Daniel Heintzman
Head of Product at Cal AI
Where to hear and read about Daniel Heintzman2 episodes and tests
SubversiveHow Cal AI Hit $50M ARR in 2 Years by Nailing Product/Channel Fit on TikTok
SuperwallI speedran a new app from 0 to $2m/year in 6 months (here's how)Daniel Heintzman is Head of Product at Cal AI, the calorie tracking app MyFitnessPal bought after it passed $50M ARR in under two years, and before that he built the Sway dating app to $2M ARR in 6 months. The most useful thing to learn from him is to pick the distribution channel first, then shape the product, onboarding and pricing to win in it.
People were actually paying us before we built the app to get feedback on their profile
I speedran a new app from 0 to $2m/year in 6 months (here's how)
Daniel Heintzman: the story and the lessons
1Daniel Heintzman's path from big-company designer to Cal AI
Heintzman moved from engineer to designer to product, chasing zero-to-one work. His stops included Facebook, Robinhood and Snapchat. He says startups teach zero-to-one best and big companies teach the other side.
Seeing both sides is what he values most from his path.
Meta, Snap and Robinhood each taught him a different way to build. Meta's growth team was data-driven and ran many A/B tests. Snap leaned on intuition and the psychology of features, with a core product design team of only 10 or 15 people.
At Robinhood Crypto he worked on the trade flow while other designers handled the branding.
He joined Cal AI after watching its growth from outside. He noticed it a couple of months after launch while building a similar app, and began talking to Zach about 8 months in, at about $1.5M MRR.
He calls it probably one of the fastest consumer AI apps of the last few years. He says it was already making a couple million a month, with onboarding and paywall conversion above industry standard.
Cal AI sold to MyFitnessPal and Heintzman stayed on. The app launched in May 2024, passed $50M ARR in under two years and was acquired in late 2025. The founders were bootstrapped and started as teenagers.
Teams work as usual and the two apps stay separate. He sees a long growth runway and keeps looking for new app ideas.
2How Daniel built Sway to $2M ARR in 6 months
He chose dating because its users pay and profile analysis was unclaimed. His earlier idea, an AI financial advisor, was dropped because personal finance apps are hard to monetize. Sway launched in summer 2024 with two co-founders.
The idea came from the Hinge subreddit, where people post profiles for feedback. A private app could reach a bigger audience than public posting.
He charged about $15 for manual profile reviews before any app existed. In the first weeks, followers DMed profile screenshots and paid for feedback by hand. The ask sat in the caption, not in dedicated videos.
He recommends this step to anyone testing an idea, because real payments prove demand before code.
Sway built an Instagram audience with a pre-order link before shipping. A co-founder ran a new account of actionable, funny prompt tips that people could copy onto their profiles. Some posts did well within a couple of weeks.
By launch, Sway had thousands of paying customers: "we actually had distribution before we launched the product."
He batched a month of TikTok slideshows in an hour or two. Backgrounds were designed in Figma and slide text lived in a spreadsheet that filled the slides. Text was borrowed from the most viral slideshows of a men's improvement account.
It was his first month ever on TikTok. Videos pulled roughly 10K to 50K views, each with a plug for the app.
He kept the account unbranded until views were reliable. A fully branded new account likely would not have gone viral at first. Once views were dependable, the whole feed took on the orange brand theme.
A branded Christmas Day post reached about 4 million views and earned $5 to 6K in one day. He guesses people being home on Christmas helped it catch the algorithm.
Paid ads took Sway from $30-40K a month to $2M ARR. Ads started around the Christmas post, and he says growth became "a lot a lot easier." A creator's organic promo videos worked better as ads on his own page, and controversial ones drew comments and performed well.
Relying on one creator, Defunct Sibling, limited creative variety. At that stage the team had not yet learned to make videos built for ads.
3How Heintzman designs onboarding and paywalls to earn more per user
He builds onboarding as a sales funnel before the paywall. The steps are: clarify why the user is there, label the problem, show the dream result, personalize with questions, explain why the app is for them.
The dream result appears again on the paywall, and one guided path means users never have to think or choose. His stated aim is to raise intention to purchase.
Sway's first onboarding failed by asking questions with no context. Users emailed asking what they were even subscribing for. He added short videos on the problem and the solution, and made each question multiple choice.
He thinks Cal AI, which goes straight to questions, misses the point. Questions should be quick to answer so nothing slows users before the paywall.
He put the aha moment after the paywall and conversion improved. He tested both orders. Asking users to upload screenshots before paying risked them leaving to fetch images and never returning.
He still thinks a cheap lead-up to the aha moment before the paywall would be better, if it takes little effort.
He defaulted Sway to annual plans to fund growth. He had noticed that apps defaulting to annual had good LTV. A weekly plan was tested and grew much more slowly.
Cash collected up front pays for growth. He says Sway's LTV is higher than any other app he has seen, a claim with no data shown.
Uncapped photo packs let a few heavy spenders carry revenue. Packs sold at $14.99 and $49.99 on top of the subscription, and a credits upsell adds about 11% of revenue.
He says the top 1% of spenders drive most revenue at Tinder, Hinge and Bumble. The idea came from a co-founder who built his own game.
4Why Cal AI was built to be understood in a 10-second TikTok
Heintzman's team treated distribution as the biggest risk and promoted at once. The founders built the app in a month or two and began promoting in month two. The first version had one plus button to take a photo and show calories and macros, with no daily total.
It went viral on that single hook and reached $1.5M MRR about 8 months after launch. Heintzman notes that "a lot of people will spend like years perfecting the product."
Cheap AI coding moved the bottleneck from product to distribution. Anyone can code an app in a weekend, and hundreds of thousands of subscription apps now compete for attention on Meta, Google and TikTok.
A product needing more than a few seconds of explanation slows growth. That flips the old build-it-and-they-will-come mantra.
Cal AI's screens are designed to be read in someone else's video. Text is legible from a distance and the logo sits on every core screen. Creator tags are often too subtle, so the logo tells viewers which app to download.
Without it, viewers must dig through comments or rewatch to find the name.
A separate creator version of the app looks better on camera. It shows a streak and a history of logged meals, which adds social proof and makes each video look like real, ongoing use.
Every influencer video shows the same three screens. They are the camera, the loading screen and the food detail page. Every design choice serves a viewer who must grasp the app in seconds.
He would judge any new app idea by whether it is easy to understand in social videos. Cal AI favors clarity over standing out.
He tests each feature by whether its title alone sells it. He borrows Tony Fadell's press release method: the title should be enough to make people want to use the feature. Cal AI's own line is "take a picture and see the calories and macros."
5Heintzman's rules for working with creators
He picks fitness creators because their bodies prove the result. There are endless numbers of them, and creators pitching calorie tracking do not conflict with the coaching many of them sell.
Cal AI's proof comes from eating well. About 300 creators were posting in one month.
He judges creators by their worst posts, not their best. Cal AI looks for about 100k views per video consistently. Paid ads usually perform worse than a creator's normal content, so he expects the worst.
"If they can't do that consistently even on their worst post, then usually it's not going to work out."
He reads comments to find creators who actually drive downloads. Good signs are questions put to the creator and praise. Passive reactions mark viral videos that do not send people to the App Store.
Creator pay has an easy minimum view threshold plus bonuses. The minimum is set at a level very achievable for that creator, so a flop does not leave anyone unhappy. Bonuses apply above it, though no amounts are given.
He saw the failure case at earlier apps: a creator balked at the minimum, the video missed, and the creator stopped working with the team.
Creators get two rules: show the core screens, and show them early. The screens are the camera, the photo being analyzed and the result. Retention on most videos "goes straight down," so late placement loses viewers.
Everything else is left to the creator. Scripts would make posts look like ads, sink views and not scale to hundreds of creators.
He layered paid ads and affiliates on top once creators were profitable. The first ads were repurposed creator videos built for organic views. The paid team then started an affiliate program of dozens of creators making only Cal AI videos, paid by commission.
When the team went all in on paid ads, growth took a step change. Creator awareness also made ads convert better, since many people had heard of the app but not downloaded it.
He prefers creator-made content over content the team posts itself. Your own content is capped by how much you can push out, while influencers keep stacking up with built-in audiences.
An influencer test measures one variable, downloads and payment. UGC also tests whether you can go viral. Paid ads were the big inflection point at Cal AI and his previous app.
6How Cal AI keeps users and spreads by word of mouth
With conversion already strong, he focused on surviving the 3-day trial. Users need enough value in three days to keep paying. The team cut bugs and redesigned the food database, the camera and the food detail page.
Many Cal AI clones exist, so the team wanted features no clone could copy.
Groups and streaks borrow from Strava and Duolingo. Groups started as private friend groups and later added public groups by goal or topic, with leaderboards on streaks.
He calls these "different ideas we've tried," not proven wins.
Shareable progress adds friend proof on top of creator proof. Streaks, milestone badges and before and after photos give users something to post to Instagram or group chats.
"People just only want to download apps that other people that they know use." Most growth still comes from influencers and paid ads, and he cannot say how much sharing drives.
A $10 referral can be claimed before the user pays. Users can refer a friend during onboarding, which puts a sharing prompt early in the funnel.
The team could not isolate its impact because private sharing is hard to track.
Cal AI shows the product demo on the first screen and the paywall. A video of someone taking a photo and seeing calories matches what users saw on social media. The same video reminds them what they signed up for.
Onboarding was already strong, so the team has touched it less than people assume.
Heintzman prices far below rivals so sharing feels like an impulse. Cal AI costs about $30 a year against roughly $80 to $120 for incumbents. The founders bet that a cheap app gets recommended freely.
He says low pricing "helped with Calorie AI a lot." Sway took the opposite path and focused on monetization.
Cal AI took the simple, cheap, younger end of the market. MyFitnessPal serves serious users who accept friction. Cal AI draws first-time trackers, many with no weight goal.
It later added barcode scanning and a food database lookup for exact numbers. He expects people to pick one or the other by life stage.
A Thanksgiving promotion gave free access around a shared moment. People already post what they ate, so the promotion turned a natural sharing moment into brand awareness.
7How Daniel picks app ideas and niches
He wants ideas that are simple to show, novel and in a big market. Cal AI fit all three: photo calorie counting was novel, easy to film and in a market large enough for MyFitnessPal to exist.
A blank slate has two unknowns, the solution and the demand: "you're solving too many variables."
He finds novelty in the top complaint about leading apps. The test is to find the biggest apps in a category and make their main complaint the idea.
He notes that one therapy app spends heavily on Facebook ads, with Sensor Tower showing about $800K a month on iOS and $200K on Android.
He looks for viral content niches before building. A huge social niche can prove an app works without similar apps. Astra plugs into horoscope content on TikTok with a soulmate framing that had gone viral with other apps.
A broad app like Cal AI can be plugged into many niches, such as science-based lifting.
He checks that creators in a niche are many, cheap and non-competing. Fitness creators are plentiful and reasonably priced. Dating coaches sell profile feedback themselves, so promoting an AI version cannibalizes them.
Beauty creators get big-margin brand deals from makeup companies, so a looks-maxing app for women would struggle to afford them.
He sketched a plan to ship many apps and give creators equity. Cheap vibe coding lets one team turn out apps quickly while big creators promote the last one. Ideas he explored were sleep tracking, language learning, workouts and a plant identifier.
His former partners knew a talent agency that could reach creators with millions of followers and would take a cut of the equity. He says spray and pray is probably just as good.
8How Heintzman works, and what he got wrong
His team ships fast, often the same day an idea appears. Speed is the habit Cal AI values most. Joining MyFitnessPal adds more structure, which the team is learning from.
"We'll have an idea and like we'll ship it like the next day or same day."
He moves before motivation fades, which he says takes about two weeks. Cal AI's co-founders say motivation on a feature lasts about 2 weeks. Robinhood Crypto went from idea to launch in about a month and went viral.
He wants products that look unlike their rivals. Robinhood Crypto looked almost like a video game while Coinbase looked like a banking product. The Robinhood team made many branding variations on purpose.
He keeps a mood board of well-designed apps and revisits it.
Sway's share features were a hypothesis that usage did not support. Users rarely used the share actions, because people do not want to ask friends about their dating profile.
Users did rerun the flow more than expected, which led the team to add credit purchases for power users.
His claims rest on his own account, with little data shown. The $2M ARR, the doubled revenue per user and Sway's high LTV come only from Daniel. The ad costs, retention and churn were not shown.
Cal AI's creator costs, ad return and channel split are also missing, and his account of Cal AI covers events he joined late.
How Cal AI Hit $50M ARR in 2 Years by Nailing Product/Channel Fit on TikTok
Cal AI designed its calorie tracking app to be understood in a 10-second TikTok video and grew through hundreds of fitness creators, then layered paid ads and affiliates on top. The result was over $50M ARR in under two years and an acquisition by MyFitnessPal in late 2025.
Heintzman moved from engineer to designer to product, chasing zero-to-one work. His stops included Facebook, Robinhood and Snapchat. He says startups teach zero-to-one best and big companies teach the other side.
Seeing both sides is what he values most from his path.
He joined Cal AI after watching its growth from outside. He noticed it a couple of months after launch while building a similar app, and began talking to Zach about 8 months in, at about $1.5M MRR.
He calls it probably one of the fastest consumer AI apps of the last few years. He says it was already making a couple million a month, with onboarding and paywall conversion above industry standard.
Cal AI sold to MyFitnessPal and Heintzman stayed on. The app launched in May 2024, passed $50M ARR in under two years and was acquired in late 2025. The founders were bootstrapped and started as teenagers.
Teams work as usual and the two apps stay separate. He sees a long growth runway and keeps looking for new app ideas.
Heintzman's team treated distribution as the biggest risk and promoted at once. The founders built the app in a month or two and began promoting in month two. The first version had one plus button to take a photo and show calories and macros, with no daily total.
It went viral on that single hook and reached $1.5M MRR about 8 months after launch. Heintzman notes that "a lot of people will spend like years perfecting the product."
Cheap AI coding moved the bottleneck from product to distribution. Anyone can code an app in a weekend, and hundreds of thousands of subscription apps now compete for attention on Meta, Google and TikTok.
A product needing more than a few seconds of explanation slows growth. That flips the old build-it-and-they-will-come mantra.
Cal AI's screens are designed to be read in someone else's video. Text is legible from a distance and the logo sits on every core screen. Creator tags are often too subtle, so the logo tells viewers which app to download.
Without it, viewers must dig through comments or rewatch to find the name.
A separate creator version of the app looks better on camera. It shows a streak and a history of logged meals, which adds social proof and makes each video look like real, ongoing use.
He tests each feature by whether its title alone sells it. He borrows Tony Fadell's press release method: the title should be enough to make people want to use the feature. Cal AI's own line is "take a picture and see the calories and macros."
He picks fitness creators because their bodies prove the result. There are endless numbers of them, and creators pitching calorie tracking do not conflict with the coaching many of them sell.
Cal AI's proof comes from eating well. About 300 creators were posting in one month.
He judges creators by their worst posts, not their best. Cal AI looks for about 100k views per video consistently. Paid ads usually perform worse than a creator's normal content, so he expects the worst.
"If they can't do that consistently even on their worst post, then usually it's not going to work out."
He reads comments to find creators who actually drive downloads. Good signs are questions put to the creator and praise. Passive reactions mark viral videos that do not send people to the App Store.
Creator pay has an easy minimum view threshold plus bonuses. The minimum is set at a level very achievable for that creator, so a flop does not leave anyone unhappy. Bonuses apply above it, though no amounts are given.
He saw the failure case at earlier apps: a creator balked at the minimum, the video missed, and the creator stopped working with the team.
Creators get two rules: show the core screens, and show them early. The screens are the camera, the photo being analyzed and the result. Retention on most videos "goes straight down," so late placement loses viewers.
Everything else is left to the creator. Scripts would make posts look like ads, sink views and not scale to hundreds of creators.
He layered paid ads and affiliates on top once creators were profitable. The first ads were repurposed creator videos built for organic views. The paid team then started an affiliate program of dozens of creators making only Cal AI videos, paid by commission.
When the team went all in on paid ads, growth took a step change. Creator awareness also made ads convert better, since many people had heard of the app but not downloaded it.
Shareable progress adds friend proof on top of creator proof. Streaks, milestone badges and before and after photos give users something to post to Instagram or group chats.
"People just only want to download apps that other people that they know use." Most growth still comes from influencers and paid ads, and he cannot say how much sharing drives.
A $10 referral can be claimed before the user pays. Users can refer a friend during onboarding, which puts a sharing prompt early in the funnel.
The team could not isolate its impact because private sharing is hard to track.
Cal AI shows the product demo on the first screen and the paywall. A video of someone taking a photo and seeing calories matches what users saw on social media. The same video reminds them what they signed up for.
Onboarding was already strong, so the team has touched it less than people assume.
Heintzman prices far below rivals so sharing feels like an impulse. Cal AI costs about $30 a year against roughly $80 to $120 for incumbents. The founders bet that a cheap app gets recommended freely.
He says low pricing "helped with Calorie AI a lot." Sway took the opposite path and focused on monetization.
Cal AI took the simple, cheap, younger end of the market. MyFitnessPal serves serious users who accept friction. Cal AI draws first-time trackers, many with no weight goal.
It later added barcode scanning and a food database lookup for exact numbers. He expects people to pick one or the other by life stage.
A Thanksgiving promotion gave free access around a shared moment. People already post what they ate, so the promotion turned a natural sharing moment into brand awareness.
His team ships fast, often the same day an idea appears. Speed is the habit Cal AI values most. Joining MyFitnessPal adds more structure, which the team is learning from.
"We'll have an idea and like we'll ship it like the next day or same day."
His claims rest on his own account, with little data shown. The $2M ARR, the doubled revenue per user and Sway's high LTV come only from Daniel. The ad costs, retention and churn were not shown.
Cal AI's creator costs, ad return and channel split are also missing, and his account of Cal AI covers events he joined late.
I speedran a new app from 0 to $2m/year in 6 months (here's how)
Daniel shows how he built an audience and proof of willingness to pay before writing code, then used influencer-made ads, an annual-default paywall and consumable upsells to reach $2M ARR in 6 months with a dating profile analyzer. He also explains what he carried over to Cal AI, including retention features and a framework for picking app ideas.
Meta, Snap and Robinhood each taught him a different way to build. Meta's growth team was data-driven and ran many A/B tests. Snap leaned on intuition and the psychology of features, with a core product design team of only 10 or 15 people.
At Robinhood Crypto he worked on the trade flow while other designers handled the branding.
He joined Cal AI after watching its growth from outside. He noticed it a couple of months after launch while building a similar app, and began talking to Zach about 8 months in, at about $1.5M MRR.
He calls it probably one of the fastest consumer AI apps of the last few years. He says it was already making a couple million a month, with onboarding and paywall conversion above industry standard.
Cal AI sold to MyFitnessPal and Heintzman stayed on. The app launched in May 2024, passed $50M ARR in under two years and was acquired in late 2025. The founders were bootstrapped and started as teenagers.
Teams work as usual and the two apps stay separate. He sees a long growth runway and keeps looking for new app ideas.
He chose dating because its users pay and profile analysis was unclaimed. His earlier idea, an AI financial advisor, was dropped because personal finance apps are hard to monetize. Sway launched in summer 2024 with two co-founders.
The idea came from the Hinge subreddit, where people post profiles for feedback. A private app could reach a bigger audience than public posting.
He charged about $15 for manual profile reviews before any app existed. In the first weeks, followers DMed profile screenshots and paid for feedback by hand. The ask sat in the caption, not in dedicated videos.
He recommends this step to anyone testing an idea, because real payments prove demand before code.
Sway built an Instagram audience with a pre-order link before shipping. A co-founder ran a new account of actionable, funny prompt tips that people could copy onto their profiles. Some posts did well within a couple of weeks.
By launch, Sway had thousands of paying customers: "we actually had distribution before we launched the product."
He batched a month of TikTok slideshows in an hour or two. Backgrounds were designed in Figma and slide text lived in a spreadsheet that filled the slides. Text was borrowed from the most viral slideshows of a men's improvement account.
It was his first month ever on TikTok. Videos pulled roughly 10K to 50K views, each with a plug for the app.
He kept the account unbranded until views were reliable. A fully branded new account likely would not have gone viral at first. Once views were dependable, the whole feed took on the orange brand theme.
A branded Christmas Day post reached about 4 million views and earned $5 to 6K in one day. He guesses people being home on Christmas helped it catch the algorithm.
Paid ads took Sway from $30-40K a month to $2M ARR. Ads started around the Christmas post, and he says growth became "a lot a lot easier." A creator's organic promo videos worked better as ads on his own page, and controversial ones drew comments and performed well.
Relying on one creator, Defunct Sibling, limited creative variety. At that stage the team had not yet learned to make videos built for ads.
He builds onboarding as a sales funnel before the paywall. The steps are: clarify why the user is there, label the problem, show the dream result, personalize with questions, explain why the app is for them.
The dream result appears again on the paywall, and one guided path means users never have to think or choose. His stated aim is to raise intention to purchase.
Sway's first onboarding failed by asking questions with no context. Users emailed asking what they were even subscribing for. He added short videos on the problem and the solution, and made each question multiple choice.
He thinks Cal AI, which goes straight to questions, misses the point. Questions should be quick to answer so nothing slows users before the paywall.
He put the aha moment after the paywall and conversion improved. He tested both orders. Asking users to upload screenshots before paying risked them leaving to fetch images and never returning.
He still thinks a cheap lead-up to the aha moment before the paywall would be better, if it takes little effort.
He defaulted Sway to annual plans to fund growth. He had noticed that apps defaulting to annual had good LTV. A weekly plan was tested and grew much more slowly.
Cash collected up front pays for growth. He says Sway's LTV is higher than any other app he has seen, a claim with no data shown.
Uncapped photo packs let a few heavy spenders carry revenue. Packs sold at $14.99 and $49.99 on top of the subscription, and a credits upsell adds about 11% of revenue.
He says the top 1% of spenders drive most revenue at Tinder, Hinge and Bumble. The idea came from a co-founder who built his own game.
Every influencer video shows the same three screens. They are the camera, the loading screen and the food detail page. Every design choice serves a viewer who must grasp the app in seconds.
He would judge any new app idea by whether it is easy to understand in social videos. Cal AI favors clarity over standing out.
He picks fitness creators because their bodies prove the result. There are endless numbers of them, and creators pitching calorie tracking do not conflict with the coaching many of them sell.
Cal AI's proof comes from eating well. About 300 creators were posting in one month.
He prefers creator-made content over content the team posts itself. Your own content is capped by how much you can push out, while influencers keep stacking up with built-in audiences.
An influencer test measures one variable, downloads and payment. UGC also tests whether you can go viral. Paid ads were the big inflection point at Cal AI and his previous app.
With conversion already strong, he focused on surviving the 3-day trial. Users need enough value in three days to keep paying. The team cut bugs and redesigned the food database, the camera and the food detail page.
Many Cal AI clones exist, so the team wanted features no clone could copy.
Groups and streaks borrow from Strava and Duolingo. Groups started as private friend groups and later added public groups by goal or topic, with leaderboards on streaks.
He calls these "different ideas we've tried," not proven wins.
Shareable progress adds friend proof on top of creator proof. Streaks, milestone badges and before and after photos give users something to post to Instagram or group chats.
"People just only want to download apps that other people that they know use." Most growth still comes from influencers and paid ads, and he cannot say how much sharing drives.
Heintzman prices far below rivals so sharing feels like an impulse. Cal AI costs about $30 a year against roughly $80 to $120 for incumbents. The founders bet that a cheap app gets recommended freely.
He says low pricing "helped with Calorie AI a lot." Sway took the opposite path and focused on monetization.
He wants ideas that are simple to show, novel and in a big market. Cal AI fit all three: photo calorie counting was novel, easy to film and in a market large enough for MyFitnessPal to exist.
A blank slate has two unknowns, the solution and the demand: "you're solving too many variables."
He finds novelty in the top complaint about leading apps. The test is to find the biggest apps in a category and make their main complaint the idea.
He notes that one therapy app spends heavily on Facebook ads, with Sensor Tower showing about $800K a month on iOS and $200K on Android.
He looks for viral content niches before building. A huge social niche can prove an app works without similar apps. Astra plugs into horoscope content on TikTok with a soulmate framing that had gone viral with other apps.
A broad app like Cal AI can be plugged into many niches, such as science-based lifting.
He checks that creators in a niche are many, cheap and non-competing. Fitness creators are plentiful and reasonably priced. Dating coaches sell profile feedback themselves, so promoting an AI version cannibalizes them.
Beauty creators get big-margin brand deals from makeup companies, so a looks-maxing app for women would struggle to afford them.
He sketched a plan to ship many apps and give creators equity. Cheap vibe coding lets one team turn out apps quickly while big creators promote the last one. Ideas he explored were sleep tracking, language learning, workouts and a plant identifier.
His former partners knew a talent agency that could reach creators with millions of followers and would take a cut of the equity. He says spray and pray is probably just as good.
He moves before motivation fades, which he says takes about two weeks. Cal AI's co-founders say motivation on a feature lasts about 2 weeks. Robinhood Crypto went from idea to launch in about a month and went viral.
He wants products that look unlike their rivals. Robinhood Crypto looked almost like a video game while Coinbase looked like a banking product. The Robinhood team made many branding variations on purpose.
He keeps a mood board of well-designed apps and revisits it.
Sway's share features were a hypothesis that usage did not support. Users rarely used the share actions, because people do not want to ask friends about their dating profile.
Users did rerun the flow more than expected, which led the team to add credit purchases for power users.
His claims rest on his own account, with little data shown. The $2M ARR, the doubled revenue per user and Sway's high LTV come only from Daniel. The ad costs, retention and churn were not shown.
Cal AI's creator costs, ad return and channel split are also missing, and his account of Cal AI covers events he joined late.
Cal AI