The capital plan should explain the business in dollars.
Cleopas is building a new way to explore life's biggest questions about faith and existence. It combines deeply governed research with AI so a person can privately explore a much wider set of evidence, arguments and perspectives than any single human could reasonably hold in their head, through a conversation that moves at their pace.
The A$15 million ask is deliberately easy to scrutinise. If the cheque does not make the research better, the product safer, the team faster or the user learning deeper, it should not be in the budget.
The working assumption is a single capitalisation event. The goal is therefore not to minimise the cheque. The goal is to fund the mission well enough that Cleopas does not have to slow down, narrow the research programme or depend on another fundraising round while the product is still proving itself.
A$9.05M to build and learn fast. A$5.95M to protect the mission.
Indicative first 24 months
Engineering, research, scholars, paid users, compute, governance, distribution and operating contingency.
Protected strategic reserve
Years three to five, sovereignty, specialist escalation, additional cohorts and downside protection.
Front-load the expensive learning.
The first two years intentionally spend more than the old operating plan because speed and evidence quality are strategic advantages. Research and engineering run in parallel while real users enter the product early.
| Workstream | 24 months | What it should create |
|---|---|---|
| Engineering + product | A$3.05M | 4 to 6 engineers early, governed product, evidence graph, retrieval, evaluations, research tooling, security foundations |
| Internal research operations | A$1.55M | Research team, claim decomposition, source mapping, tagging, provenance, quality control and external-review preparation |
| External scholars + specialist review | A$0.80M | 150 to 250 independent review assignments across the first 50 major questions, plus premium specialist reserve |
| Paid alpha + beta users | A$1.50M | Staged cohorts building toward about 1,000 intensive participants and roughly 10,000 hours of real product use |
| Compute + security | A$0.65M | Inference, embeddings, model bake-offs, evaluation runs, observability, secure environments and sovereign-hosting experiments |
| Legal + governance + IP | A$0.35M | Licensing, participant consent, scholar agreements, IP assignment, privacy, entity and governance work |
| Distribution + partnerships + brand | A$0.50M | Controlled partner cohorts, launch preparation and measured channel testing |
| Operations + contingency | A$0.65M | Operating overhead and room for a research-heavy international AI project to encounter the unexpected |
| Total | A$9.05M | Front-loaded build and learning programme |
More engineers early because time to market matters.
Cleopas should not spend a year producing research while a tiny engineering team slowly builds around it. The plan assumes 4 to 6 engineers early, with product and design support, so the Evidence Landscape, conversational system, evaluations and research tooling can develop at the same time.
A narrow governed product should be in real hands quickly.
Users start changing the product and research roadmap well before year one is over.
Evidence system, product, evals, data tooling, security and infrastructure can move in parallel.
The research team is one of the company's two production engines.
Internal researchers do the repeatable work: assemble sources, split questions into claims, map evidence families, tag provenance, identify competing explanations and prepare material for external attack. Expensive scholars should spend their time challenging difficult judgments, not doing work the internal team can do.
Fifty major questions. Hundreds of independent review assignments.
The first 50 major questions should be unusually expensive and unusually well challenged. A simple planning assumption is US$1,000 per review, with 3 to 5 independent reviews per question. That creates 150 to 250 review assignments before premium specialists are added.
Highly governed early topic set.
No single scholar becomes the authority.
Independent attack before product confidence.
Different disciplines, countries and worldviews.
The A$0.80M budget is deliberately larger than the simple per-review calculation. Some globally recognised specialists may cost materially more. Some questions will require several disciplines. Others will need methodological or scientific review rather than historical review. The budget buys the ability to use the right person rather than the cheapest person.
Confidentiality and anti-contamination can be the same design choice.
One of the strongest features of the research model is that external work can be siloed. A scholar can receive a claim, source set and review task without being told the wider product thesis, the desired conclusion or what another reviewer has already said.
Scholar research
Reviewers work independently. They can disagree. Their disagreement is preserved. They do not need to know the intended product or be exposed to the conclusions of the other reviewers.
User research
Participants receive proper consent information about the product test, compensation, recording and data use, but they do not need to be primed with the product's preferred hypotheses or the conclusions of earlier cohorts before giving their own reactions.
This is purpose-blind independent research and product validation, not a formal clinical trial. Any withholding of hypothesis or product context should be designed with appropriate consent, privacy and research-governance advice.
Pay 1,000 people enough to stay long enough to find the flaws.
The first 1,000 beta participants are not vanity registrations. The working protocol asks for roughly 8 to 10 hours of use over one to two weeks, across several question families, followed by structured feedback and interview work.
We want to know what made them trust Cleopas, what made them distrust it, where it felt biased, where it became too technical, where it overclaimed, what changed their thinking and what made them want to leave. That feedback then becomes product changes, research priorities, evaluation cases and conversation-training material.
Cohorts should be staged. A practical sequence could be 50, 150, 300 and 500. Each cohort should receive a materially better version than the cohort before it.
The money creates a compounding system, not a static library.
This loop is the real capital efficiency story. Research creates better software. Software creates better user conversations. User conversations reveal which research matters. Over time AI should make the research operation faster while human governance remains responsible for what becomes trusted knowledge.
Raise it once. Spend it only when the evidence says to.
The remaining A$5.95 million is deliberately not tied to permanent first-24-month headcount. It protects years three to five and creates the option to move faster when evidence justifies it.
Possible uses
Slower supporter revenue, premium specialists, additional beta cohorts, sovereignty infrastructure, security events, high-demand research branches or proven distribution channels.
Governance
Milestone release, board oversight, budget variance reporting and explicit gates before reserve capital becomes operating spend.
The cheque should be inspectable in the product and the IP.
Each decomposed into the many smaller propositions underneath it.
With premium experts where the question deserves them.
Approximately 10,000 hours of real use plus qualitative research.
Research, governance, evaluation and AI tooling that can expand beyond the first topic set.
If the consumer product fails, the cheque should still leave something unusually valuable behind.
The capital does not primarily disappear into ephemeral model inference. It creates structured research, atomic claim maps, provenance, source relationships, scholar disagreement records, review protocols, evaluation sets, behavioural training material and a repeatable governance method.
That does not mean the research has a guaranteed A$15 million resale value. It does mean a mission-led funder can be left with a substantial reusable intellectual asset rather than a failed interface and a pile of model invoices.
For a church network, seminary, educational organisation, publisher or other Christian institution, a deeply structured body of independently challenged research could be reused across education, training, digital products, apologetics critique, curriculum, pastoral resources and future AI systems. Much of the research methodology is also portable to other fields where evidence is contested.
A$15 million is being raised to make the evidence and the learning unusually hard to attack.
The capital thesis
A$15 million is not being raised to build a chatbot. It is being raised to buy the research depth, independent challenge, engineering speed and real-user learning required to make a new kind of evidence-governed conversation trustworthy enough for life's biggest questions.