Hit on the Fifth Try: an AI Design Tool at $10K/Month in Six Weeks with Zero Ad Spend

Mattia Pomelli and two friends built four design-related tools over the past year. None reached a decisive result. Their fifth product, Sleek, took three weeks to build, cost zero in ad spend, and hit $10,000/month (about ¥1.5 million) within six weeks of launch.

Their technical skill didn’t suddenly improve, and no outside capital was injected. All figures are disclosed in dollars; the parenthetical yen figures are reference conversions at ¥150 to the dollar.

The six-week breakdown

The short three-week build time owed largely to being able to reuse code written for the previous four products. The four failures left no revenue behind, but they did leave code as an asset. The fifth product’s early velocity rides on top of that accumulation.

What’s being sold

Sleek is a tool that has AI design mobile app interfaces. Users describe their app concept via chat, and screen designs are generated from that. Pomelli calls this “vibe design.” The target customer is app developers and founders who lack design skill or can’t afford to hire a designer.

The gap the team identified: coding and design assistance tools are plentiful for the web, but scarce for mobile apps. While no-code/AI website builders are becoming saturated, mobile had remained relatively untouched.

The stack is Next.js, Supabase, Vercel, PostHog, Stripe, and Resend. Billing is a monthly/annual tiered structure based on AI credits, with the free plan restricted to a single generation, editing or additional generations require an upgrade. The team has also tested region-specific pricing for lower-purchasing-power markets.

The one thing changed for product #5

There’s no “one day it went viral and everything changed” moment in this case. The launch tweet on X did perform well, but that was a result, not a cause. What worked was narrowing the definition of the customer.

Pomelli sums up the problem with prior products as being “useful to many people but perfect for no one.” When the target is broad, there’s no clear answer to which feature to build first, what copy will land, or where to even post. With Sleek, the target was fixed to “mobile app developers without design skill.” Once that was fixed, priorities and messaging and distribution followed automatically, in his account.

The difference between the first four products and the fifth comes down to whether, before building, you can state in one sentence who it’s for, not to feature count or development time.

What actually worked

The launch’s starting point was a single X post, built with a strong hook, a demo video, and structured to invite comments, on the premise that more comments get picked up by the algorithm. The roughly 8,000 existing followers were the fuel for that spark.

But what worked persistently was less the single post than what came after. Pomelli kept posting mobile-app-design content on X, daily. He’d design someone’s app for free in the comments, and the person watching would ask “what tool are you using?”, only then would the product enter the picture. In his words: “make content that’s genuinely valuable or interesting to people, and let it lead indirectly to your product.” He kept the same posture on Reddit. Instagram creators voluntarily made videos about Sleek (unpaid, in both cases).

There’s a logic here that fits neatly with the zero-ad-spend condition. A design tool’s own output is itself an advertisement. Showing a generated screen is stronger proof than any feature description. Because the product and the demo are indistinguishable, delivering value and acquiring customers happen in the same motion.

The billing design is equally lean. Restricting the free tier to one generation keeps AI inference cost down while letting only genuinely committed users move to paid. Per Pomelli, users who actually experienced the value didn’t hesitate to pay $25/month to evaluate the product further. Widening the free tier increases inflow but also cost, and it retains a segment who “try it, feel satisfied, and stop.” The one-generation design cuts both.

A premise not to overlook

First, roughly 8,000 X followers can’t be built in six weeks. The “zero ad spend” claim in this case worked because the distribution network already existed. Run the same playbook from zero followers, and the launch post reaches no one. It looks like free acquisition, but it was actually pre-paid, in time, well in advance.

Next, the three-week development period rested on the code assets from the four prior products. Add up the development time of those four failures, and the time actually invested in reaching Sleek exceeds a year.

And Pomelli himself names “market and timing” as one factor. Riding the wave of AI-built apps happening to surge at the moment they launched is precisely what surfaced the adjacent demand for mobile design. Riding a trend isn’t something reproducible through willpower.

Worth flagging as risk too: what’s disclosed is a single point-in-time figure of $10,000/month, with no churn rate or retention period revealed. AI-credit-based products see rising cost per generation as usage grows, so revenue growth doesn’t automatically translate to profit growth. $10,000 across a 3-person team is $3,333/person/month (about ¥500,000), and the level after subtracting living costs and inference costs is unknown.

The scope of replication

What’s transplantable is the approach to narrowing the target, and the idea of fusing content with acquisition. Question the “useful to many, perfect for none” state and cut down to a customer profile you can state in one sentence. Then, rather than promoting, put the output itself on display. Neither of these requires capital.

What’s hard to transplant is the 8,000 existing followers, the code assets from four prior products, and the timing. The follower count in particular is the variable that explains nearly the entirety of this case’s initial velocity. If the same product had been launched by a developer with zero followers, the $10,000 in six weeks doesn’t happen.

Put differently: Sleek’s six weeks aren’t six weeks. The preceding year and the 8,000 followers were simply cashed in, by coincidence, on attempt number five.

Related reading

• Pieter Levels, running Photo AI solo — a precedent for converting X-based reach directly into revenue
• Carrd’s growth to a roughly ¥200M-a-year scale — an example of a tightly scoped tool growing over the long run