Our Story

Built between
New York
and Hawaiʻi.

A father-daughter team combining decades of production engineering with institutional finance and frontier research. We were building the intelligence layer before it had a name.

Paniolo is the Hawaiian word for cowboy — ranchers who herded cattle across the slopes of Mauna Kea since the 1830s. We herd your coding agents with the same discipline: structured, purposeful, and built to endure. Rooted in Honolulu, operating from New York.

The Founders

Two perspectives.
One infrastructure thesis.

Co-Founder & CTO

Ben Kinsey

Over 15 years building production-grade TypeScript and React systems at scale. Ben led engineering teams across staffing, fintech, and enterprise software — and spent years watching the industry adopt AI tools without the discipline to make them reliable. Long before harness engineering had a name, he was architecting the intelligence layer: skills, custom agents, lint rules, deep-linked documentation, and structured tooling that made AI output converge on professional standards. Paniolo is that methodology, productized.

15+ Yrs Production Engineering TypeScript · React · Node · GraphQL XTIVIA · ALTRES Harness Engineering Pioneer

Co-Founder & CEO

Anabel Kinsey

U.S. Presidential Scholar. Economics at Columbia University, perfect 4.0 GPA. Incoming quantitative research intern at Millennium Management. Anabel built AI knowledge systems at Prudential Financial, presented investment strategy at the Clinton Global Initiative, and supported LP fundraising at Chai Ventures as an Investor Relations Fellow. She brings the analytical discipline of institutional finance and the conviction that infrastructure — not models — is where enterprise AI value is captured.

U.S. Presidential Scholar Columbia University · 4.0 GPA Millennium Management Clinton Global Initiative Chai Ventures · IR Fellow

What We Believe

The principles that
define the work

Every AI error is infrastructure debt.

When an agent makes a mistake, the right response is not to fix the output and move on. It is to update the harness so the same failure is structurally prevented. Errors are signals. Signals become infrastructure.

Harness quality is model-agnostic.

A well-engineered harness outperforms a better base model with a weak one. Empirically validated at ICLR 2026. The infrastructure layer is the lever — not the model.

Observability before optimization.

You cannot reliably improve what you cannot see. Every component, trajectory, and decision must be auditable before any evolution loop can be trusted. Structure precedes speed.

Build agents that improve with experience.

CLI early access — limited seats