Our research analyzed thousands of life stories to understand how AI models identify pivotal moments. We discovered that they consistently recognize struggle as a catalyst for transformation. Supernova reframes our stories and imagines new timelines.
Brought to life by researchers from Mila and Stanford.
Our Discovery
From 1,000,000+ sentences to 5 turning points
We scraped, cleaned and annotated 617 biographies, crises and Nobel‑winning discoveries. Each article is chunked, embedded and stored in a FAISS index—ready for retrieval and analysis.
Using advanced AI models to analyze hundreds of life stories, news events, and historical events, we found a pattern: different AI models, each with their own "personality", consistently identify different critical turning points. When analyzing the same events, models reveal distinct reasoning patterns and values.
Strategic models: Prioritize outcomes and achievement milestones
Emotional models: Focus on interpersonal dynamics and relationships
Creative models: Emphasize innovation and paradigm-shifting moments
Understanding AI personality patterns helps us build more interpretable and aligned systems.
Scientific Discovery Reasoning Patterns
When analyzing scientific breakthroughs, we discovered three distinct reasoning personalities emerge from advanced models like OpenAI's o3, Gemini 2.5 Pro, and Claude Sonnet 3.7:
Causality-centric: o3 focuses on direct cause-and-effect pathways, emphasizing mechanisms and critical junctures that create cascading effects
Enablement-centric: Gemini 2.5 Pro prioritizes foundational methods, barrier removal, and empirical validation steps that make outcomes possible
Synthesis-centric: Claude Sonnet 3.7 emphasizes conceptual integration, paradigm-level connections, and transformative insights
For example, analyzing machine learning breakthroughs: o3 ranks Hopfield's 1982 energy-based network first for its causal impact, Gemini prioritizes the 1986 backpropagation paper as a key enabler, while Claude focuses on conceptual paradigm shifts across the timeline.
AI Personality in Action
The Chandrasekhar Limit Discovery
At just 24, Subrahmanyan Chandrasekhar was publicly humiliated by the famed astronomer Arthur Eddington, who dismissed his groundbreaking work on stellar collapse. Despite this devastating setback, Chandrasekhar's persistence led to the discovery of the "Chandrasekhar limit"—a revolutionary insight that determines when a dying star will collapse into a neutron star or black hole, ultimately earning him the 1983 Nobel Prize in Physics.
Strategic AI
Prioritizes Nobel Prize (1983) as the defining moment
"Achievement milestones demonstrate tangible success and career outcomes."
Emotional AI
Focuses on Chandrasekhar Limit Discovery as the breakthrough
"The human journey of discovery and foundational scientific understanding matters most."
Creative AI
Emphasizes Philosophy of Systematization as innovation
"Conceptual frameworks and intellectual contributions drive paradigm shifts."
Real-World Impact
Our research reveals that AI models exhibit distinct personality patterns when analyzing complex events. Using our Supernova Event Dataset, we demonstrate how understanding these patterns enables interpretability and human-AI collaboration.
Model Interpretability
Our critical event analysis makes AI decision-making more transparent by revealing consistent behavioral patterns. This interpretability is crucial for safe deployment in high-stakes domains like healthcare, law, and finance.
Model Behavior Understanding
We demonstrate that LLMs exhibit consistent personality-like patterns without explicit prompting—from emotional reasoning in Orca 2 to strategic analysis in Qwen 2.5—advancing our understanding of AI alignment and behavior.
Human-AI Collaboration
Our framework enables better model selection by understanding how different AI personalities complement human expertise. Teams can leverage causality-focused, enablement-driven, and synthesis-oriented reasoning for comprehensive problem-solving.
Human-AI Co-Discovery
In scientific discovery, our research shows that o3 emphasizes causal chains, Gemini 2.5 Pro focuses on methodological enablement, and Claude Sonnet 3.7 prioritizes conceptual synthesis—enabling collaborative research workflows.
Try the Demo
Share a story from your life. The AI model will identify the pivotal turning points and connect you to someone who faced a similar journey.
Our Story
In late December 2023, Pranav, a PhD researcher in Reinforcement Learning at Mila, and Ioana, an astrophysicist and AI enthusiast now working at Stanford, found each other through words of hope exchanged on X/Twitter during their moments of challenge. When the path ahead seemed uncertain, they found hope in each other's stories — and realized AI could help millions of people find the same. Every Saturday for several months, they shaped this insight into research. From this experience, Supernova was born to help people.
The Vision
A world where people discover hope with the help of AI.
The Mission
Empower people to transform their lives by imagining new stories.
Contact Us
We'd love to hear from researchers, mentors, tech innovators, and storytellers everywhere.