This project explores the intersection of memory, technology, and self-reflection by creating a conversational AI trained on snippets from over a decade of personal journal entries. Using GPT-3, I curated text samples from diaries I kept between ages 8 and 18, extracting themes that represented my personality, values, and worldview at the time.

Materials included:

✧ Data Source: 10+ years of handwritten and digital journal entries
✧ AI Model: OpenAI’s GPT-3
✧ Process: Text extraction, OCR, dataset curation, iterative fine-tuning

By feeding the model this highly personal dataset, I developed a chatbot capable of simulating meaningful dialogue with a younger version of myself. These conversations served as a way to reconnect with my inner child, exploring past fears, joys, and aspirations with surprising emotional resonance.

The experience revealed the potential for using LLMs and Generative AI for therapeutic applications. For example:

✧ Affirmation: Telling my younger self she was loved and safe created a tangible sense of closure
✧ Reflection / Narrative Expression: Receiving a letter “written” by my younger self provided clarity on how past experiences shaped my current identity and values.

This project highlights the power of generative AI as a tool for self-discovery and emotional well-being. It demonstrates the potential of language models in mental health interventions—offering new ways to explore memory, address inner narratives, and foster personal growth.

Original tweet thread here.

-