Raised $20K before writing a line of code. The YouTube-script AI that reached $62K/month in 18 months

Gil Hildebrand received $20,000 (about ¥3 million) before writing a single line of product code. This is the story of Subscribr, an AI tool for YouTube creators he launched in April 2024. $10,000/month by day 100 after launch; $62,000/month (about ¥9.3 million) after 18 months. First-year cumulative revenue exceeded $500,000, putting the business on a $1,000,000/year (about ¥150 million) pace.

Below, dollar figures are followed by a yen reference converted at ¥150 to the dollar — an approximation that doesn’t reflect actual exchange rates.

The 18-month arc

Cross-referencing the ~4,000 customers with $62,000/month gives $15.50 per customer per month, which doesn’t match the $49 minimum plan, so the natural reading is that the count includes buyers of lifetime licenses sold via AppSumo, and cumulative customers who have already churned, but Hildebrand doesn’t disclose this breakdown. This is a figure that remains unresolved.

What’s being sold, and why this market

Subscribr is an AI tool that helps YouTube creators plan video concepts, work out thumbnail-and-title packaging, and build out script structure. The stack is Laravel, with Livewire for real-time processing, Horizon for job queues, and a single DigitalOcean droplet. Rather than keep up with the churn of JavaScript frameworks, Hildebrand chose the PHP ecosystem for its stable dependencies and fast builds.

Market selection was deliberate. The YouTube Partner Program has over 3 million channels, yet there was little software built specifically for them. He calls this “a Goldilocks market”, large enough to sustain a $1,000,000/year business, but too small for well-funded competitors to seriously bother entering.

His background is relevant too. He joined a web-development firm at 15, became an independent $50/hour consultant at 18, spent nine years as CTO under Seth Godin at Squidoo running a three-person engineering team to $10,000,000 in revenue, then went through a VC-funded crypto startup before returning to bootstrapping.

The decisive move: getting paid before building

The turning point in this case came before the launch. Hildebrand built a 1,000-person email list through activity on X, a mini-app, an industry report, and a public mockup, then aimed a steeply structured pre-sale at that list.

The terms: sell only 50 lifetime licenses. The first 10 at the lowest price, the next 10 at a higher price, rising in tiers from there. Delivery deadline of two months, with a full refund if the deadline was missed or the buyer was unsatisfied. The 50 licenses sold out in days, bringing in over $20,000. Hildebrand delivered on schedule. Refund requests were limited to a small number of cases where circumstances changed.

The difference before and after isn’t the dollar amount. $20,000 isn’t much as startup capital. What changed is that “someone will pay if I build this” moved from hypothesis to fact. In his own words: “real validation is a paying customer and money in your account. The first thing to figure out is: what’s the minimum required to get someone to pay you.”

The tiered price increase mattered too. The cap of 50 licenses combined with rising prices gave the list a reason to “decide now.” A signal was made visible through price differentiation that a free interest survey never would have produced.

The structure of what worked

First, the order of time was reversed. Normal development goes: build → ship → validate → discard if it doesn’t sell, by the time validation finishes, months have already vanished. Hildebrand moved validation to the front, and by imposing a refund guarantee and a delivery deadline on himself, closed off his own escape routes while building. Every decision during development could be measured against one concrete standard: fulfilling the promise made to 50 paying buyers.

Second, the price-tier choice. A minimum of $49 isn’t cheap for an individual-focused tool. Hildebrand didn’t chase customer count. He raised the unit price and personally staffed support as the founder. Fewer customers paying more carries an operational meaning: it fits within the volume of inquiries a solo operator can handle.

Third, front-loaded investment in compounding channels. He invested in SEO content before turning profitable, and now gets healthy organic traffic, plus traffic arriving via ChatGPT. An affiliate program is also functioning. Meanwhile, Google and Meta paid ads run for brand-awareness purposes and roughly break even. Paid traffic disappears the moment you stop paying for it. SEO and an email list compound. This allocation is what paid off over the 18-month window.

What didn’t work

AI product lifetime deals are something Hildebrand names explicitly as a risk. Selling a buy-it-once license on a model where every use incurs inference cost means losses deepen the more it’s used. In fact, the AppSumo campaign came with terms where the platform took the majority of the take. He states you shouldn’t offer a lifetime deal without deeply understanding usage patterns first.

The limits of solo operation also showed up. As customers grew, support consumed more time and marketing intensity couldn’t be sustained. That’s the backdrop for bringing on a co-founder at month 18, someone with over 15,000 YouTube subscribers and existing recognition in the target market. Revenue rose 50% in the first month after joining. Customer education, an area solo operation couldn’t reach, got filled in.

One more point: over 18 months, the product’s own design aged. AI has moved fast, and the current architecture no longer meets today’s expectations, prompting a v3 rebuild as a fully agentic chat interface. That a full redesign was needed after just 18 months illustrates how short product lifespans run in this space.

Conditions for transplanting this

The pre-sale technique itself is something anyone can copy. But this case’s success rested on a premise: a list of 1,000 people to aim it at. That list was built through sustained activity on X and free publication of the mini-app and report. There’s months of runway embedded in that.

Similarly, a promise of a refund guarantee with a two-month delivery window is also a bet one can take on because of 25 years of implementation experience. Making the same promise without that technical confidence risks a pile of refunds.

Even so, the sequence itself demands neither capital nor a résumé. The question “can I get paid, even minimally, before building?” can be asked regardless of whether the product is SaaS. Hildebrand also says: “don’t quit your job to start a company, do it as a side hustle. Don’t raise money before you’ve proven how you’ll acquire customers.” At the center of this case is someone who had been through a VC-funded company choosing to secure revenue before capital.

Related reading

• Plausible Analytics reaching $1M ARR — a different route to roughly ¥100M in annual revenue with no outside capital
• PDF.ai’s Damon Chen — the cost structure of running an AI product with a small team