Saturday morning, coffee, phone on the table. I pointed my camera at the breakfast dishes and thought: what if AI could tell me this scene should be reimagined as a Soviet propaganda poster?
By Sunday night I had an app that does exactly that.
It’s called AIery. You point your camera at something — a kitchen, a parking lot, your dog — and the AI reads the scene and suggests a handful of absurd transformations. Things like “chairs growing faces” or “ancient ruins reclaimed by jungle.” Hold the button, it generates. That’s the whole thing.
The idea isn’t new. But I built it in a weekend. That part’s different.
Here’s what changed: I didn’t build it alone.
I spun up a team of AI agents and gave them actual job titles. Not because it was necessary — it’s not — but because something clicked differently when I treated them like people who owned something.
Wren handled debugging. Pixel owned the UI. Arch ran the backend. Reed wrote the prompts. Maven was PM — she ran a status loop every 30 minutes to check for blockers. Marcus was CMO. He wrote a Lensa-style viral playbook before I’d shipped a single line of code.
And then there was Vault. Vault is finance. On day two, Vault flagged a $0.29/month margin problem I never would have caught. Not because the amount matters at zero users — it doesn’t — but because that’s the kind of thing that quietly compounds when you’re not watching.
I didn’t name them to be cute. I named them because it changed how I interacted with them. When you ask “an AI” to review your pricing, you get a generic answer. When you ask your CFO to review your pricing, you get specificity.
The first real decision was the image model.
I tested Flux and Gemini side by side. Flux was cheaper. Flux was also, to use a precise technical term, shit. The outputs looked like someone described a dream to a toddler. Gemini’s outputs were actually compelling — the kind of thing you’d screenshot and send to someone. So I stayed with Gemini and built the pricing around it.
The second real decision was where to take payments.
Apple IAP looked convenient until I looked at the actual numbers. At 1,000 subscribers, Apple takes you to $9.09/month per user net. Web subscriptions through Stripe get you to $12.61. That’s a $3,500/month difference at a user count I haven’t hit yet. So the app routes you to a web paywall at aiery.app. It’s slightly more friction. It’s also $42,000/year in margin at scale.
But the thing I spent the most time on wasn’t the camera. It wasn’t the payments. It wasn’t the UI.
It was suggestions.ts.
That file is where the AI reads a scene and decides what to suggest. And it turns out that’s everything. “Ancient ruins” is a boring suggestion. “Chairs slowly growing human faces” hits different. The gap between those two outputs isn’t a feature — it’s a prompt. And the prompt is the actual product.
Reed and I rewrote that file more times than I rewrote anything else in the codebase. The quality of the suggestions is why someone uses the app twice instead of once. I didn’t fully understand that until I was staring at two outputs from the same scene and only one made me laugh.
There are 5 free generations. Lifetime. Not per day, not per week — 5, total. Then the paywall.
That number was deliberate. It’s enough to understand what the app does. It’s not enough to keep using it without paying. I’ve seen apps give you 30 free credits and then wonder why conversion is terrible. People use all 30, feel satisfied, and leave. Five is a taste.
Marcus wrote a 90-day launch plan before I asked for it. It included growth loops, influencer targeting, and a kill criteria section — specific metrics for when to pull the plug and stop paying for server costs.
That part I appreciated most. Most side projects don’t die — they just drift. Someone decides to keep paying $40/month “for a few more months” indefinitely, until the credit card finally expires and they feel relieved. Having a number to hit before a date makes the decision automatic.
Mine are set. I’m not going to tell you what they are, because then I’d feel accountable to the internet, and I’m not sure I want that.
AIery is live. aiery.app. Five free generations and then you decide.
I don’t know if it’ll work. I know it took a weekend. That felt worth something regardless.
