No. 046Mobile app

Brawler: an AI boxing coach app paid mostly up front by annual plans

MRR: $22,552
Total revenue: $154,724
Active subscriptions: 4,282

Credibility★★★★4/4Difficulty3/5

Verified by TrustMRR

In one line: Brawler is an AI boxing coach app for iPhone. A third party verifies its subscription revenue at about $22k a month. Most users pay for a full year at once, so whether they stay will only be known in May 2027.

Founder: George LampropoulosX ↗

Timeline

Apr 28First version submitted to the App Store
May 1Live on the App Store; on May 6 it added a paywall that offers a discounted annual plan when you close it
MayRevenue $4,152
JuneRevenue $17,996; 330 ratings
JulyWeight-loss mode launched; revenue $26,161
AugustSwitched to a stronger AI model and added beginner courses; revenue $23,482
SeptemberTikTok ads brought zero paying users; best month (see the table above); 1,200 ratings
OctoberAdded an AI sparring partner; price points went from 3 to 10

Compared with similar products

BrawlerMaxed (AI lifting coach) ↗CalBuddy (AI calorie app) ↗ChefGPT (AI recipes) ↗

What they started withA playbook that already worked (our inference) plus a dev partnerNo app experience; filmed his own videosCopied a similar app that already worked in the USAn AI recipe tool

Monthly revenue now$22.5k (verified)About $14k (founder's claim)Over $80k (founder's claim)$2k (verified)

What got them growingMeta ads, creators and annual plansAds made from his own footageAds first, then revenue share with fitness influencersNot yet; revenue is falling

5 moves Brawler made

Reused a playbook the founder had used before.
That it was reused is our inference.
Lists only two acquisition channels: Meta ads and creators.
Lists only two acquisition channels: Meta ads and creators.
On day 5 after launch, added a discounted annual offer that appears when you close the paywall.
Now $9.99 a month, $59.99 a year, or $29.99 a year discounted.
Tried many price points.
The App Store lists 10 subscription prices (our inference: price testing).
Shipped often and fixed weak spots raised in reviews.
More than twenty versions from late May to October; a July review said the coaching lacked depth, and beginner courses came in September (our inference: the two are related).

Method: Build an AI mobile appCompare all methods

How to copy it

Pick a niche you know and write the list of mistakes an AI coach should correct
Choose something you've practiced and can judge, and list the mistakes coaches correct most: for each, what it looks like, why it's wrong and how to drill it, grouped by the training plan's main categories. This list is the AI's scoring rubric for videos. What the founder did: the app builds plans by training category, reviews users' uploaded videos and gives corrections across several areas (App Store page).Expand ▾Collapse ▴
Choose something you've practiced and can judge, and list the mistakes coaches correct most: for each, what it looks like, why it's wrong and how to drill it, grouped by the training plan's main categories. This list is the AI's scoring rubric for videos. What the founder did: the app builds plans by training category, reviews users' uploaded videos and gives corrections across several areas (App Store page ↗).
Author post 1 ↗Original page 2 ↗
A mistakes list grouped by category, each with what it looks like, why it's wrong and a drill.
Someone who knows the field reviews the list and agrees these are the most common mistakes.
I've practiced <your niche> for <years> years and want to build an AI coach app. Below is my draft of common mistakes. Please: 1) complete and organize it, grouped by <training categories>; 2) for each: how it looks on video, why it's wrong, one corrective drill; 3) flag mistakes a phone video can't reveal; 4) add no expert claims I didn't make and mark uncertain ones "to confirm". Draft: <paste>
Generate the app with AI and build v1 of video analysis plus training plans
Build v1 with an AI mobile app builder doing three things only: training plans by the step 1 categories; upload a video and get the most important corrections from the list; a Q&A screen. Put the step 1 list verbatim into the prompt as the rubric. What the founder did: by his account Rork built the entire app (X post by @GeorgeLampro20), with Supabase as the backend (X article by @GeorgeLampro20).Expand ▾Collapse ▴
Build v1 with an AI mobile app builder doing three things only: training plans by the step 1 categories; upload a video and get the most important corrections from the list; a Q&A screen. Put the step 1 list verbatim into the prompt as the rubric. What the founder did: by his account Rork built the entire app (X post by @GeorgeLampro20 ↗), with Supabase as the backend (X article by @GeorgeLampro20 ↗).
Author post 1 ↗Author post 2 ↗
On a real phone, an uploaded video returns corrections citing list items; the plan screen shows plans by category.
One of <your target users> uploads their own video and feels the corrections hit the real problem.
Build v1 of an AI coach app for <your target users>: 1) a training plan screen with plans by the categories in the list below; 2) upload a video, call a model API that can read video or images, and return the top 3 corrections from the list, each citing its item; 3) a Q&A screen; 4) Supabase backend with secrets in environment variables. Follow official docs. List: <paste step 1 list>
- Rork
  Step 2: generate the mobile app with AI (our optional recommendation; the founder says he built Brawler with it, but his bio says Building with @Rork, so there may be a partnership).
- Expo
  Step 2: an alternative where you write the app yourself with an AI coding tool such as Cursor (our optional recommendation).
- Supabase
  Step 2: store users and analysis history; the founder's article says his app backends use Supabase.
  Author post ↗
Add monthly and annual subscriptions with RevenueCat, with the paywall after the first analysis result
Create monthly and annual subscription products in App Store Connect, connect them with RevenueCat, and show the paywall after the first video analysis: part free, all with a subscription. You set the prices. What the founder did: the paywall runs on RevenueCat (X article by @GeorgeLampro20), with monthly and annual plans (TrustMRR).Expand ▾Collapse ▴
Create monthly and annual subscription products in App Store Connect, connect them with RevenueCat, and show the paywall after the first video analysis: part free, all with a subscription. You set the prices. What the founder did: the paywall runs on RevenueCat (X article by @GeorgeLampro20 ↗), with monthly and annual plans (TrustMRR ↗).
Author post ↗
The paywall appears after the first analysis; a sandbox account can buy both plans; RevenueCat shows the first real subscription.
RevenueCat shows the first subscription not paid by you.
My AI coach app shows the paywall after the first video analysis: one correction free, all with a subscription, at $<price>/month or $<price>/year. Write: 1) English paywall copy: a headline referring to the correction they just saw, 3 things a subscription unlocks, a line per option; 2) English names and descriptions for the App Store subscription products. Use only my prices.
- RevenueCat
  Step 3: connect App Store subscriptions and the paywall; the founder's article names RevenueCat for the paywall, and TrustMRR verifies revenue through it.
  Author post ↗
Run small Meta ad tests with rough screen recordings and measure cost per paying user
Record a few short phone screen captures: uploading a video and the moment the AI points out a mistake. Run each on a small budget for a few days, match new subscriptions in RevenueCat, and compute cost per paying user. You set the budget. What the founder did: his channels are Meta ads and influencers (TrustMRR), and the notes of a podcast he joined say rough screen-recording ads beat polished ones (the episode covers several apps) (Superwall podcast).Expand ▾Collapse ▴
Record a few short phone screen captures: uploading a video and the moment the AI points out a mistake. Run each on a small budget for a few days, match new subscriptions in RevenueCat, and compute cost per paying user. You set the budget. What the founder did: his channels are Meta ads and influencers (TrustMRR ↗), and the notes of a podcast he joined say rough screen-recording ads beat polished ones (the episode covers several apps) (Superwall podcast ↗).
Original page 1 ↗Author post 2 ↗
A few screen-recording ads are live; a table of spend, subscriptions and cost per paying user per ad.
At least one ad's cost per paying user is below what the user pays.
I'm running Meta ads for an AI coach app using rough phone screen recordings. Write scripts for a few 15-second recordings: 1) each shows real use: upload a video, the AI points out a specific mistake; 2) a different 2-second opening for each; 3) English captions, no hype; 4) also give me table headers for computing cost per paying user per ad.