48 signups in one day, zero conversions. Here's what I learned building an AI ad tool.
3 weeks ago I shipped Silo, an AI tool that generates ad creatives for Meta/Facebook. Today I have 99 signups, 213 generations, and exactly $59 in MRR.
Let me tell you what actually happened.
The idea
I was running ads for DTC brands and got tired of the creative bottleneck. You need fresh creatives constantly. Meta's algorithm eats them alive. Hiring designers is $500+ per batch and takes days.
So I built a tool that generates complete ad images using AI. Not templates you fill in yourself. Actual AI-generated images based on proven ad formats that convert. Fake iMessage conversations, steamed mirror text, social proof screenshots, before/after comparisons. You drop in your brand URL, it extracts everything automatically, and generates ready-to-run creatives.
The "viral" day that taught me everything
A Polish AI educator discovered Silo and wrote a full tutorial for his Skool community. 48 signups flooded in on a single day. I was thrilled.
Zero of them converted to paid.
Why? They were tech-curious AI enthusiasts browsing cool tools. Not brand owners with ad budgets who actually NEED this. The lesson hit hard: the right 10 users matter more than the wrong 1,000.
What's actually working
My ONE paying customer ($59/mo) came from organic Reddit. A real brand owner who tested it and saw value immediately. The comment screenshot and steamed mirror templates get the most "whoa" reactions. Nobody cares about the tech. They care about "will this make me ads that convert?"
The numbers (fully transparent)
- 99 signups
- 1 paying customer at $59/mo
- $59 MRR
- 213 total generations
- Infra cost: ~$10/mo (Fly.io + Gemini API)
- Built the whole thing in 3 weeks as a solo dev
What I'd do differently
- Skip the free tier entirely at launch. Free users cost money and rarely convert.
- Target brand owners on Reddit and Twitter from day one, not directories.
- Build 3 killer templates before 280 okay ones.
Currently focused on getting in front of DTC brand owners and media buyers. Generated sample ads for The Oodie (Davie Fogarty's brand) yesterday and the CreativeOS president followed me on X. Small wins.
If you run Meta ads, I'd love your honest take: siloai.app (free tier, no card required)
What creative formats are working best for you right now?
@moniantov Totally agree on the time allocation gap. I tracked my own hours for a week and realized I was spending 70% of my time on product and 30% on distribution. Flipped it to 50/50 and that's when things started moving. The product was already good enough — I was just hiding behind "one more feature" instead of doing the uncomfortable work of putting it in front of people.
@JeffTech74 The irony of building a distribution tool but struggling with distribution yourself is painfully real. What are you building? I found that the best distribution channel for dev tools is literally just being helpful in communities where your target users hang out. No links, no pitching — just answer questions and let people find you through your profile. It's slow but it compounds.
It's a new social media platform. I could try and explain it here, but if you go to Pr3ss3D.com I'd appreciate your impression of what I am doing and how it's different.
I've found a few of these lessons out myself, but my product has some harder nuts to crack. My last one being distribution. But I wish my app was already launched for you. No algorithm blocking reach. It's so ironic...I built a solution for the very thing that is stopping it from becoming fully real.
The "right 10 users matter more than the wrong 1,000" line is the whole post distilled.
I've been thinking about a version of this problem but applied to time, not users. Most founders (myself included) have a massive gap between where they think their hours go and where they actually go. You can spend weeks "doing marketing" and when you actually measure it, 80% of those hours went to things that produced zero signal.
Your Polish AI educator moment is a perfect example. It probably felt productive - 48 signups in a day. But the allocation was pointed at the wrong audience. The one Reddit comment that landed your paying customer probably took 5 minutes.
The lesson isn't just "target better." It's "measure what your time actually produces, not what it feels like it produces."
Curious - now that you know your one paying customer came from organic Reddit, how has that changed how you allocate your marketing hours?