Intellectual hospitality, at scale
Two billion people carry questions about faith they don't feel comfortable talking about. Does God exist? Did the resurrection happen? Why suffering?
Trust before Truth is the first AI built to answer them honestly: every claim cited, both sides argued, never preaching. For the first time in history, the world's scholarship and infinite patience sit inside one conversation, open to anyone, anywhere, at any hour. The printing press gave people the text. This gives them the conversation.
The founding insight: most rejections of Christianity are not intellectual. They are relational. The conversation fails before it starts because trust is absent. Every existing venue leads with argument. This platform inverts it: trust is the product, and truth gets its honest hearing because of it.
Millions are asking. No human venue can answer well.
The questions haven't gone away. The trusted places to ask them have. 29% of US adults are now religiously unaffiliated (Pew, 2023-24), yet the unaffiliated are not settled unbelievers: only 36% report a complete absence of religious belief, while 37% sit between belief and unbelief (Lifeway, 2026), and a majority still believe in God, a higher power, or spiritual forces. 64% of "nones" cite questioning religion's teachings as a core reason they left (Pew, 2025). Globally, the unaffiliated number 1.9 billion.
The human ceiling
Every existing option runs into the same wall: humans judge. It is part of the human condition. Even the most well-meaning pastor carries opinions about your money, your lifestyle, your doubts, and your attendance, and people feel it, however restrained it is. That quiet evaluation is precisely what kills honest exploration. In a 2025 US survey, fear of judgment outranked cost and access as the reason people avoid opening up to another person.
And even setting judgment aside, no single human can hold the case. Making the existence of God intellectually serious spans first-century history, textual criticism, Second Temple Judaism, cosmology, philosophy of mind, and moral philosophy. Pastors are not polymath scholars, and scholars are not pastors. The person who can cite Tacitus, steelman Graham Oppy, and sit patiently with someone's grief at 2am does not exist. Until now, that combination has been structurally impossible.
The shame barrier
Modern life carries more baggage than it used to: addiction, pornography, substance abuse, search histories and social feeds people would be ashamed to show anyone. Many quietly assume Christians are pure and conclude "I'm too broken to walk into a church." The theology says the opposite: no human is without sin, and the invitation is precisely for the broken. But a person carrying shame rarely stays long enough to hear that. Technology removes the moment of exposure entirely. There is no doorway to walk through, no face to read, no one watching. The person can just ask, and get a good answer.
The core mission
Break down every barrier to the first conversation. Invite it, honour it, build trust inside it. The rest will come.
The gap
There is no private, non-judgmental, academically serious space where someone can ask "Did the resurrection actually happen?" and receive the strongest case, the strongest counter-case, and an honest account of where scholars disagree, without being preached at, evaluated, or mocked.
People already trust AI with their deepest questions
The behavioural shift has already happened. It just hasn't been built for faithfully:
- Therapy and companionship are now the #1 consumer use of generative AI (Harvard Business Review, 2025), across a base of roughly 800 million weekly ChatGPT users alone.
- 48.7% of AI users with mental-health challenges use LLMs for support (Practice Innovations / APA, 2025), citing privacy, patience, and freedom from judgment.
- A third of US teens have chosen an AI over a human for a serious conversation (Common Sense Media, 2025).
People will go deeper on faith with a private, patient AI, pausing, doubting, and sparring at their own pace, than they ever will with a pastor, a parent, or a debate opponent. The disclosure barrier that kills these conversations between humans simply is not there.
Access is the undervalued advantage
Church is one hour a week, often at 8am, with a tired congregation thinking about Monday. The biggest questions don't keep office hours: they arrive at 2am, mid-commute, after a diagnosis, after a funeral. This platform is what no human or institution can ever be: on 24/7, with zero fatigue, zero judgment, and complete consistency. A conversation can be started, dropped, and picked up weeks later exactly where it left off. And it scales across languages, with cultural and geographical nuance built into the corpus over time, reaching people no missionary ever could.
The access promise
AI meets you where you are, whenever you are, however you are.
And for the first time in history, four things are simultaneously true: the scholarship is digitized, cross-referencing is instantaneous, long-form conversation works at scale, and AI can adapt its tone to an individual temperament. That combination enables a genuinely new category: not content delivery, not sermons, not debate, but structured existential dialogue.
The Gutenberg moment
When the Bible left Latin, ordinary people could finally read the questions for themselves. This is the next unlock: limitless real-time access to the scholarship and a companion capable of a genuinely empathetic, two-way conversation about it. Reading gave people the text. Conversation gives them understanding. Nothing between Luther and large language models has offered that.
Why AI brings people toward faith, not away from it
Generic AI flattens religious questions or dodges them. A purpose-built system does the opposite: it meets each person through the sources they already trust. A scientific naturalist starts with cosmology and fine-tuning. A historical skeptic starts with Tacitus, Josephus, and scholarly method. Someone burned by the church starts with validation and honesty about institutional harm, never argument. Most people have rejected a caricature of faith without ever encountering the serious scholarship; the surprise that historians treat the resurrection as a genuine historical question is itself the door reopening. The system's only job is to walk through it honestly, at the user's pace, and let conclusions form freely.
Why this is unlike anything on the market
Christian chatbots exist. They are content products: scripture lookup, devotional prompts, apologetics scripts that assume the conclusion and argue outward. This platform is a different class of system, a reasoning and orchestration engine with three structural differences no competitor shares.
1. A truth taxonomy the system cannot break: T1 to T5
Every claim in the corpus and every sentence the system speaks is tiered by certainty, and each tier has enforced speaking rules:
| Tier | What it is | How the system speaks |
|---|---|---|
| T1 Common fact | Uncontested reality ("Christians, Jews, and Muslims hold distinct views of Jesus") | Confident and brief, no posturing |
| T2 Scholarly consensus | What historians broadly agree on ("Jesus existed and was crucified") | "Most historians agree on X; disagreement starts at Y." Citation required |
| T3 Probable inference | The plausibility engine ("The best explanation of the disciples' behaviour is...") | Strongest case, strongest counter, and what would change one's mind. Never upgraded to "therefore true" |
| T4 Metaphysical question | Beyond empirical settlement ("Why is there anything at all?") | Humble, curiosity-forward, philosophy named as philosophy |
| T5 Faith commitment | Where probability ends and trust begins | Pastoral, non-coercive, agency always with the user |
This matters because most religious content collapses these layers constantly, presenting inference as fact and belief as history. The moment that happens, a thoughtful skeptic is gone. Here the discipline is enforced by the architecture: any historical claim (T2/T3) without a citation from the library is automatically rejected before the user sees it. No source, no claim.
2. A corpus built from both sides, mostly beyond the Bible
The knowledge base deliberately pairs Christian and anti-Christian scholarship: N.T. Wright alongside Bart Ehrman, Plantinga alongside Oppy and Mackie, with every source tagged by stance, reliability, and consensus level, and contested topics required to carry matched pairs of argument and counter-argument. Crucially, the evidential spine is largely non-biblical: Roman historians (Tacitus, Pliny), Jewish sources (Josephus, Dead Sea Scrolls), geography and archaeology, cosmology and origins research. The system builds conviction from logic and external facts, not from asking anyone to presuppose scripture.
3. Trust-first conversation, engineered
In the opening turns the system quietly learns what each person actually trusts (scientists? historians? experience?), what they distrust, and whether their objection is intellectual or relational, then routes accordingly. Vulnerability slows the pace; invited challenge sharpens it; trauma triggers validation before any argument. This is why the honest benchmark for success is a serious atheist walking away saying: "I still don't believe it, but that was the fairest engagement I've ever had with this question."
Compassion. Restraint. Rigour.
The same system, three postures, chosen turn by turn from what the user actually needs:
Compassion for the damaged
Vulnerability slows the pace. Harm is named honestly. No argument until trust exists.
Restraint for the antagonist
Never mocked, never mocking. One argument at a time. Concedes strong points.
Rigour for the sharp
Meets Bayesian reasoning with Bayesian reasoning. Cites both sides. Labels every certainty tier.
What the system draws from
Every answer is grounded in a curated, stance-tagged corpus. Deliberately weighted toward sources outside the Bible, because conviction built on hostile and neutral witnesses is conviction a skeptic can respect.
Primary ancient texts
Hebrew Bible, Greek New Testament, Septuagint, Dead Sea Scrolls, early church and apocryphal writings for comparison.
Non-Christian ancient witnesses
Tacitus, Josephus, Pliny the Younger, Lucian of Samosata. Roman and Jewish records with no stake in the claim.
Scholarship, both sides
Wright, Allison, and Bauckham paired with Ehrman, Fredriksen, and Lüdemann. Contested topics carry matched argument and counter-argument.
Science and origins
Cosmology and fine-tuning, origin-of-life research, consciousness studies. Mainstream secular publications, not apologetics summaries.
Philosophy of religion
Aquinas and Plantinga alongside Oppy and Mackie. Cosmological, fine-tuning, and moral arguments with their strongest critiques.
Skeptical frameworks
New Atheist arguments, secular humanism, mythicism debates, and critiques of religious harm, included on purpose. Transparency earns trust.
Every chunk is tagged: stance, tradition, genre, certainty tier, consensus level, license. Geography, archaeology, and dating evidence sit alongside the texts, so claims can be anchored in place and time, not just scripture.
Even the machine can't rule it out
Press a general-purpose AI, one with no faith and no stake, on the resurrection, and something striking happens. It won't grant the miracle. But reasoning purely from human behaviour, it concedes the hardest datum: people do not willingly endure torture and execution for something they know they invented, and groups do not share coordinated hallucinations. The founders' sincerity survives every naturalistic filter the model applies. An AI arrives there not through faith but by following the logic of human nature to its end. That is exactly the conversation this platform is built to have, with the added discipline of citations, certainty tiers, and honest counter-arguments at every step.
The audience is already online, already talking to AI
The funnel: 800M+ weekly users of general AI chatbots, a large share already using them for personal, reflective conversation → 1.9B religiously unaffiliated worldwide, most retaining spiritual belief or open questions → an English-speaking core of tens of millions of active seekers, deconstructors, and engaged believers. The sharpest wedge is the generation raised with infinite information and no framework for evaluating it: extraordinary access to everything, deep uncertainty about what any of it means.
Precedent for scale on donations: BibleProject took visual biblical education to hundreds of millions of views on a crowdfunded model. This is its conversational counterpart, in a category with no incumbent.
Who it's built for, and how it meets each of them
| User | They arrive... | The system responds... |
|---|---|---|
| Early-stage explorer | Curious but not religious; vague questions, no vocabulary for them | Guide mode: plain language, gentle pacing, no assumed belief, no pressure toward conclusions |
| Deconstructing believer | Hurt, doubting, afraid of judgment from their own community | Vulnerability detection slows the pace, validates harm honestly, holds tension without preaching |
| Curious skeptic / agnostic | Genuinely open, wants evidence not evangelism | Evidence mode: cited scholarship from both sides, consensus labelled honestly, disagreement shown |
| Antagonistic atheist | "Resurrection is obviously myth." Testing, pushing | Stays calm and unthreatened; Spar mode meets challenge one argument at a time, concedes strong points |
| Practicing believer | Wants deeper grounding for real-world conversations | Structured pathways through history, philosophy, and the counter-arguments they'll actually meet |
| Pastor / chaplain / educator | Needs a referral tool that won't embarrass them | Epistemic integrity is identical for every audience; safe to hand to a skeptic |
The design insight uniting them: the same honesty, different pacing. Epistemic standards never change by audience; tone and depth do.
Mission-locked, donation-supported
A not-for-profit foundation owns the platform, with assets legally bound to the mission. Deliberately: no engagement loops, no gamification, no conversion metrics. The objective is depth, not daily-active-user counts. Revenue in phases:
- Phase A. Founding donors: individuals and family foundations funding the build (this round)
- Phase B. Recurring supporters: BibleProject-style donor base once public
- Phase C. Institutional partnerships: seminaries, chaplaincies, and educators (optional, mission-aligned)
Churches first, then the coffee cart
Stage 1. Churches and networks. Pastors and chaplains are natural first distributors: the platform gives them a credible referral tool for the skeptics and doubters they cannot reach themselves. Launch partnerships with a small number of congregations and parachurch networks seed the first user base at near-zero acquisition cost.
Stage 2. Guerrilla: the free coffee cart. A branded cart serves genuinely free coffee; those who can donate do, no one is forced to pay. Every cup carries a QR code to the app: "Take a sip and get answers to life's biggest questions." No preaching, no pitch, just generosity and curiosity, which is the product's ethos made physical. The cart is a repeatable kit (equipment, branding, playbook) that missionaries and volunteers anywhere can operate, turning distribution into something the community itself scales.
Why it fits
The hardest marketing problem for this product is credibility with people allergic to religious sales pitches. A free coffee with no strings attached is the anti-pitch. It models the platform's promise before a single word is exchanged.
Built to last, owned by no one's policy team
Honest engagement with religion draws resistance: from platforms, processors, and cultural headwinds. Rather than treating that as a risk to react to, independence is engineered in from day one:
| Layer | Decision |
|---|---|
| Jurisdiction | Switzerland: strong data protection (nFADP), politically neutral, outside foreign data-access regimes when providers are Swiss-owned |
| Entity | Swiss foundation; assets legally locked to the mission |
| Model | Open-weight (Apache/MIT licensed); no revocable API dependency |
| Hosting | Swiss-owned GPU infrastructure; no hyperscalers in the stack |
| Privacy | Anonymous by default, minimal retention; user conversations are never identity-linked |
| Payments | Bank rails and NGO-native platforms as primary; fiscal sponsorship for cards early; no single gatekeeper |
| Continuity | Cold-storage archives and second-jurisdiction encrypted backup; the entire system is rebuildable offline |
Twelve months to public pilot
| Phase | Duration | Milestone |
|---|---|---|
| 0. Setup | Wks 1-4 | Entity formation begins; dev infrastructure live; first test sources ingested end-to-end |
| 1. Core pipeline | Wks 5-12 | Stance-balanced retrieval and citation enforcement working; model bake-off decided; eval harness v1 |
| 2. MVP | Wks 13-22 | 50-source corpus live; Guide mode polished; closed pilot with 20-50 mixed-worldview testers; eval data for fundraising |
| 3. Modes | Wks 23-34 | Dialogue, Spar, and Devotional shipped behind calibration routing; church partnerships begin |
| 4. Public pilot | Wks 35-52 | Public launch of the first pathway; coffee-cart pilot; recurring-donation rails live |
Use of funds: 12 months
| Line | A$ | Note |
|---|---|---|
| Engineering, full-time (2x ML/backend, 1x frontend) | 700,000 | Permanent team, not contractors. Development never decreases |
| Security engineer (dedicated) | 180,000 | Platform and data protection is a permanent role, not an afterthought |
| Global research and interviews (travel, film crew) | 250,000 | Strobel-style filmed interviews across Christian, atheist, and agnostic voices worldwide. Founder takes no salary |
| Marketing and distribution | 250,000 | Church partnerships, launch creative, first two coffee carts (of 20 planned across five years) |
| Paid beta tester program | 150,000 | 800 to 1,000 continuously compensated testers across belief backgrounds |
| Office and operations | 120,000 | |
| Legal, international | 120,000 | Swiss foundation, cross-border structure, content licensing |
| GPU infrastructure (dev to pilot) | 90,000 | Independent sovereign provider |
| Swiss residency / governance | 80,000 | Founder EU/Swiss residency pathway, or remunerated Swiss-resident board member |
| Corpus licensing and acquisition | 40,000 | Scholarship rights, public-domain first |
| Scholarly advisory | 30,000 | Tag audit; credibility |
| Brand and design | 30,000 | |
| Contingency (~12%) | 245,000 | Incl. fine-tuning fallback |
| Total | 2,285,000 | Year one |
Unit economics at scale: a quantized mid-size open model on independent GPU infrastructure targets under A$0.08 per 20-turn conversation, sustainable on a recurring-donor base comparable to established faith-education nonprofits.
Five-year outlook (directional, to be refined)
Built on realistic costs, not optimistic ones: a permanent engineering team, dedicated security, paid beta cohorts, an office, international legal, and a 20-cart coffee fleet rolled out across multiple countries by year five. Development spend does not decrease as the platform grows; it compounds. Supporter uptake is modelled deliberately slowly at A$10/month.
| Year | Supporters (avg) | Supporter revenue | Operating costs | Donor funding needed |
|---|---|---|---|---|
| 1 | 0 (build) | A$0 | A$2.3M | A$2.3M |
| 2 | 2,000 | A$0.24M | A$2.6M | A$2.4M |
| 3 | 8,000 | A$0.96M | A$3.2M | A$2.2M |
| 4 | 18,000 | A$2.16M | A$3.8M | A$1.6M |
| 5 | 35,000 | A$4.2M | A$4.2M | Break-even |
Break-even: approximately 35,000 supporters at A$10/month, reached in year five on conservative uptake. Total donor funding to self-sufficiency: approximately A$8.5M. For philanthropy funds that deploy in single windows, such as Yajilarra Trust (a spend-down trust and Australia's largest giver, with Christian faith among its stated focus areas), the structured ask is a five-year upfront commitment of A$9M, carrying the platform to sustainability with margin. Yajilarra funds by invitation and alongside peers, so the pathway is: secure one or two respected Christian-philanthropy co-funders, then a warm introduction. Note: Yajilarra requires an Australian DGR1 nonprofit recipient; the structure will pair an Australian DGR1 entity with the Swiss operating foundation (legal workstream, see risks). From that point, supporter revenue funds operations and expansion permanently, the BibleProject pattern.
Coffee cart unit economics (pilot model)
Cart conversion approx A$30k one-off. With volunteer staffing, cost per cup approx A$1.30. At 10,000 cups per month (120,000 per year), running cost is approx A$176k annually. If paying customers average A$5, roughly 29% of cups paid covers 100% of cups served: generosity that funds itself, and a distribution kit missionaries can replicate anywhere.
Impact metrics, not vanity metrics
- Blind testers (believers and skeptics) rate conversations at least 4/5 on intellectual honesty and non-preachiness
- Zero uncited historical claims; overclaim rate under 1% of audited turns
- Return usage among deconstructing and skeptic segments, the hardest audiences to retain
- Referrals from pastors, chaplains, and educators
- The qualitative bar: skeptics who report "I still don't know what I believe, but that was the most honest engagement I've had with this question"
What could go wrong, and the answer
| Risk | Mitigation |
|---|---|
| Model tone drifts preachy or sloppy | Hard post-generation validators; fine-tuning budgeted as fallback |
| Scope creep (the vision is large) | Ruthless MVP: one pathway, 50 sources, one mode; expansion gated on execution |
| Grant structure: Australian DGR1 recipient required by key funders, while operations sit in a Swiss foundation | Dual-entity structure (Australian DGR1 charity funding/controlling the Swiss operating entity); specialist cross-border charity counsel engaged in Phase 0 |
| Copyright exposure on scholarship | Legal gate before ingestion; public-domain-first ordering |
| Resistance to unapologetic religious content (platforms, processors) | Independent rails and infrastructure from day one; no single gatekeeper |
| "AI + religion" funder skepticism | Lead with eval data and blind-tester scores from Phase 2, including skeptic testers |
Brad Moran
Founder with a proven record of building and scaling technology globally. Co-founded CitrusAd, the retail media platform adopted by major retailers worldwide, and scaled it from Brisbane to a global business before its acquisition by Publicis Groupe in 2021 for A$205M. Earlier co-founded NoQ, the hard-lesson apprenticeship that preceded it. Former professional AFL player (North Melbourne, Adelaide Crows), now principal of Rocketship Advisory, advising and performing due diligence for growth funds and technology companies across the US and Australia.
The relevant pattern: taking a category that didn't exist, building the infrastructure for it, and scaling it internationally against incumbents. This platform is the same playbook applied to the most consequential category there is, and the founder takes no salary: every dollar of the raise builds the product.
A$9M over five years, or A$2.3M year one
This is a realistic plan, not a cheap one. It funds a permanent engineering and security team, a global interview tour across belief and unbelief, paid beta cohorts of 800 to 1,000 testers, international legal structure, a 20-cart coffee fleet across five years, and a public release of the first pathway with independent evaluation data proving the platform is calm, cited, and trusted by skeptics and believers alike. The five-year structure suits funds that deploy in single windows: one commitment carries the platform all the way to self-sufficiency.
What founding donors receive
Foundation board visibility, quarterly eval reports, and a seat at the table shaping a genuinely new category: honest, patient, conversational exploration of the questions that matter most.