Build AI that
stays with people.
Willow started with one very normal frustration: AI could sound brilliant and still forget what we had already figured out. So we built the part that keeps the thread: what happened, why it mattered, and what should happen next.
I never set out to build an AI company.
I was simply trying to understand why AI lost the thread.
During one of the hardest periods of my life, I relied on AI to help me think, organize my thoughts, and find clarity. The longer our conversations became, the more I noticed something that felt fundamentally wrong.
The AI wasn't just forgetting information, it was gradually losing continuity. Goals shifted, constraints disappeared, and the conversation slowly drifted away from what mattered.
I couldn't accept that as an unavoidable limitation.
That question led to years of independent research into continuity, coherence, and inference-time reasoning. The result became Willow, a continuity layer designed to help AI stay aligned over time, giving developers a practical way to build systems that remain coherent, trustworthy, and reliable across long-running conversations and workflows.
Today, the mission is simple: make AI feel less disposableand more human.
Rigor by design.
Every claim, stress-tested.
Soma is a theoretical physicist and researcher specializing in complex system modeling and computational validation.
As Willow's Technical & Scientific Advisor, Soma designs rigorous stress-test protocols, including our “Physics Trap” benchmarks : to evaluate LLM drift reduction, temporal anchoring, and multi-model stability in production environments.
Her work ensures every Willow benchmark traces back to a reproducible test, not a marketing claim.
A continuity layer for every model you ship.
Provider-agnostic. Inspectable. Engineered for production.