Loading page…
One PDF split into age-band variants, sold through a quiz that routes each visitor to their own version — so the question stops being “do I need this” and becomes “which one do I need”
3 conversions at ~$8 CPA · $29 product price · the tier-1-country ad set produced 0 conversions
The competitor's landing page had a quiz, and each answer led to a different version of the product. That single mechanic changes the buyer's internal question from whether to buy to which one to buy.
Fewer bars = easier, cheaper, or faster for an AI-assisted solo builder. Editorial judgments based on the case details.
Be clear about scale before anything else: the numbers here are tiny. $26 of ad spend returned $59 in sales on a brand-new account. This entry earns its place because those numbers are real, documented on camera from a fresh Gumroad account with nothing hidden, and because the mechanic behind them is genuinely under-used. The creator opens on his Gumroad dashboard — three digital products built with AI, all solving one problem — and then shows the ad that gave him the idea. Someone is selling a weight-loss programme, and their landing page has a quiz. Each answer leads to a different version of the product. His observation: instead of thinking do I need this, people end up thinking which one do I need. His product-selection rule comes first and is a hard reversal of the usual advice: do not sell what you love. Sell what is already working — products people are already making money from — and build an upgraded version. His research method costs nothing. Make a fresh social account, search planners, guides or templates so the algorithm learns your interest, and let the feed fill with ads for digital products. Engage with the ones you find, save them, and the feed compounds. More engagement on an ad is a better sign that people are buying. In parallel, use TikTok Creative Center or the Facebook Ads Library, search the same keywords and filter by oldest ads — a long-running ad usually means a profitable one. Collect names and links in a Google Sheet. His three checks for picking one: the problem recurs so people keep buying the solution; there is room to make it better; and you have seen the same type of product multiple times while scrolling, which suggests the seller is scaling it and there is plenty of content to learn from. He picks a weight-loss programme and upgrades it into a 100-day transformation programme.
The first buyers were bought, cheaply, on a brand-new account.
The quick-reference card — acquisition channels, replication playbook, risks & traps, and our verdict — is members-only.
Join the Unicorn Club — $5/moCancel anytime, in one clickPaste into Claude Code / Codex and get a working version of this product end-to-end.
Two production-grade prompts per idea — a full build spec and an SEO growth plan — plus 10 tailored build specs a week. Members only.
Join the Unicorn Club — $5/moCancel anytime, in one clickEvery claim traces back to a listed source — open any of them and check the number yourself.
Data credibility: Founder-reported and demonstrated on camera, at very small scale. The creator (publishing as Mr. AI CASH) built the entire funnel during the video on a brand-new Gumroad account specifically so the results would be visible from zero, and showed the TikTok Ads Manager and Gumroad screens: two ad groups at $2/day, zero conversions from the five-tier-one-country set, three conversions at roughly $8 each from the broad high-income set, only two of which Gumroad recorded, and — with the losing ad set excluded — $26 spent against $59 made. He states the pixel discrepancy openly and attributes it to routing traffic through his own site before Gumroad. He also says at the start that his dashboard shows three digital products built with AI, without giving a figure for those. The absolute numbers are trivially small and every derived ratio rests on three conversions, so nothing here is a validated CPA. The video contains affiliate-shaped promotion of the tools used (notably Kimi) and links a blog post containing the prompts. One element we flag rather than report neutrally: the demonstrated scenario ad uses an AI-generated person delivering a first-person weight-loss testimonial that did not occur, which is a fabricated endorsement and not a practice this directory endorses.
The product was priced at $29 on a Gumroad account opened the same week, and the first sale landed inside the first day the campaign ran.
The mechanic that turns is the quiz, and it was lifted deliberately from a competitor's ad: a landing page where each answer routes to a different version of the product, so "instead of thinking, 'Do I need this?' people end up thinking, 'Which one do I need?'" Once the first variant exists, every…
This build is designed so the silence never really starts.
One digital product deliberately split into variants, plus a quiz that routes each buyer to theirs. Both halves are necessary; the variants without the quiz are just a bigger catalogue. His variant logic: take the product idea and look for the different versions it can have. Most transformation programmes divide by age range. You can also split by gender, by a specific sub-problem, or by a specific goal. He notes seeing examples with so many variations that a customer does not quit until they find exactly what they are looking for — and says that is the goal. He builds three age-band versions (18-24, 25-39, and older). Production is one prompt per variant. He uses Kimi's Agent Swarm, which splits the work across multiple agents each handling a different part, and continues running in the background if you open another chat. The prompt instructs it to output the product as HTML divided into A4 pages, so the design exports cleanly as PDF. Roughly 15 minutes per document. The output he shows is a printable programme with clean, simple colours and — the part he singles out — worksheets people can write in and use to track progress daily. His framing: multiple digital products inside one digital product. Variant two and three do not need the full prompt. He tells it to create a new version using the previously generated 18-24 PDF as the exact template, adapted for ages 25-39, and repeats. The brand is generated the same way: he asks for a social-media brand name for the programme, picks The 100 Club, then asks for a logo prompt and generates several variations.
$29 per digital download on Gumroad, sold through the quiz page, with paid traffic as the accelerator and organic short-form as the free path. The ad test is documented in full and the honesty of it is the reason to trust it. He built one campaign with two ad groups, each with three to five creatives and a $2 daily budget. Ad group one targeted five tier-one countries with relevant interests. Ad group two targeted a broader set of high-income countries — tier one plus European — also by interest, language set to English. After more than 24 hours: the tier-one ad set produced zero conversions. The broad ad set produced three conversions at around $8 each. Gumroad showed only two, which he attributes to pixel attribution loss from routing visitors through his own site before Gumroad, and expects to improve after the first few sales. Turning off the tier-one set, the arithmetic he shows: $26 spent, $59 made — more than a 50% profit margin. His scaling plan from there is one campaign with three to five broad ad sets, each targeting a single interest, but only after letting the current ad run its full three-day test window. Our read on the economics: at $29 with a roughly $8 CPA there is real headroom, but three conversions is not a validated CAC — it is a signal. The structural strength is that the quiz routes buyers into the variant that matches them, which should lift conversion relative to a single generic page, and that the variant catalogue costs almost nothing to produce because each new version is a re-skin of the last.
Two channels, run in parallel, and he sequences them deliberately: free content first for durability, paid ads to compress the timeline. The free side is a fixed daily format. One scenario video where people can see themselves in the situation the product solves, plus two quote-or-advice videos teaching something or delivering information that stays with the viewer. He posts the same content across every short-form platform, with the link in every bio behind a clean call to action, and uses an AI prompt as a content strategist that returns two advice ideas plus one full scenario scene and script per day. His scenario-video production line is the most transferable part: pick a voice in ElevenLabs, paste the script, generate, then trim the gaps between sentences in CapCut so it sounds human. Generate the scene image (his example: a fit woman standing next to her overweight husband), then lip-sync it with Kling's avatar tool by uploading the image and the audio. Assemble in CapCut. The resulting clip has the character say she used to be overweight too, before the 100-day plan — a first-person testimonial delivered by a generated person. That is worth flagging plainly rather than admiring: it is a fabricated personal testimonial for a weight-loss product, which is both an advertising-standards problem and the exact thing this directory's own rules prohibit. Use the format for education and demonstration, not for invented before-and-after claims. Quote videos are trivial by comparison: free stock footage, text overlay, one repeatable editing style, then swap the clip and the text for each new one. Three videos in under 30 minutes, scheduled across the day. The paid side is the TikTok pixel wired to the Gumroad product page, verified in test events, then the two-ad-group structure.
Under $10 of fixed cost plus whatever you spend on ads, which is the genuinely notable part. Production: Kimi's Agent Swarm for both the PDFs and the quiz site — he says he chose it over Claude and ChatGPT for this specific work and shows heavy usage without hitting limits. ElevenLabs for voice. Kling for the avatar lip-sync. CapCut for editing. Gumroad for hosting and checkout, which takes its cut per sale and requires no monthly fee. Deployment is the frugal trick. Kimi deploys the quiz site to a shareable link, but he judges that link unprofessional, so he downloads the project as a zip, extracts it, and uploads the index file to Vercel as a new project — live in seconds. Then the domain: his brand name's .com was taken, and instead of adding numbers or picking a random extension, he uses a .online domain, which he argues tells people the business is online and is one of the cheapest extensions available. He buys it at around $1 on Namecheap, connects it to the Vercel project, verifies the email link, and it is live. Margin per sale at $29 is the price minus Gumroad's fee minus your blended ad cost. At his ~$8 CPA that is healthy; at a $20 CPA it is gone. The hidden cost he does not name: attribution. Routing traffic through your own quiz page before Gumroad breaks pixel tracking, which he encountered directly — three conversions in TikTok, two visible in Gumroad. That gap makes optimisation guesswork at low volume.
Copy the quiz funnel. It is the mechanic the whole entry rests on and it is under-used outside the weight-loss and supplement categories: segment one product into variants, then let a short quiz route each visitor to theirs. Copy his fix list from the demo too, because those details are what make it convert — the first question should ask the segmenting attribute, answers should be card-style with images, selecting an answer should advance automatically without a next button, and prices should not appear during the quiz. Copy the ad-library research method: search the keywords, filter by oldest ads, and treat longevity as the profitability signal. Copy the variant-as-template production trick — variant two is a re-skin instruction, not a new prompt. Copy the two-ad-group test structure at $2/day, and specifically copy the willingness to kill the ad set that produced nothing after 24 hours rather than defending it. Copy the domain frugality: a $1 domain on a sensible extension beats a brand name with numbers bolted on. Change the fabricated testimonial, unambiguously. Generating a person who says she used to be overweight before this programme is an invented endorsement for a health product. Use AI video for demonstration, explanation and stock-style scenes; never for fake personal results. Change the attribution setup. Either sell on your own domain with a proper checkout so the pixel fires on your page, or pass tracking parameters through to Gumroad and reconcile manually. Optimising ads on three conversions with a broken pixel is not optimisation. And change the validation expectation. $59 of revenue proves the machine runs, not that the product works. Get to 30-50 sales before you believe any CPA number.
The scale is the first honest risk: three conversions and $59 in revenue is a signal, not a business. Every ratio derived from it — the ~$8 CPA, the 50%+ margin — has a sample size of three and could move dramatically in either direction. He says himself that nothing is ever 100% guaranteed. The category risk is significant and specific: weight loss is one of the most heavily policed advertising categories on every major platform. Health and body-image claims get ad accounts restricted, and TikTok and Meta both enforce rules about before-and-after framing and body-related targeting. Building an ad business on that category means building on a permanent compliance edge. The integrity risk is the one to take seriously. The scenario video features a generated person delivering a first-person weight-loss testimonial that never happened. That is a fabricated endorsement, exposed to advertising-standards enforcement in most jurisdictions, and it is the single reason we would not recommend copying the acquisition step as demonstrated. The attribution risk compounds the small sample: routing through your own page before Gumroad broke his pixel, showing three conversions in TikTok and two in Gumroad. Without clean data, scaling spend is gambling. The product-quality risk: three PDFs generated in roughly 15 minutes each, where variants two and three are re-skins of the first, sold as a 100-day transformation programme. Refund rates and reviews will decide whether that survives contact with buyers, and the video ends before any of that is known. And the platform risk: Gumroad's terms, TikTok's ad policies and the ad account itself are all outside your control, with a brand-new account carrying the least trust and the most scrutiny.
The four founder playbooks — first customers, first dollar, the flywheel, the silent stretch — are in the Unicorn Club, on this idea and on all 1,435 others. The story, the numbers, the sources and every later section stay open to everyone.
Join the Unicorn Club — $5/moCancel anytime, in one click