Dec 11, 2024 · 49m · startup-ideas
Inside the Brain of a 17-Year-Old $1M/Month AI Founder
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
Seventeen-year-old CalAI founder Zach Yadegari joins Greg Isenberg to build and share a complete founder playbook for Dr. AI, an AI-powered consumer health app. Together, they break down end-to-end Figma UI/UX prototyping, prompt engineering workarounds, viral TikTok creator marketing, and subscription unit economics.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Greg holds 20.1% of the talking time here. How this is scored →
Greg as informed peer 3.3 Guest teaching 5.1 Guest disagreement 0.7 Greg pushing back 0.6
speaking balance: gold is Greg, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 17:34 Debunking Viral Shareability in Utility Apps
Zach directly rejects Greg's assumption that consumer utility apps should rely on social sharing, explaining that users find health marks embarrassing to post.
Hardest push from Greg ▶ 18:44 Questioning Low-Shareability Consumer App PrioritiesGreg presses back on the concept by arguing that as a consumer investor/builder, he prioritizes high-shareability apps over low-shareability products.
Biggest teaching moment ▶ 41:10 Influencer Deal Unit Economics MasterclassZach systematically breaks down the mathematical calculation of predicting creator views and calibrating upfront pay against estimated RPM to guarantee post profitability.
Greg holds their own ▶ 10:10 Educating Zach on Product Management 'Primitives'Greg intervenes with senior Silicon Valley product design terminology, explaining how Cal AI and RizGPT created reusable UI primitives for other developers.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Greg as informed peer | Guest teaching | Guest disagreement | Greg pushing back | Why |
|---|---|---|---|---|---|---|
| Ideating Dr. AI and Analyzing Healthcare AI Potential | 3 | 4 | 1 | 1 | Zach introduces the concept of Dr. AI inspired by personal anecdotes and recent LLM medical diagnostic studies. Greg validates the market need by citing healthcare access constraints in Canada and the US and anticipates pushback from traditional medical professionals. | |
| Wireframing Core Navigation and Activity History Feed | 4 | 4 | 0 | 0 | Zach sketches the core navigation and feed history in Figma, explaining the utility of single-player logs. Greg adds product design insight regarding user investment and switching costs against competitors. | |
| Designing the Scan Flow and UI Primitives | 6 | 3 | 0 | 1 | As Zach maps out the camera scan flow, Greg introduces the product management framework of UI 'primitives,' educating Zach on industry terminology which Zach politely acknowledges. | |
| Formatting Results, Gamification, and Growth Mechanics | 4 | 6 | 2 | 2 | Greg questions the shareability and viral mechanics of the app. Zach pushes back on viral loops for health apps, noting users find medical conditions embarrassing, and explains why paid acquisition and fear triggers work better than organic sharing. | |
| Wireframing the Symptoms Quiz and Diagnostic Log | 2 | 4 | 0 | 0 | Zach walks through wireframing a simplified, multi-step multiple choice questionnaire to diagnose symptoms cleanly without overwhelming the user. | |
| Sponsor Break: Boring Marketing SEO Solutions | 3 | 5 | 0 | 0 | Following an ad break and a discussion about citing sources, Zach demonstrates how to reverse-engineer app screens specifically to maximize visual impact during brief influencer video spots. | |
| Prompt Engineering and Bypassing LLM Guardrails | 2 | 7 | 1 | 0 | Zach conducts a live screen share demonstrating prompt engineering and jailbreaking techniques, using roleplay to bypass ChatGPT's standard medical refusal guardrails. | |
| TikTok Creator Discovery and Demographic Audience Targeting | 3 | 6 | 1 | 1 | Zach navigates TikTok to analyze influencer audience demographics by manually inspecting comment sections, demonstrating that doctor influencers reach medical students while mom creators reach the true target buyers. | |
| Influencer Unit Economics and Subscription Pricing Strategy | 3 | 7 | 1 | 0 | Zach breaks down the exact math behind profitable influencer campaigns using RPM vs CPM modeling and explains why weekly subscriptions capitalize on episodic repeat utility rather than ongoing retention. |
Statements from this episode (16)
Yadegari builds initial app UX himself before outsourcing visual polish
“Usually what I do is I do all of the UX myself, and then I'll hire someone either on Upwork or some somewhere else. Now I know a lot of talented designers, but in the past, that's what I would have done and have them make my bad designs look a lot prettier.”
Yadegari favors Instagram-style activity feeds for single-player utility apps
“I've actually really liked the concept of just showing a log, kind of like an Instagram feed of your content, even if it's a single person utility app.”
Isenberg: 'CalAI for X' represents a new multi-action scanning UI primitive
“Between, you know, between Riz, GPT, Cal AI, you guys have kind of invented this whole new primitive, and there's probably like a thousand different apps that like Cal AI for X could work for it.”
Yadegari: AI apps with high latency should use background loading
“What a lot of apps do is they bring you to an analyzing screen where it loads, and you just wait on the screen. I honestly hate this, and especially for something that can take 30 to 60 seconds, depending on how complex your prompt is, and when you're waiting …”
Yadegari: Cal AI Grew Primarily Through Influencer Marketing
“Well, Cal AI was grown through influencer marketing.”
Yadegari: Utility App Users Do Not Naturally Share Food Scans
“And what I noticed with a utility app like Cal AI is that people don't really care to share their food.”
Yadegari: AI health apps must use fear-mongering marketing to drive downloads
“But I think that the primary lever to drive downloads to something like this would have to be paid ads or influencer marketing, where we could get into it more later, but it would play on the emotions of a little bit of fear mongering on the idea that they mig…”
Yadegari: Sequential multiple-choice questions beat checkbox lists for app quizzes
“I think it would be, kind of to your point of making it digestible and gamified, instead of asking them, like, what are you feeling, and then giving a bunch of checkboxes and they could click anything, it would probably be better to ask a series of multiple ch…”
Yadegari: AI chatbot apps differentiate on prompting, not UI design
“This is another thing where you could just look up chatbot on the App Store and pull out the UI. It does not need to reinvent the wheel. The only thing that's important to reinvent here is the prompting, and that's probably, so before we go into that, I want t…”
Yadegari: App UI must be immediately clear in two-second influencer clips
“Showing it on an influencer video, when it, when the phone is only on the screen for two seconds, you need the viewer to understand what this app does and why they want to get it for themselves. And if it's unclear, if the app is too convoluted, then influence…”
Yadegari: Marketers must evaluate an influencer's audience rather than their niche
“When thinking about influencers, I always try to, I, it's always important to look at the audience of the influencer, not at the actual influencer. Their niche could be one thing, but their audience could be something completely different.”
Yadegari: Influencers refuse pay-per-view and require upfront payment
“Influencers never like to be paid on a basis where it's pay per view. So you always have to pay, tell them what they're getting paid up front.”
Yadegari benchmarks consumer app influencer campaigns at a $5 RPM
“Assuming an RPM of five dollars, which is what I usually start with an app.”
Yadegari: AI wingman apps monetize repeat resubscriptions on weekly models
“I have many friends who have built the apps like RizGPT and other of these AI wingman response apps. And what I've learned from them is that, first of all, they're on the weekly pricing model, and two, they're mostly capitalizing on the repeat users.”
Yadegari calls $7 per week app pricing steep and potentially unethical
“So when I think of weekly, usually I go for four dollars or seven dollars a week. I think seven dollars is pretty steep, and you could debate whether it's ethical to price that much for an app like this, because it comes out to close to 30 dollars a month if s…”
Yadegari: Most Cal AI clones rely on basic inaccurate prompts
“A lot of people will make GPT wrappers for things like Cal AI. We have so many clones right now, and most of them have done nothing to make the app more accurate. They just put a basic prompt in and have GPT estimate the calories, whatever that means from the …”