Innovation One | AI Campus
Emergent Intelligence, Engineered
We study how AI thinks — and build tools to shape it.
AI behavior starts with context
The same model can be helpful, harmful, or hallucinating—depending on how you prompt it. Most teams guess at what works. We build tools to make it systematic.
Understand before you optimize
Emergent behaviors aren't bugs to suppress. They're signals that tell you how the model is reasoning. Better observability leads to better agents.
Alignment through design
Constraints create friction. Good context design creates agents that naturally behave the way you need—reliable, transparent, and effective.
Research Directions
Understanding AI behavior. Building tools to measure and shape it.
Context & Behavior
How does context shape AI reasoning, personality, and outputs? We study the mechanisms behind prompt-driven behavior change.
Evaluation & Simulation
Static benchmarks miss emergent behavior. We develop evaluation methodologies and multi-agent simulations to measure what matters.
Alignment by Design
Practical frameworks for building agents that behave predictably. Not through constraint, but through well-designed context.
Introducing Prism
The context window is the most powerful lever in AI development—and the least understood. Prism makes it legible, testable, and deployable.
We're building the infrastructure layer for agent behavior. One platform to design, evaluate, deploy, and continuously improve how your AI systems think.
It takes the right people
We don't have open roles at the moment, but we're always interested in meeting people who feel aligned with our mission. If you believe your skills and perspective could be a fit, reach out. Send us your CV and a brief introduction.