Story
The bet: a middle ground between scale-weighers and non-trackers
Cal AI's hypothesis was that there was an untapped middle market between hardcore users weighing food on a scale for decimal precision and people who don't track at all. They put capital behind testing it and hit their first six-figure month by June.
“The hypothesis was that, yeah, it would make people's lives easier, but we are almost hitting this interesting intersection between people are very hardcore with tracking their calories, weighing their food on a scale. Needing the precision decimal point accuracy, and then on the other end of people that don't track their calories at all. So our hypothesis was that there was a middle ground, and that's where we threw the capital to test, and it worked.”
Number
Cal AI hit $2M in a single month, a $24M run rate
Just 10 months after launch, Cal AI did $2M in revenue in one month, putting it on a roughly $24M annual pace, all bootstrapped.
$2M
Monthly revenue · USD/month
“Well, last month we just did our first $2 million.”
Framework
AI wrappers are the new dropshipping: validate before you build
Zach argues every app should start as a thin AI wrapper, the same way e-commerce founders start by dropshipping. The wrapper is the cheap proof of concept; once it works, you build your own deeper product.
“And it's true that we started as an AI wrapper. And I think that that's something all apps should do. Just like in e-com, it's very common to start as a dropshipper and dropship a product. And then once you find success, actually manufacture yourself, store it yourself, create your own brand out of that. So it's just the proof of concept is the dropshipping. For us, the proof of concept was using ChatGPT and other AI tools kind of as a wrapper app, frankly.”
Steal thisShip a thin AI wrapper as your proof of concept; only build deep tech after it sells.
Tactic
Lead the influencer DM with 'Paid promo?' to win the preview line
Cal AI A/B-tested cold influencer DMs and found leading with 'Paid promo?' works best, because the influencer only sees the first line or two in their inbox preview, so the money offer has to come first.
“paid promo, putting that first, the question mark, and then jumping into Our app lets you track calories just by taking a picture of your food. We think it would fit your audience and would love to work for you, with you. If you're interested, let us know and we can hop on a quick call. And so this works really well because it optimizes for that preview message the influencer will see in their inbox. They only see the first line or two.”
Steal thisOpen cold influencer DMs with 'Paid promo?' so the offer survives the inbox preview cutoff.
Story
He funded Cal AI by selling an unblocked-games site for schools
Before Cal AI, Zach built totallyscience.co, a site that let students play games past school content blockers. He grew it to 5M users via TikTok, monetized with ads, and sold it, using that money to seed Cal AI.
“So before building Cal AI, I actually built an unblocked gaming website my freshman year of high school. This website lets students play games in class while their teacher was teaching, bypassing the website blocking protocols. And I grew that to 5 million users through TikTok. It was generating money by putting ads on the site and then I sold it. So that's where most of the money came in that we put into this.”
Number
Sold a side-project games site for $100K at age 16
Zach's unblocked-games site totallyscience.co made $60K over two years, then he sold it for $100K when he was 16, his first real exit.
$100K
Sale price of website · USD
“It made $60,000 for 2 years and then I sold it for $100,000 when I was 16.”
Tactic
The 'comment what to add next' trick that farms viral comments
Growing his games site, Zach recorded his screen in class and ended each TikTok with a caption: 'comment what game to add next.' The prompt flooded videos with comments, which the algorithm rewarded with reach.
“And then I would put a little thing, a caption that says, comment what game to add next. And that would gather a ton of comments, making the video go viral.”
Steal thisEnd your short-form videos with a low-effort 'comment what to add next' prompt to farm engagement.
Tactic
Price influencer deals by predicting views and reading the comments
Influencers want paying up front, not CPM, so Cal AI predicts a post's views from past videos before paying. Crucially, not all views are equal: they read the comment section to gauge how strong the creator's community is and weigh the payment accordingly.
“So you have to predict before they make any video how many views they're going to get. You'll have to look at their previous videos, but then you also have to keep in mind not all views are worth the same. You have to analyze the comment section and see how strong their community is. And based on that, weigh how much you're actually going to pay them.”
Steal thisBefore paying an influencer up front, forecast views from past posts and discount for weak comment sections.
Tactic
95% of subscriptions are annual, so churn is unknown but cash is up front
Over 95% of Cal AI's subscriptions are annual plans. They literally can't measure churn yet because they haven't been around a full year, but annual billing pulls cash forward to fund growth.
“We actually don't know because it hasn't been a year and more than 95% of our subscriptions are annual.”
Steal thisDefault subscribers to annual plans to pull a full year of cash forward for marketing.
Framework
Marketing-first: build the app backwards from one AI aha moment
Zach's idea-generation method: take anything that doesn't yet use AI, find the single magic 'aha moment' AI can perform (snap a photo, get calories), make that the marketing hook, then wrap a sticky app around it to retain users.
“I want to find something looking at it from marketing-first principles. Is how I always think. So almost going backwards, and I look for an aha moment that I could capture within some sort of experience and then wrap a whole app around that. So for Cal AI, the aha moment is take a picture of your food, get the calories, and that's great for marketing material. They come on, they do that, but then there's a whole app around that that gets them to stay.”
Steal thisFind the one AI 'aha moment' that markets itself, then build a retention engine around it.
Idea
AI journal that surfaces life insights from your entries
Take an existing journal app (voice or typed) and add the aha feature: the AI periodically reads your entries and surfaces patterns, e.g. 'every day you hang out with Sally you have a bad day.'
“To build a journal app, and these journal apps already exist. So take an existing one, put your— and this is the spin. So you could make it voice notes, you can make it typing, whatever, doesn't matter. But the key feature, the aha moment AI feature you implement is that periodically you will have these insights generated from the AI on how you can improve your life. Like, hey, on Monday and Tuesday you hung out with Sally and you had a bad day. Maybe Sally is the cause of your bad days.”
Steal thisBolt periodic AI pattern-insights onto a journaling app as the retention hook.
Idea
An AI service that ports iOS apps to Android (and back)
Building Android cost Cal AI tens of thousands and stole dev time from iOS. Zach pitches an AI-powered system that ingests a native codebase and converts it to the other platform, run as an AI-heavy agency now, fully autonomous later.
“I think with all these AI tools, there is definitely the possibility to build something that lets you upload the codebase to One Native Project, and then it will convert it to the other. Now, right now, I think AI can probably do 90% of the work, but there will need to be a tiny bit of human intervention. So maybe this would be best done as an agency that's very AI-powered at the moment, but very soon it's going to be something where an AI agent can do it all for you.”
Steal thisRun an AI-powered agency that ports native iOS apps to Android; close the last 10% by hand.
Idea
Remotely-configurable onboarding flows so you skip App Store reviews
Every app rebuilds its onboarding survey from scratch, and any tweak requires an App Store update that takes days. Zach pitches a system to swap questions and A/B test onboarding screens remotely, tracking conversion and drop-off without resubmitting.
“But someone could really easily make a system where anyone can swap out the questions, remotely do A/B tests on these, which another problem here is that anytime you want to test out something new within your onboarding flow or within your app in general, you have to submit an update to the App Store, which could take a few days. So building out a system where you can build out the whole onboarding survey questions And then also change what the screens are, see how that affects conversion rates, see how that affects completion rate, drop-off rate remotely would be huge.”
Steal thisBuild remote-config onboarding so app teams A/B test paywall screens without an App Store resubmit.
Story
Why they killed the app-studio plan to stay all-in on Cal AI
Cal AI briefly tried being an app studio, reusing its marketing 'sauce' across many AI apps. They scrapped it: the time to scale a new app to six figures could instead add seven figures to Cal AI because retention, LTV, and funnel compound on the winner.
“But at the scale Cal AI is and the rate it's growing, we realized pretty quickly that it made more sense to stay full-time on Call AI because the same time it would take to build another app and scale it to 6 figures revenue, we could have added an additional 7 figures in revenue to Call AI just because everything boosts each other. Increasing retention will increase LTV, and as we increase retention, we could increase a funnel. So 1 1 can equal 3 instead of 2.”
Steal thisPour effort into compounding your winner rather than spreading thin across new apps.