In a technical talk on The MAD Podcast with Matt Turck, Slingshot AI co-founder and CEO Daniel Cahn opened the hood on how Ash's foundation model for psychology is built. He framed the need around a 60-year-old shortage, roughly 1,500 people in need for every therapist in the US, and noted that therapy has become AI's top use case, mostly via general chatbots that are "not very good at it." Cahn described training Ash in three stages: pre-training on a large, ethically sourced dataset spanning modalities like CBT, DBT, ACT, IFS, and psychodynamic therapy; alignment by a full-time clinical team that adds guardrails; and reinforcement learning that draws on rich conversational signals. He explained that Slingshot optimizes toward four sequential goals, intent, therapeutic alliance, psychological change, and behavioral change, and shared a first study with NYU showing gains across them.