Confidential Briefing

Prepared for the OpenAI Foundation

Confidential

Nature-made antibodies cleared 28% of brain plaque in 10 days, with zero inflammation.

Demonstrated in Alzheimer mouse models

Join the Abzyme Research Foundation to leverage ground-breaking first-principles AI to optimize this technology, and bring it to the millions of patients waiting for an end to Alzheimer's.

01The Stakes

The most common neurodegenerative disease on earth. And this is the ceiling of medicine.

→ 13M Americans with Alzheimer's · today → 2050
→ $1T+ Annual U.S. cost · today → 2050
40–41% of aducanumab patients developed brain abnormalities (ARIA), vs ~10% on placebo
Anti-tau antibody trials, all failed

The drugs we have fight Alzheimer's by provoking the immune system to engulf plaque, a mechanism (phagocytosis) that brings inflammation with it. The result is a hard ceiling. Aducanumab's EMERGE/ENGAGE trials needed 18 months at high dose to produce a clinically questionable benefit, and did so at the cost of brain abnormalities in 40–41% of patients. Lecanemab and donanemab carry black-box warnings for brain bleeds that have killed trial patients, cost upward of $26,000 a year, and clear only one of two pathological proteins.

For the most common neurodegenerative disease on earth, this is the state of the art.

02The Reframe

The field is using AI to invent. We use it to perfect.

The default setting for AI in medicine is “Here is a disease. Generate a molecule.” Point a model at a blank page and hope something synthesizable, safe, and effective falls out. Billions have been spent this way. Alzheimer's has buried most of it.

We are not using AI to discover a drug. We are using AI to understand why a process that already works in human biology doesn't work well enough, and to engineer the molecular changes that close the gap.

Every improvement we make is an improvement to something already proven in humans. Not generative. Not hypothetical. That reorientation is the whole difference between a moonshot with a path and a moonshot without one.