Ammonix Launches a New AI Architecture for High-Stakes Specialist Agents That Outperform Frontier Models at a Lower Cost

BLONAY, Switzerland and WILMINGTON, Del., Sept. 22, 2026 /PRNewswire/ -- Ammonix today emerged from stealth with a new AI architecture that combines a fast, specialized System One decision-making architecture with a System Two language model for reasoning and interaction.

The architecture is designed to create specialist AI agents that learn from a company's operational data and workflows. It enables fast responses, auditable recommendations, control over sensitive data, and predictable computing. Within broader agentic systems, Ammonix agents can serve as intelligent control layers for critical functions, making AI systems safer, faster, and more reliable.

The company is led by co-founder and CEO Peter Ruppersberg, a scientist and serial entrepreneur whose medical AI company, Cortex, was acquired by Boston Scientific in January 2025 for up to $300 million.

Unlike general-purpose AI, Ammonix agents are built around specific principles: they learn from an organization's data, reach expert-level performance from a few hundred examples, and improve over time by updating their memory. Every recommendation is fast and auditable, with supporting examples available for verification. When an agent does not know, it asks for human input rather than guessing, an approach designed for high-stakes applications. By capturing expertise once and reusing it across tasks, Ammonix agents can run on local hardware, dramatically reducing compute requirements and costs while enhancing data privacy, security, and operational control.

Ammonix demonstrates its architecture across four applications: ECG interpretation, healthcare claims collection, waste-to-energy control-room support, and applied computer vision. In these applications, the agents matched or outperformed OpenAI's GPT-6 at lower cost and significantly faster speeds.

"Enterprise AI cannot simply be another expensive model with more parameters," said Peter Ruppersberg. "Companies need agents built around their own operational requirements that respect boundaries, provide verifiable evidence behind their conclusions, and remain economically deployable."

Peter Ruppersberg began his career in a Nobel Prize-winning research team and later became a full professor and department head, publishing in Nature and Science and filing numerous patents. He subsequently founded several companies, including Mentasys, an AI product-search pioneer acquired by Lycos in 2006.

Ammonix will soon release AmmonixCode, a source-available coding platform for building agents with the Ammonix architecture. It will be free for research and evaluation, with commercial and enterprise licenses available at [email protected].

To learn more, visit www.ammonix.ai.

Media Contact: [email protected]

Photo - https://mma.prnewswire.com/media/3009048/ammonix.jpg
Logo - https://mma.prnewswire.com/media/3009049/ammonix_Logo.jpg