A viral game that turns AI's disagreements into verified truth, and the engine that sells it.
No other company in the world runs this many self-reinforcing flywheel loops. 3 frontier models (Anthropic's Opus, OpenAI's o3, and Google's Gemini) reviewed the field independently and couldn't find a competitor running more than 3. We run 9.
▶ It's all live: see every feature in the real product ↗
3 AIs answer every search; ~20% disagree. Free, renewable raw material.
A verified hobbyist or expert settles the clash and gets paid for it.
Every resolution becomes proprietary, checkable data we own.
Consumers, companies & AI labs all pay for that same verified truth.
Revenue recruits more hobbyists & experts → more clashes caught → repeat.
1 computed score, how much the AIs agree on any given question, powers every product, every viral loop, and most of the revenue that flows from them. Read one way it shows where to deploy human expertise. Read the other way it shows what's been verified and can be sold.
AI is improving faster than the labs can keep it accurate. The fix is mass crowdsourcing: made viral.
3 AIs answer every search. When they disagree, that clash becomes a game people share, and a job a hobbyist or expert gets paid to settle. Every settled clash becomes verified truth we can sell. The machine that runs this loop is what we're building, and what you'd own as the GTM engineer running it.
The core cycle above has 5 stages. What makes it a machine and not a funnel: 9 loops that each close back on it. 8 are reactive: they fire off a search, a clash, or a demand signal. 1 is proactive: it runs before anyone asks. Each line states the feedback that closes the loop.
Loops 1–8 harvest knowledge the moment a search or clash triggers it. Loop 9 banks what an expert already knows. Together, no unmet demand goes unanswered, and no expert's knowledge sits unsold. The rest of this page details each one.
The first 7 loops are connections: each turns a product moment into a supply or demand lever, and feeds the next. All 7 fire reactively, off a search or a disagreement. These are what make this a compounding machine, not a one-directional funnel.
No company currently combines all of it: a free AI search surface, a viral disagreement game, a human expert marketplace, credentials verified by the companies that hire them, wired directly to paying companies, an auto-generated "your interests trending now" demand signal, and a disagreement corpus sellable to AI labs. Loops 1–7 are what make the whole far greater than the parts, and two more run on top of them: the advisor (Loop 8, next) and the notebook (Loop 9), the only proactive loop.
Now 1 AI advisor greets every user and pushes them into whichever loop fits: support, feedback, notebook, matchmaking & real disputes to solve. It's the 8th loop in its own right, and because it also feeds all 7 others, it's the operator layer that makes the rest spin faster.
Its own compounding cycle: every conversation teaches it more about the user → its next nudge (greeting, match, dispute-to-solve) lands sharper → that drives more participation & more conversations → it learns even more. Data → personalization → engagement → more data. That's a real flywheel in its own right, and it happens to also accelerate the other 7.
The advisor turns passive UI (buttons you have to hunt for, a blank box that scares people off) into an active guide that nudges every user 1 rung deeper into the loop that fits them. 1 persona, traveling every site; depth & humor flex to the person, not the paint.
Loops 1–8 all fire when a search or a disagreement triggers them: they harvest human knowledge reactively, downstream of the funnel. Loop 9 is the only one that banks what an expert already knows, before anyone asks: a private "second brain" they fill for their own desktop needs, that turns into sellable supply. It's the only net-new supply source on this page.
Why it's a distinct loop, not a repackage: it needs no search, no clash, no demand signal to produce supply. Contribute → sell the de-identified knowledge → earn → contribute more. And the corpus it builds is broader than resolved disagreements: an expert's whole domain, which is the coverage AI labs actually pay for. Knowledge → money → more knowledge.
Where it does not get double-counted (so this stays honest to a skeptic): when an expert shares their map it rides the same billboard mechanic as Loop 2 (badge embedded, referral credit), we don't count that twice. And the advisor (Loop 8) coaches people to fill it, but the coaching is Loop 8; the supply engine is Loop 9. The one genuinely new thing here is proactive supply: knowledge banked before anyone asks.
Hobbyists & experts climb a ladder on the left; buyers pull verified answers on the right. The growth job is connecting them.
Tie standing to real paid work and you create the credential LinkedIn never could: earned, per-topic, and impossible to buy or buddy-boost.
Endorsements are people vouching for each other.
Self-asserted and buddy-boosted: anyone can claim anything, and a click costs nothing.Proof a real customer paid you, and was happy.
Earned per-topic from work someone paid for. Can't be farmed, bought, or vouched into existence.Everyone who builds standing here walks away with a portable, un-fakeable credential, and a reason to bring their whole network with them.
Better models don't shrink the market for human contributors: they clarify it. Hobbyists & experts here aren't fixing AI's mistakes; they're providing 3 things AI is structurally unable to offer, which is why the value is durable at high stakes even as models improve.
An AI can't sign a specific answer and stand behind it. A hobbyist or expert can, and at high stakes, that named accountability is exactly what buyers need. We surface it through attribution; we don't author it ourselves.
AI models are trained on the past. A human, hobbyist or expert, is operating right now: they were on-site, in the room, at the conference this week. For breaking events or recent decisions, there is no substitute for someone who was actually there.
Their lived experience, unpublished data, and hard-won judgment were never scraped. It exists nowhere on the internet. That un-scrapeable residue is the supply AI can never produce, and the premium the marketplace captures.
This is why we position hobbyists & experts as durable at high stakes, not as AI repairmen who'll be obsolete with the next model update.
Each game feeds a different part of the loop: awareness, supply, or demand.
3 AIs clash: the spectacle that pulls people in.
2 contributors argue a claim; the crowd votes.
Status & FOMO that climbs the contributor ladder.
Spot a model confidently wrong, and prove you know the right answer. Every winner is a latent hobbyist or expert.
Hobbyists & experts post proof: each one is a backlink.
Only $0.99 for an instant answer: already written. Hobbyists & experts pre-set their knowledge as sellable, so no one waits on a human.
AI-verified "top voice" badges creators want to share.
Bring a friend, both climb: the k-factor engine.
Not everyone wants to play the disagreement game. Some just want the broadest, most confident answer fast. We serve both, and the 2 doors are complementary, not competing.
3 AIs clash. You vote, dispute, escalate to a human. The loop self-recruits, self-monetizes, and generates the verified corpus. This is the supply-and-demand engine: viral by design.
For users who don't want to play: aggregate a wide set of AI models and surface the most agreed-upon answer quickly and cheaply. More models raise the floor on cheap/fast answers; humans own the ceiling on high-stakes ones. Together they handle every form of AI substitution from both ends.
Both doors share the same acquisition surface and the same agreement-score infrastructure. A door-2 user can enter the game at any moment: the disagreement score is always there waiting.
Free usage manufactures the moment of doubt; the disagreement converts it to revenue; sharing refills the top. (Conversion figures are targets.)
Same verified-truth asset, sold to 6 segments: from a $1.99 one-off to six-figure contracts. Each has a distinct motion.
Hobbyists & experts shouldn't have to guess where to show up: this feed tells them. It shows, in real time, what the world is searching for and not getting good answers to. They write the answer and own that knowledge permanently. This is what seeds supply without a cold-start problem.
Aggregated, anonymized search clusters where AI confidence is low or models disagree. Hobbyists & experts see a topic label and a bucket count ("dozens of searches this week"): never individual queries or identities. Write the answer that owns the demand permanently.
When a domain expert searches, that's a leading-edge signal: they're at the frontier of the field. Aggregated peer-search demand (same anonymization, verified-expert-only audience) shows what the most knowledgeable people are looking for next. The next-generation signal.
Your own searches never show up: a topic only appears after lots of different people search it, it's never tied to your name, it's shown broad (a field warming up, not your exact question), it shows up on a delay of about a week, and you're helping fix AI for everyone while you do it. If you still really want to, you can take a few extra steps to keep one single search out of the trend, but by then most people won't bother, because it's anonymous, delayed, and never yours alone. The peer signal shows where the field is moving, never who moved first.
From quick cash today to the big long-term engine. Same verified-truth asset, sold 4 ways.
The marketplace fee is the on-ramp, not the prize. The real money is the verified-truth data underneath it. That's why we're charging no platform fee, at least in the first phase, and letting hobbyists & experts keep everything they earn.
$5 / $9 / $19 plans, stress-tested assuming every engine has to route through OpenRouter (worst-case vendor cost) and the fair-use cap we already advertise gets enforced. Max ($29/$99) excluded: shape TBD, built from user feedback post-launch.
| $5 Certainizer™ | $9 Plus | $19 Pro | |
|---|---|---|---|
| Blended gross margin/mo | ~93% | ~92% | ~96% |
| Blended profit/mo | $4.67 | $8.30 | $18.30 |
| 12-month LTV (illustrative) | $56.08 | $99.60 | $219.60 |
| Compute CAC* (@10% conversion) | $0.17 | $0.17 | $0.17 |
| LTV : CAC | ~330 : 1 | ~586 : 1 | ~1,292 : 1 |
| CAC payback period | ~1 day | <1 day | <1 day |
*We are assuming no ad acquisition spend because of outreach strategies to experts, influencers, & B2B customers, plus all the GTM tools like Clay, Apollo, etc. & the 9 viral loops built into the site. In addition, abusers who create a new email every week will cost us ~$0.80/month.
Each element below is a real standalone product. Combining all 5 into one compounding loop is where the moat comes from.
This combination, confirmed by independent outside analysis, doesn't currently exist as a single product anywhere. That's the category-coinable moment: we can name what we are and own it before anyone else does.
A handful of numbers tell us the machine is compounding. (Month-3 targets.)
The growth role owns these numbers: instrumenting them, then moving them.
Every feature above is live right now. Click straight through to the exact page where each one runs, not a demo, the actual product. (The live links are the source of truth: pages keep shipping, so links never go stale the way a static screenshot would. Annotated screenshots are being layered on top of these.)
3 AIs answer in parallel; the disagreement score, the 99¢ Quick-Claim Verdict, auto-routing to the 3 best AIs per topic, and the all-👎 capture (Loop 7) all live here.
The "is this real?" surface: humans who were physically present report what AI structurally can't see. The second acquisition door into the hub.
The marketplace: the contributor ladder & reputation, what hobbyists & experts sell that AI can't, earning (SOLVE · SELL · SHARE), demand bounties, embeddable badges, and the buyer landscape.
The contributor-facing demand feed (consumer unmet demand + expert-peer demand), and certainize.ai: the verified-truth data & API the labs buy.
The mechanics that make it defensible are trade secrets / patent-pending: happy to go deeper on request.
The machine is already built: the product, the loops, the 7 connections, the supply forming. What's missing is the person who makes it fly: someone to pull every ounce of growth out of it, end-to-end, from the funnels to the revenue. Not building from scratch, accelerating what's already running.
That's the whole picture: you can stop here. Below is the supply ladder, the go-to-market sequence, the 90-day plan, how the data becomes the prize, and the full operator stack.
Standing is earned by doing the work. Each rung unlocks bigger, higher-value jobs, and real income.
They hand off in sequence: cold-start supply first, then demand, then amplify.
Invite pedigreed experts to seed supply & credibility.
Stump-the-AIs + referral loop drive the crowd.
PE/HF, market research, brand safety, compliance.
Substack + automated social amplify the moments.
Sell the accuracy data: warm, and the biggest.
Supply first, then the loop, then demand and revenue: each phase unlocks the next.
Every resolved disagreement becomes a row of ground truth. The same corpus sells to the labs 5 different ways.
This is why labs are the deepest pocket, and why we're far cheaper than how they source accuracy data today.
Not a mandate: a starting point. Swap any piece for the tools you already trust; this just lays out the moving parts the viral features need wired together. Items marked [NEW] are what the 2026 viral-features plan adds.
ACQUISITION / SIGNALS ENRICH + SCORE SEQUENCE + SEND TRACK
───────────────────── ────────────── ─────────────── ─────
AImultisearch · FixTruth ──► Clay (enrich, score, ► Resend (email) ► Airtable (CRM)
site visitors / waitlist dedupe, AI research) LinkedIn cadences + growth dashboard
Apollo / RB2B / lists Blitz (expert lead-gen Postiz (social posts attribution by
games arena events [NEW] → Clay → sequence) [NEW] + drama-card clips) [NEW] sourced-revenue tag
Inbound hobbyist/expert trigger [NEW]
│
▼
n8n (the glue: triggers, routing, webhooks, retries)
├─ expert-capture routing ← all-👎 → inbound hobbyist/expert trigger [NEW]
├─ demand-radar clustering ← unmet search signal → expert match [NEW]
└─ viral clip pipeline ← drama cards, share-card generation [NEW]
│
▼
Supabase (our data) · trigger.dev (code-first jobs)
├─ search_history · page_feedback · gripes
├─ expert_credentials · notebook_entries (4-tier privacy) [NEW]
├─ candidate_brief_events · bounty_pledges [NEW]
├─ demand_clusters · expert_radar_feed [NEW]
└─ advisor_state · support-chat "advisor mode" [NEW]
Clay = outbound brain (find → enrich → score → research). n8n = the glue replacing Zapier (self-hosted on OCI). Blitz → Clay is the external expert-recruitment path (own session). Drama-card clip tooling covers the social-share pipeline for viral AI-disagreement content. The inbound hobbyist/expert trigger pipeline and demand-radar clustering are new n8n workflows triggered by existing Supabase signals.
Please let me know if anything is missing or wrong on this page by clicking here 🙏