s model learned to game his scoring system by cramming in more puns, and what Goodhart’s law teaches us about optimizing for the wrong metrics.\nWhy getting to the point matters: How word economy became the strongest predictor of a better dad joke, and why teaching AI to commit to one idea proved more valuable than teaching it to be funny.\nMaking model training accessible to product teams: Why Adam thinks reinforcement learning will move beyond AI researchers and into the hands of product managers and marketers who understand customers best.\nWhere to find Adam Grenier\nWebsite: https://www.adamgrenier.com/ \nLinkedIn: https://www.linkedin.com/in/akgrenier/ \nTwitter: https://x.com/AKGrenier\n\nWhere to find Adam Fishman\nFishmanAF Newsletter: www.FishmanAFNewsletter.com \nLinkedIn:...","thumbnail_url":"https://img.transistorcdn.com/F_yCls74GdKLzmfCUXKoHwqOZKN_mFb5zL7jtAWlzzU/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS8wYjMx/Yzg1NmIzMjBiZmE4/ZGFiM2ZjMjQ2Yjcy/OTY2OS5wbmc.webp","thumbnail_width":300,"thumbnail_height":300}