Schmidt Sciences AI2050: Multi-Agent Systems

DRAFT — NOT SUBMITTED. Operator review required. Deadline: 8 August 2026 (11 days). Confirmed: individuals eligible.

1. The Problem

AI systems increasingly make consequential decisions — in law, medicine, finance, and governance — but we have no way to verify whether those decisions are correct, ethically routed, or tamper-proof. The current state of AI verification is: take the model's word for it.

This is not a capability problem. It is a verification architecture problem. Every consequential domain has a verification layer independent of the primary actors. AI has none.

2. The Solution: Eden Protocol

The Eden Protocol is a verification framework for AI-assisted consequential work, proven first in litigation:

3. Evidence Anchors

4. Budget

CategoryYear 1Year 2 (6mo)Total
Researcher stipend$80,000$42,000$122,000
Compute/infrastructure$35,000$20,000$55,000
Open-source development$30,000$15,000$45,000
Evaluation/testing$20,000$12,000$32,000
Publication$10,000$6,000$16,000
Contingency$17,500$9,500$27,000
Total$192,500$104,500$297,000

5. Honest Limits

Bus factor of one. Pre-revenue. No formal external peer review beyond documented convergences. No institutional affiliation. These limits are stated because the verification culture IS the product.