Loading page…
Sell the taste that AI coding tools do not have: a library of templates, components and skills you feed the agent so it stops producing the same grey card layout — $15K MRR and 21,700 users in a month.
21,700+ users in about one month (posted publicly on X by the founder, relayed by the video's creator)
"You shouldn't just vibe code, you should also vibe design." AI writes acceptable code and mediocre interfaces, so the scarce thing is not generation — it is the guiding template you hand the model so the output does not look like everything else built that week.
Fewer bars = easier, cheaper, or faster for an AI-assisted solo builder. Editorial judgments based on the case details.
Aura is described in this video as "a site full of templates for beautiful websites with assets, components, and skills, all tailored towards strong design." It was built by Meng To, a designer who the video credits as the person behind it. The traction, as reported: "He posted on X that the product hit $15,000 in monthly recurring revenue or MRR and gained over 21.7 thousand users in just a month." Two details make this more credible than most cases in this video. First, the number traces to a public post by the founder himself rather than an analyst's estimate — it is still unaudited, but its origin is identifiable. Second, the founder's own working practice changed alongside the product: "He also shared that he now uses Cursor for design and he's no longer using Figma like in his previous workflow." A designer abandoning Figma for a code editor is a strong signal about where the workflow is actually moving, and it is the insight the product monetises. The product exists because of a gap the video states plainly: "AI tends to generate basic UIs. So, instead of letting it work on its own, you need to give it guiding templates to diversify the look." Millions of people can now generate a working app. Almost none of them can make it look like it was made by someone with taste. Aura sells the missing input.
The reported start is unusually steep: more than 21,700 users in roughly one month, announced by Meng To himself on X.
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. The exact words are quoted below, with a timestamp that jumps to the second each figure is said.
“The whole project was built by Meng To, who was the person behind Aura. He posted on X that the product hit $15,000 in monthly recurring revenue or MRR and gained over 21.7 thousand users in just a month.”
9:16 · hear it in the videoData credibility: Creator-reported, with an identifiable origin. The revenue and user numbers — $15,000 MRR and over 21,700 users in about one month — come from a public post the founder made on X, relayed by the video's narrator. We have not independently verified that post, no dashboard is shown, and the figures are unaudited. Not disclosed anywhere: the price point, the free-to-paid conversion rate, churn, the split between the three product lines (templates, components, skills), or how users were acquired. The one-month window is a real limitation — MRR reported one month after launch says nothing about retention. What is directly quoted and internally consistent is the method: designing in Cursor rather than Figma, starting with Claude models for coding and switching to Gemini or GPT on failure, building in small incremental steps, keeping prompts under three sentences, and supplying minimal correct context rather than full documentation. The founder is named (Meng To); no country or city is given in the video.
The money is subscription revenue: $15,000 MRR against more than 21,700 users in about a month.
The engine is that the teaching and the product are the same thing.
There is no reported quiet stretch here -- the figures cover about one month -- and that is itself the thing to plan for, because a month of MRR says nothing about whether anybody renews, and no churn number exists.
Three things bundled: templates, components and skills. Templates are complete page and app designs. Components are reusable interface pieces. "Skills" is the interesting third category — in the current agent tooling, a skill is a packaged instruction set the model loads to perform a task a particular way, which in a design context means the model can be taught house style rather than being asked for it in prose every time. So the product is not a design tool. It is an input library for someone else's design tool. The customer already owns Cursor or Claude Code; what they lack is anything good to point it at. Aura is the pointing. The founder is explicit that this is complementary rather than proprietary — the video notes "he recommends components from existing libraries like 21.dev." That is unusual and worth noting: he is not claiming to be the only source of good components, he is claiming to have assembled and curated a coherent one. Our read: this is a curation business wearing a software coat, and that is a feature rather than a limitation. The moat is not code, it is that a working designer's judgement is the scarce input and the customer knows they do not have it. It also explains the pricing shape — MRR rather than one-time sales — because a library that keeps adding to itself justifies a subscription and a static template pack does not.
Subscription. The figure given is $15,000 in monthly recurring revenue, alongside more than 21,700 users, in about one month. The video does not state the price or the paid-conversion rate, so the honest arithmetic is a range rather than a fact: 21,700 users producing $15,000 MRR implies either a small paid subset at a meaningful price or a large paid subset at a low one. Both are plausible for a design library and the video does not let us choose between them. What the shape does establish is that the free tier is doing real work. Twenty-one thousand users in a month is a marketing asset, not a cost centre, provided the hosting of static assets stays cheap — which for a template library it does. Our read: recurring revenue is the correct model here and it is not obvious. Most people selling templates sell them once, on a marketplace, and then spend forever acquiring new buyers. Charging monthly for continuous additions turns the same work into an annuity, and it also aligns the product with the actual pace of the market: every few weeks a new model or framework convention lands, and the library that keeps up is worth paying for again.
The founder's prompting advice is unusually concrete and it doubles as the sales argument for the library. On model choice: "not relying on a single model while building the app. Instead, it is more effective to start with Claude models because they are more powerful for coding tasks. And if it fails to do the task, then switch to Gemini or GPT models when needed." On scope: "building the app step-by-step with incremental changes" and "keeping the prompt simple by breaking the app into smaller parts and iterating on them one at a time." On prompt length, which is the most counter-intuitive line in the video: "prompts should ideally stay under three sentences so the AI stays focused." On context: "You don't need to dump all the documentation on the AI either. Instead, you should give it the minimal but correct context so it delivers what you actually want." His summary: "keep the agent setup simple and focused." Our read: notice how these two claims interlock. Prompts should be under three sentences, and the output should still look designed. Those are only compatible if the taste lives somewhere other than the prompt — in the loaded template, component set and skill file. The advice is not marketing dressed as generosity; the advice is the reason the product has to exist.
Copy the wedge, which is the most valuable idea here: find the input that AI tools are worst at supplying and sell that input. Design is the obvious one and it is now partly taken. The same gap exists for copywriting voice, for information architecture, for data models in regulated domains, for accessibility compliance, for motion and interaction detail, for email and notification design, for onboarding flows. In each case the agent will produce something functional and generic, and someone with real craft can package the guiding material. Copy the subscription framing. Sell continuous curation, not a one-time pack, and let the market's own churn justify the renewal. Copy the openness. Recommending other libraries costs nothing and buys credibility, because the customer is buying judgement and judgement that pretends to be the only source is less believable. Change the assumption that you can do this without craft. This case works because a practising designer's taste is the product. Our read: if you cannot personally tell a good interface from a bad one, you cannot build this business — you will assemble a library that looks exactly like what the model already produces, which is worth nothing. Pick the domain where you actually have the eye. Change the platform bet. Tie the library to a format and a workflow, not to one editor, so a shift from Cursor to whatever comes next does not strand the catalogue.
Model improvement is the direct threat. Every generation of frontier model produces better default interfaces, and the value of a guiding template shrinks as the baseline rises. This is a business that must keep moving up-market in taste to stay ahead of the default. The tooling could absorb it. Cursor, Claude Code and their competitors have every incentive to ship curated design systems, skills and templates natively and free. An input library is exactly the kind of thing a platform bundles. Curation is a treadmill. Subscription revenue is only defensible while additions keep arriving, and additions require the founder's own hours. This does not obviously delegate — the whole value is one person's judgement. Distribution is undisclosed. The video says nothing about how 21,700 users arrived in a month. Given the founder is an established designer with an existing following, audience is the likely answer, and a copier without one has a materially harder problem. Evidence quality. The revenue and user figures come from a founder's X post relayed by a YouTube analyst — no dashboard, no price, no conversion rate, no churn. And the one-month timeframe means there is no evidence at all about whether this MRR survived month two.
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