The money makes the product stronger.
The first 50 questions are expensive because we want them attacked properly. The first 1,000 beta participants are expensive because we want hours of use and detailed feedback, not vanity registrations. More engineers and researchers early buy speed. The remaining reserve buys independence if revenue, research or infrastructure takes longer than expected.
Method + research proof
Burial pilot externally stress-tested. Claim decomposition, source dependence, evidence-family separation and maximum defensible language now form the current architecture.
People
Founder-led work, current engineer, external reviewers
Capital
Existing capital / founder-funded
Gate
Ready to encode the reviewed method into a real product.
Build + private alpha
4 to 6 engineers begin the governed product while the internal research team starts the first 10 Gold Questions and prepares the wider 50-question programme. Private alpha begins in weeks 6 to 12.
People
Engineering 4-6; research ops 4-6; external specialists; 25-100 alpha users
Capital
Front-loaded engineering, research setup, compute, legal and first scholar reviews
Gate
Can the live system answer unscripted questions without outrunning evidence?
Closed beta + research factory
Staged paid cohorts begin. Research and engineering operate in parallel. Questions 11 to 20 introduce harder counterarguments. The system starts assisting claim decomposition and research preparation.
People
Hundreds of users; research and engineering teams both active
Capital
Paid users, specialist review, tooling, evaluation and product iteration
Gate
Does deep paid use expose repeatable failure modes and real question demand?
Expanded beta + first 50 governed questions
Build toward approximately 1,000 intensive paid participants and 50 deeply governed user-facing questions. User conversation data begins changing research priorities.
People
Up to ~1,000 intensive testers over staged cohorts; broader scholar network
Capital
Major beta spend, 50-question peer review, continued engineering and research
Gate
Are research hours per governed topic falling while integrity holds?
Demand-led Evidence Landscape
Topic ordering becomes increasingly user-led. System-assisted research improves. External experts focus on contested frontier judgments. Revenue and supporter economics are tested without making survival depend on them.
People
Growing research network, engineering scale, selected partners and broader beta/public cohorts
Capital
Product scale, research expansion, sovereignty pathway, controlled distribution
Gate
Do users, researchers and the system form a repeatable learning flywheel?
North Star
A living human-governed Evidence Landscape where real user questions reveal what should be researched next, AI accelerates the work, and humans retain authority over governed knowledge.
People
Research organisation potentially as strategically important as engineering
Capital
Deploy reserve only against proven learning and growth
Gate
Can Cleopas expand breadth without integrity degrading?
Raise once does not mean spend once.
The recommended A$15M commitment should be governed through annual budgets, board controls and milestone gates. Unused capital remains runway. If the product learns faster than expected, the reserve accelerates sovereignty, research breadth and distribution. If it learns slower, the same reserve prevents a forced second raise at exactly the wrong time.