You Must Be Some Kind of Therapist

192. Inside Detrans.AI: The Tool Revealing Why People Leave Trans Identity | Peter James Steven

Dec 15, 2025
Peter James Steven, a web developer from Wellington, New Zealand, shares his journey of creating Detrans.AI, a tool that amplifies detransitioners' perspectives. He discusses the ethical choices in building the AI, including its design to avoid biases found in Western models. The conversation reveals critical insights, such as age patterns in transition, gender differences in motivations, and the powerful impact of therapy on detransition. Peter also confronts the 'born this way' narrative and highlights the importance of proper support for individuals questioning their identities.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
ANECDOTE

Personal Catalyst For The Project

  • Peter built detrans.ai after a family member transitioned and he wanted to understand why people transition and detransition.
  • His curiosity about a loved one led to a public project amplifying detransitioner stories and data.
INSIGHT

Using RAG To Preserve Detrans Perspectives

  • Retrieval augmented generation (RAG) lets a chatbot answer using a curated corpus instead of retraining a model from scratch.
  • Peter used RAG to surface detrans subreddit content and avoid mainstream model guardrails.
INSIGHT

Model Choice Shapes Narrative Framing

  • Mainstream models like ChatGPT often enforce gender-affirming narratives via safety layers.
  • Peter selected a Chinese model (Kimi K2) that was more open to gender-critical perspectives.
Get the Snipd Podcast app to discover more snips from this episode
Get the app