Aiolum Submit a story
All stories
AI language tutor app · Canada

Verbi

Small teamB2CEducation
Outcome
US$251K verified revenue in 6 months, October 2026Revenue falling
Launch
2 April 2026
Team
The founder and his girlfriend at first; the profile says 2 to 5 people, plus paid video creators
Outside funding
None reported; the profile says bootstrapped
Group
Learning and study
Main channel
Creators and short videos

Verbi is an iPhone and Android app for learning to speak a foreign language. The learner holds a button, speaks, and an AI tutor answers in that language, gives corrections and scores the session. This needs speech recognition, software that turns recorded speech into text, and a language model, a program trained on large amounts of text that writes the tutor's replies. Verbi rents both from outside AI providers, which its privacy policy does not name. The app is free to download and sells a subscription as an in-app purchase, a payment made inside the app that Apple or Google collects, keeping a commission.

Pedro Machado, a software engineer in Vancouver who also teaches web development on YouTube, released it on 2 April 2026. The publisher is Acadia Labs Inc., a Canadian company. It was his second app: the first, a dream journal released in February 2026, reached about US$100 a month. For Verbi he and his girlfriend posted short videos on about 12 TikTok and Instagram accounts, most of them slideshows generated by a tool he built, and later he paid creators for UGC, user-generated content: videos that look like an ordinary user's post but are paid for by the company. Revenue was US$685 in April, US$53,969 in July and US$128,225 in September. By 13 October 2026 the total was US$250,824, with 4,569 active subscriptions and MRR, monthly recurring revenue, of US$21,136. MRR is the sum of all active subscriptions with each one converted to its price per month, so a yearly plan of US$47.99 counts as US$4.00.

The revenue, MRR and subscription figures are grade A: TrustMRR read them from the company's account at RevenueCat, a service that manages in-app subscriptions for app developers and records the payments made through Apple and Google. The profile does not say whether its revenue is before or after the store's commission; RevenueCat's own documentation defines its standard revenue figure as before store fees and taxes. Everything about the method comes from the founder's posts, grade C. The app is six months old, half of its revenue came in one month, and no yearly subscriber has reached a renewal date yet. The page also shows what happened to earlier companies that sold conversation practice with human teachers.

Verbi: US$251K verified revenue in 6 months, October 2026. AI language tutor app, Canada, team of 2 to 10, evidence grade A.
The story in one card. Share it with the address of this page.
01

The numbers, graded

Each row shows the evidence grade and the source numbers from the list at the end of this page. The grades are explained in How Aiolum verifies.

US$250,824 in verified revenue in the six months after launch, half of it in September 2026, read from the company's RevenueCat account
FactValueGradeHow we knowSources
Verified revenue, all time to 13 October 2026US$250,824ATrustMRR, read from the RevenueCat API, last synced on 13 October 2026 at 17:59 UTC. The first revenue is in April 2026. The cached profile of 12 October showed US$249,315. This is the headline figure.[1]
Verified revenue, 30 days to 13 October 2026US$64,495 in one monthAOne month of sales. It is not a yearly rate and should not be multiplied by 12: the profile marks it 38.3% below the 30 days before.[1]
Verified revenue by calendar month, 2026April US$685, May US$3,247, June US$9,561, July US$53,969, August US$36,812, September US$128,225, October to the 13th US$18,325AThe monthly table of the profile. The seven values add up to the all-time total.[1]
Share of September 2026 in all revenue51.1%calcCalculated: 128,225 / 250,824. July, August and September together are 87.3%.[1]
Revenue per day, early September against early October 2026about US$6,312 a day, then about US$1,451 a daycalcCalculated: the September total minus the profile's daily values for 14 to 30 September leaves US$82,053 for 1 to 13 September, divided by 13. The daily values for 1 to 12 October add up to US$17,417, divided by 12. The second value is 77% lower.[1]
Best single dayUS$16,889 on 2 September 2026CRevenueCat chart in the founder's X post of 3 September, read from the image. He wrote that he had never imagined a "16k day". His posts do not state the cause of that day.[18]
MRR, 13 October 2026US$21,136AUp 33.3% in 30 days according to the profile. RevenueCat converts each active subscription to one month: a yearly price is divided by 12 and a weekly price is multiplied by 4.[1, 5]
Active subscriptions, 13 October 20264,569ARevenueCat counts paid subscriptions that have not expired. A subscriber who has cancelled stays in the count until the paid period ends.[1, 4]
Average MRR per active subscriptionUS$4.63 a month, or US$55.51 a yearcalcCalculated: 21,136 / 4,569, and that result times 12. In MRR a yearly plan at US$47.99 counts as US$4.00 a month and one at US$59.99 as US$5.00, while a weekly plan at US$7.99 counts as US$31.96. An average of US$4.63 fits a base that is almost all yearly plans. The real mix of plans and countries is not published.[1, 5, 6]
The same average on earlier datesUS$7.21 on 12 July, US$5.01 on 21 July, US$4.49 on 2 September 2026calcCalculated from the founder's RevenueCat screenshots: MRR of US$6,781 with 941 subscriptions, US$5,899 with 1,178, and US$10,070 with 2,241. Between 12 and 21 July the subscriptions rose and the MRR fell, which fits weekly plans ending while yearly plans were added. On 16 July he wrote that his MRR does not grow because he optimizes for the yearly plan.[11, 16, 18]
30-day revenue compared with MRR3.05 timescalcCalculated: 64,495 / 21,136. A yearly plan is paid in full on the day of purchase but adds only one twelfth of its price to MRR. Most of the revenue is therefore new sales, which stop when new buyers stop.[1, 5]
Is the revenue before or after the store's commission?the profile does not sayCRevenueCat's documentation defines its Revenue figure as gross: before store fees and taxes, minus refunds. It reports the amount after taxes and commission separately, under the name Proceeds. TrustMRR's FAQ says only that it imports "the available revenue" from RevenueCat, and that its rebuilt figures can differ from a provider's own by up to 30%.[1, 3, 4]
Apple's commission on a subscription30% in a subscriber's first year, or 15% in the Small Business ProgramBApple's developer pages: the developer receives 70% of the price in the subscriber's first year and 85% after it, or 85% from the start when enrolled in the Small Business Program, which is open to developers with up to US$1 million in proceeds in the prior calendar year. Whether Acadia Labs is enrolled is not public.[20]
What would remain after Apple's commission, as an illustrationUS$213,200 at 15%, or US$175,577 at 30%calcCalculated: 250,824 times 0.85 and times 0.70, as if the whole amount were gross App Store sales. The split between the App Store, Google Play and web sales is not published, so this row only shows the size of the commission.[1, 20]
Profit margin, last 30 days55%CA field of the TrustMRR profile that is not in its list of verified metrics. 55% of US$64,495 is US$35,472 (calculated). On 12 July 2026 the founder wrote "80% margins". The costs are not published.[1, 11]
Figures on the day Verbi joined TrustMRR, 28 August 2026MRR of about US$8,500, 30-day revenue of about US$29,000, about 1,900 active subscriptionsATrustMRR's own X post. It also gave the total revenue as about US$29,000, the same as the 30-day figure, while the profile today shows US$53,969 for July alone. The first import probably held only 30 days of history; TrustMRR does not explain the difference.[2, 1]
Founder's dashboard, 28 days to 21 July 2026US$46,092 of revenue, 1,178 active subscriptions, 1,134 active trials, 33,240 new customersCRevenueCat Overview screenshot in his X post, read from the image. He wrote "45k in the last 28 days", up from 10k, and that one working video format is enough. A new customer in RevenueCat is a newly created user ID, which is not the same as a store download.[16, 4]
Founder's dashboard, 2 September 20266,167 active trials, 2,241 active subscriptions, 56,323 new customers in 28 daysCRevenueCat Overview screenshot in his X post of 2 September, read from the image. On 21 July the same screen showed 1,134 active trials.[18, 16]
First six weeks2,570 downloads and US$387 of MRRCFounder's X post of 23 May 2026, with "around 3k USD" in the same line. TrustMRR shows US$3,932 for April and May together (calculated: 685 + 3,247). He also gives about 10% to 14% "initial conversion" and 15% to 40% "trial conversion".[12, 1]
Who posted the videos, June 2026the founder and his girlfriend, on 12 TikTok and Instagram accountsCFounder's X post of 3 June 2026. No paid creator had been hired yet.[14]
Volume and cost of the videos, May 2026more than 60 posts a day, of which he made 2 himself; under US$5 for thousands of generated videosCFounder's X post of 13 May 2026, when Verbi had 100 active subscriptions. He built an internal tool that generates TikTok and Instagram slideshows. On 2 June he added that each extra phone costs US$100 and holds 3 TikTok and 3 Instagram accounts.[13, 11]
Advertising spend up to 9 June 2026US$0CFounder's X post: US$4,300 of revenue in 28 days with about US$500 of MRR and no ad spend. In the same post he starts to hire creators and to boost videos on TikTok, which normally means paying the platform to show a post to more people.[15]
Paid creatorsa first group of 10 by 21 July 2026; a campaign manager sought on 10 SeptemberCFounder's X posts. What the creators are paid is not published.[16, 11]
App Store rating, United States, 13 October 20264.6 out of 5 from 386 ratingsBApple's listing. The Canadian storefront has 57 ratings with an average of 4.4. Google Play shows 4.0 stars from 501 reviews and more than 50,000 downloads.[6, 7, 8]
Versions released18 between 2 April and 2 September 2026BCounted in the version history of the App Store listing. The notes first mention a voice practice mode in version 4.0 on 3 June 2026. No version is listed after 5.6.3 of 2 September.[6]
Team2 to 5 peopleCA field of the TrustMRR profile filled in by the founder, not verified. His posts name himself and his girlfriend, then paid creators. His X profile also lists a job as a software engineer at Twitch.[1, 11, 14]
Founder's audienceabout 293,000 YouTube subscribers and 1,425 X followersCThe PedroTech channel page and his X profile on 13 October 2026. The channel teaches web development. His posts credit TikTok and Instagram videos for the sales, not this audience.[19, 11]
Outside fundingnone reportedCThe TrustMRR profile says bootstrapped, a field filled in by the founder.[1]
AI model providersnot namedCThe privacy policy says that microphone audio is streamed to third-party AI providers for transcription, replies and lesson delivery, and that scoring, translations and grammar feedback are processed by AI providers. It names none of them.[10]
Benchmark for renewals, not about Verbinearly 30% of annual subscriptions are cancelled in the first monthCRevenueCat's State of Subscription Apps 2025 report, drawn from apps that use its platform. It also says that most apps with cheap annual plans keep up to 36.0% of users after a year. Verbi's own renewal rate cannot exist before April 2027.[21]
Time from launch to the headline figure6 monthscalcCalculated: 2 April 2026 to 13 October 2026 is 194 days. Apple's record gives the release as 1 April 2026 at 07:00 UTC; the version history and the founder say 2 April.[6, 12]

How to read this table: a grade A value was read by TrustMRR from the company's RevenueCat account through an API key; it is not a founder statement. A grade B value is published by Apple or Google. A grade C value is a founder statement, a profile field that the founder filled in, or a vendor's document. A "calc" value was computed on this page from the rows named in its note.

Verbi in-app purchases listed in the US App Store on 13 October 2026 (grade B); the listing does not say which price a new user is shown
PlanPer month, billed yearly (US$)Per month, billed monthly (US$)Limit
Pro, yearly at US$47.994.00listed under three names at this price
Pro, yearly at US$59.995.00named Verbi Annual Pro Plan
Pro, yearly at US$69.995.83named Verbi Yearly - Pro
Premium, yearly at US$79.996.67named Verbi Premium Yearly Plan
Pro, yearly with discount at US$29.992.50named Verbi Yearly Pro Plan - Discount
Pro, weekly7.99 a weekalso listed at US$8.99 a week; counted as US$31.96 or US$35.96 a month in MRR
Download00lessons need a subscription

The app is free to download and a subscription is required for full access. Apple's listing shows ten in-app purchases: seven yearly and three weekly, and no monthly plan. Several plans appear more than once at different prices, which usually means that the developer tests prices. The first column shows the yearly price divided by 12, computed on this page. The Canadian storefront lists the same plans in Canadian dollars: yearly from 39.99 to 108.99 and weekly at 9.99 or 11.99. A Google Play reviewer describes a free trial of 3 days. Prices on Google Play and on the web purchase page were not read.

02

The timeline

From a first app with US$45 of monthly recurring revenue in March 2026 to US$250,824 of verified revenue for the second app by October 2026. Each entry carries its grade.

  1. 2025 and earlierCMachado publishes web development tutorials on his YouTube channel, PedroTech, and announces them on X. [11, 19]
  2. 5 February 2026CHis first app, Lucidia, a dream journal, is released in the App Store by Acadia Labs Inc. He built it with his girlfriend. [11, 6]
  3. 10 March 2026CLucidia has 6 subscribers, 57 trials and US$45 of MRR from about 12 TikTok posts a day. On 27 March it reaches US$100 of MRR, and he moves his attention to a new app, Verbi. [11]
  4. 2 April 2026BVerbi version 1.0 appears in the App Store. Four updates follow in April, with a home screen widget, verb drills, crosswords and three more languages. [6]
  5. 13 May 2026C100 active subscriptions. He makes 2 videos a day himself and posts more than 60, generated by his own tool. [13]
  6. 23 May 2026C2,570 downloads and US$387 of MRR after about six weeks. [12]
  7. 3 June 2026CVersion 4.0 adds a voice practice mode. The same day he writes that sales are rising fast with no paid creator yet: he and his girlfriend post on 12 accounts. [6, 14]
  8. 9 June 2026CUS$4,300 of revenue in 28 days and no ad spend. He starts to hire creators. [15]
  9. 21 June 2026CPaid subscriptions have doubled in 20 days. He has cut the number of TikTok accounts by half and kept the ones that work. [11]
  10. 5 to 12 July 2026CHe writes that one video spread widely. On 12 July his dashboard shows US$33,484 in 28 days and 941 active subscriptions. [11]
  11. 21 July 2026CUS$46,092 in 28 days. A first group of 10 paid creators is complete. [16]
  12. 22 July 2026CA web version and a web purchase page go live. The first 9 customers pay outside the App Store. [17]
  13. July 2026AVerified revenue for the month: US$53,969, after US$9,561 in June. [1]
  14. 28 August 2026AVerbi joins TrustMRR with about US$8,500 of MRR and about 1,900 active subscriptions. [2]
  15. 2 September 2026CUS$16,889 of revenue in one day, 6,167 active trials, and MRR passes US$10,000. Version 5.6.3 is released the same day. [18, 6]
  16. 10 September 2026CHe looks for a manager for what he calls a mid-sized group of creators. [11]
  17. September 2026AUS$128,225 of verified revenue, the best month and 51.1% of all revenue to date. [1]
  18. 14 September to 12 October 2026ADaily revenue falls from US$4,096 on 14 September to between US$863 and US$1,951 a day in October. [1]
  19. 13 October 2026AUS$250,824 in total, 4,569 active subscriptions and MRR of US$21,136. The app is live in the App Store and on Google Play. [1, 6, 8]
03

Why it worked

Four causes, in the order they mattered. Each one is a reusable lesson.

01

Many cheap videos on many accounts

In May 2026 he made 2 videos a day himself and posted more than 60, most of them slideshows that his own tool generated at under US$5 for thousands of videos. Each extra phone cost US$100 and held 3 TikTok and 3 Instagram accounts. His stated rule is to copy the kind of video that other apps in the category already post, then to vary it. In June he cut the number of TikTok accounts by half and kept the ones that performed best. Up to 9 June 2026 he had spent nothing on advertising.

02

One format that worked, then paid creators

In early July 2026 one video spread widely, and revenue for 28 days went from about US$10,000 to US$46,092. He wrote that one working format is enough. He then paid a first group of 10 creators to make more videos, and in September he looked for a manager for the group.

03

Yearly plans, paid on the first day

The App Store lists weekly and yearly plans and no monthly plan. The average MRR per subscription, US$4.63, fits a base that is almost all yearly plans. A yearly buyer pays the full price at once, so revenue for 30 days was 3.05 times MRR, and that cash can pay for the next videos immediately. The cost is that MRR looks small, and that nobody knows yet how many of these buyers will renew.

04

A tutor that is rented, not employed

The voice lessons run on outside AI providers that transcribe the learner's speech, write the tutor's replies and score the session. There is no teacher to recruit, schedule or pay by the hour, which is what allows a yearly price of US$47.99 to US$79.99. The voice practice mode arrived in version 4.0 on 3 June 2026, and 87.3% of all revenue came in the three full months after it.

05

A first app that failed at low cost

Lucidia, the dream journal, did not pass about US$100 of MRR. He wrote that it did not really solve a problem, so videos with many views brought few buyers. He kept the method and changed the product to language learning.

How it was built

The technology list is the founder's entry on the TrustMRR profile. The other details come from his posts, the privacy policy and the store listing. They are grade C, except the version history, which Apple publishes. [1, 6, 9, 10, 11, 13, 17]

  • React Native for the app, Node.js for the server and RevenueCat for subscriptions. His first app was built with Expo, a toolkit for React Native.
  • Outside AI providers transcribe the learner's speech in real time, generate the tutor's replies, and do the scoring, translations and grammar feedback. The privacy policy names none of them.
  • He writes code with Cursor, an AI code editor, together with Claude, and runs a long audit prompt for bugs, leftover text, security and performance before each release.
  • 18 versions in five months: vocabulary and a home screen widget first, then verb drills and word games, three more languages on 29 April, the voice practice mode on 3 June, a new voice curriculum on 14 July and an interface in more languages on 24 August.
  • Videos: an internal tool that generates TikTok and Instagram slideshows, phones at US$100 each for 3 TikTok and 3 Instagram accounts, and later paid creators.
  • Sales through Apple and Google in-app purchases, and since 22 July 2026 also through a web app and a web purchase page with other billing partners.

Tools named in the sources

Each one opens its page in the Tools tab, with every story that uses it.

React NativeCNode.jsCNext.jsBCursorCClaude (assistant)CVercelBRevenueCatAApple App StoreAGoogle PlayATikTokCInstagramCTrustMRRA
04

The dead that came before

Five companies in the Loot Drop database sold the same promise, speaking a language with a tutor, and used human teachers. The values below are from that database and are not independently verified.

Died 2013 to 2021Burned US$1.1B

VIPKid

The promise

One-to-one video English lessons for Chinese children with North American teachers.

Why it died

China's 2021 tutoring rules banned foreign teachers and foreign curriculum, which was the whole product.

Legal/RegulatoryLoot Drop entry ↗
Died 2006 to 2024Burned US$600M

Meten EdtechX

The promise

Face-to-face English classes for adults in China, at a price far above apps.

Why it died

Winning and teaching a student cost more than the student paid. The pandemic and regulation ended it.

Unit EconomicsLoot Drop entry ↗
Died 2004 to 2021Burned US$300M

TutorGroup (iTutor)

The promise

Live online English lessons for learners in Asia with native-speaking tutors.

Why it died

A high cost to win each student and tutors who left, and then the 2021 ban on foreign tutors in China.

Legal/RegulatoryLoot Drop entry ↗
Died 2001 to 2021Burned US$187M

Hujiang

The promise

Online language courses for a very large free audience.

Why it died

Most users never paid, and those who paid did not cover the cost of acquiring them. A failed stock listing exposed it.

Unit EconomicsLoot Drop entry ↗
Died 2015 to 2019Burned US$50K

Teacher Finder

The promise

Match language learners directly with private tutors, without a language school.

Why it died

Larger tutor marketplaces offered more, and winning each customer became too expensive.

CompetitionLoot Drop entry ↗

What changed

All five sold what Verbi sells: practice in speaking a language with someone who answers. In all five that someone was a human teacher, paid by the hour, and the students had to be found by a sales team or by expensive advertising. The database records two ways they ended: the cost of winning and teaching a student was higher than what the student paid, or a regulator removed the teachers.

Verbi replaces the teacher with AI models that it rents, which makes a yearly price of US$47.99 possible and let a very small team build and run it. It replaces the sales team with short videos that cost almost nothing to produce. What has not changed is the question that ended Hujiang and Meten: whether a learner stays long enough to repay the cost of winning that learner. Verbi's cost per new subscriber is not published, its margin is a founder entry, and its first yearly subscribers reach their renewal date in April 2027.

05

Can one person copy it now?

The method, yes. The September result, no.

The parts can be repeated by one or two people with little money: a narrow product on rented AI models, a yearly plan, and a very high volume of short videos tested on many accounts. The result cannot be ordered. The best month depended on videos that spread, and daily revenue was 77% lower one month later.

Where it is still open: Speaking practice for one narrow need: a job interview in a second language, an oral exam, the vocabulary of one profession. Verbi's own reviews ask for more reliable speech recognition and for real grammar correction.

  1. 1Release a small first app to learn the store, the paywall and the video platforms, and stop it when the numbers stay low.
  2. 2Choose a product in which an AI model does work that a paid person used to do.
  3. 3Generate many cheap video variants, post them on several accounts, and keep only the accounts and formats that bring trials.
  4. 4Offer a yearly plan, so that the cash arrives before the cost of the next videos.
  5. 5When one format works, pay creators to repeat it, and add a web purchase page to pay less commission.
  6. 6Read revenue month by month next to MRR, and do not multiply the best month by 12.

What to watch

The risks are the mirror image of the causes above.

Half of the revenue came in one month

September 2026 brought US$128,225, which is 51.1% of everything the app has earned. Early September ran at about US$6,312 a day and early October at about US$1,451 a day, 77% lower. TrustMRR marks the last 30 days as 38.3% below the 30 days before. The 30-day figure is one month of sales and says little about the next twelve.

Apple and Google collect the money and keep a share

Apple's developer pages say that a developer receives 70% of a subscription price in the subscriber's first year, or 85% in the Small Business Program. Apple also decides on refunds: Verbi's terms say that the company cannot refund an App Store purchase itself. The TrustMRR profile does not say whether its revenue is before or after this commission, and RevenueCat's documentation defines its standard figure as before it. The web purchase page of July 2026, which the founder describes as selling "without the apple tax", reduces this dependence only for the customers who use it.

Customers come from short-video platforms

The founder's posts credit TikTok and Instagram videos for the sales, and a Google Play review begins with "TikTok brought me here". The method uses generated slideshows and many accounts on many phones; in March 2026 he asked on X whether a device may hold only 3 TikTok accounts. A platform can close accounts or show such videos to fewer people. His posts mention no other source of customers of similar size; on 22 July he wrote that he would soon try paid ads for the web version. Paid creators and promoted posts are a cost that the first months did not have: he wrote "80% margins" in July, and the profile shows 55%.

No yearly subscriber has renewed yet

The first yearly plans were sold in April 2026, so the first renewals fall due in April 2027. RevenueCat's State of Subscription Apps 2025 report says that nearly 30% of annual subscriptions are cancelled in the first month, that most apps with cheap annual plans keep up to 36.0% of users after a year, and that on weekly plans fewer than 5% of payers reach month six. RevenueCat's count of active subscriptions includes people who have already cancelled and are waiting for the paid period to end.

The AI providers are not named

The privacy policy says that the learner's audio is streamed to third-party AI providers and names none. If they charge by use, as model providers usually do, a subscriber who practices a lot costs more than one who stops, and a change of price or policy by a provider changes the margin. Reviews in the US App Store in July and October 2026 report speech recognition that rejects correct answers, a limit of 3 minutes on a voice session, and mistakes in generated exercises.

Team, costs and the founder's time are not verified

The founder's X profile lists a job as a software engineer at Twitch and a YouTube channel next to Verbi. The team size of 2 to 5, the bootstrapped status and the 55% margin are fields that he filled in. TrustMRR verifies revenue, MRR and subscriptions only, and its FAQ says that its rebuilt figures can differ from a provider's own by up to 30%. Two reviews in the Canadian App Store, answered by the developer on 9 September 2026, say that a rebuilt version is harder to use.

Sources

The grade next to each source is the best grade that the source can support. "Read in full" means that the text of the page was downloaded and the numbers were checked in it.

  1. A TrustMRR profile of Verbi, revenue read from RevenueCat by API key
    Read the markdown profile (https://trustmrr.com/startup/verbi.md) live on 13 October 2026, and the cached copy of 12 October. Revenue, MRR and active subscriptions come from the RevenueCat API, last synced on 13 October 2026 at 17:59 UTC. The description, team size, funding, technology list and profit margin are entered by the founder, grade C.
  2. A TrustMRR on X, 28 August 2026: Verbi just joined TrustMRR
    Read in full through the fxtwitter API, as a repost in the founder's timeline. Source for the figures on the day of listing.
  3. C TrustMRR FAQ, read 13 October 2026
    Read the sections on payment providers, on the RevenueCat and App Store connections, and on why figures can differ from a provider's own.
  4. C RevenueCat documentation: Overview Metrics, and the Revenue chart page
    Read both pages in full. Source for the definitions of Revenue (gross, before store fees and taxes), Proceeds, Active Subscriptions, Active Trials and New Customers. The Revenue chart page is at /docs/dashboard-and-metrics/charts/revenue-chart.
  5. C RevenueCat documentation: Monthly Recurring Revenue (MRR) chart
    Read in full. Source for the conversion of weekly and yearly prices to a monthly value.
  6. B Apple App Store listing, United States: Verbi: Speak a New Language, read 13 October 2026
    Read in full, with Apple's public lookup API for the app and for the developer account. Source for the description, the seller name, the rating, the in-app purchase prices, the release dates of Verbi and Lucidia, the version history and three user reviews.
  7. B Apple App Store listing, Canada: Verbi: Speak a New Language, read 13 October 2026
    Read the rating, the in-app purchase prices in Canadian dollars and the reviews, with Apple's lookup API for the Canadian storefront.
  8. B Google Play listing: Verbi: Learn a New Language, read 13 October 2026
    Read the header, the description and the first reviews. Source for the stars, the review count, the downloads, the date of the last update and the review that mentions TikTok and a 3-day trial.
  9. C Verbi website, with its comparison page, read 13 October 2026
    Read the home page and the comparison page in full. Company's own text. Source for the product description and the link to the web app.
  10. C Verbi privacy policy and terms, read 13 October 2026
    Read the sections on the operating company, AI and voice processing, data sharing, and in the terms (https://appverbi.com/terms) the sections on subscriptions and refunds.
  11. C Pedro Machado on X: profile and timeline, read 13 October 2026
    Read the profile and four pages of the timeline through the fxtwitter API, 74 posts back to April 2025, and four attached screenshots by eye. Source for his profile text and for the posts of 9, 10, 11 and 27 March (first app), 16, 19 and 21 May, 2, 4 and 21 June, 5, 12 and 16 July and 10 September 2026.
  12. C Pedro Machado on X, 23 May 2026: launched on April 2nd, downloads, conversion and MRR
    Read in full through the fxtwitter API. Founder's own figures.
  13. C Pedro Machado on X, 13 May 2026: 100 active subscriptions and the slideshow tool
    Read in full through the fxtwitter API. Founder's own figures.
  14. C Pedro Machado on X, 3 June 2026: two people posting on 12 accounts
    Read in full through the fxtwitter API.
  15. C Pedro Machado on X, 9 June 2026: 4.3k in 28 days with no ad spend
    Read in full through the fxtwitter API. Founder's own figures.
  16. C Pedro Machado on X, 21 July 2026: 45k in the last 28 days and 10 creators
    Read in full through the fxtwitter API, and the attached RevenueCat screenshot by eye.
  17. C Pedro Machado on X, 22 July 2026: web app and web purchase page
    Read in full through the fxtwitter API.
  18. C Pedro Machado on X, 3 September 2026: a 16k day, quoting his post of 2 September on 10k MRR
    Read both posts in full through the fxtwitter API, and the two attached RevenueCat screenshots by eye. The post of 2 September is https://x.com/pedrotech_/status/2095173251957948756.
  19. C YouTube channel PedroTech, read 13 October 2026
    Read the channel name, description, subscriber count and the link to his X account in the page data.
  20. B Apple Developer: Auto-renewable subscriptions, and the App Store Small Business Program, read 13 October 2026
    Read the section on net revenue on the first page and the rules of the program at https://developer.apple.com/app-store/small-business-program/. Source for the 70% and 85% shares and the US$1 million limit.
  21. C RevenueCat: State of Subscription Apps 2025
    Read the summary figures and the retention section. A vendor's benchmark from its own customers' data, not about Verbi.
  22. C Loot Drop database export, 1,749 failed startups
    Local export. AI-written summaries, not independently verified. Source for the failed predecessors.

More in Learning and study

Other businesses in the same group, best evidence first. All 8 stories in the group.

AI writing add-in for Microsoft Word · Germany

fastwrite.io

Evidence A
US$183,602still running, revenue falling · 17 months after launch
The product

An add-in for Microsoft Word, plus a web app, for German-speaking students who write seminar papers and theses. The student uploads the PDFs of the literature. While the student writes, fastwrite suggests the next sentence or a whole paragraph with a citation from those PDFs, and the Tab key accepts it. It also answers questions across all uploaded PDFs with the exact page, searches more than 20 million papers, and checks a draft against nine criteria that universities use for grading.

Why it worked

It puts the AI where German students already write, inside Word and in German, and each suggestion carries a citation from the student's own PDFs, so the student does not have to fear invented sources. A free plan without a time limit brings students in, and a monthly plan sold for the months in which they write turns some of them into paying customers.

Evidence strength3/4
Copyable pattern2/4
Capital needed1/4
Launch 8 May 2025Status October 2026Channel Inside another platform Read the story
AI study app for students · USA

ChatEDU

Evidence A
US$120,117still running, revenue falling · 22 months after launch
The product

An AI study app. A student uploads class materials, such as PDFs, lecture slides, notes, images or YouTube videos, and gets answers with citations to those files, practice questions, flashcards, study guides, audio summaries and a topic-by-topic progress view. It runs on the web and, since May 2026, as an iOS app.

Why it worked

It starts from the student's own course files instead of general knowledge, so answers and practice questions point back to the material that the class uses. A free plan with monthly limits lets students try it, and Pro is sold by the week or by the month. Stripe shows about US$13,200 a month at the end of 2025; the seller says growth came with little to no marketing, and no source explains how students found it.

Evidence strength3/4
Copyable pattern2/4
Capital needed1/4
Launch 12 December 2024Status October 2026Channel Not stated Read the story
AI note-taker and study app for students · USA

Coconote

Evidence C
US$7Msold · 22 months after launch
The product

A phone and web app that records a lecture, or imports audio, video, a PDF, a document or a link, and makes a transcript, organized notes, flashcards, quizzes, a chat with the note and a short AI-narrated podcast of it. A student gets one free note; after that the app needs a subscription, which starts with a free trial.

Why it worked

Two former Loom staff built a premium-priced app for something students do several times a week, going to class, and sold it through short videos by a small group of contracted creators instead of paid ads. It charged from the first day, and revenue paid for growth with about 50% EBITDA margins until Quizlet bought it.

Evidence strength1/4
Copyable pattern2/4
Capital needed2/4
Launch April 2024Exit February 2026Channel Creators and short videos Read the story

Built with the same tools

Businesses whose sources name tools that Verbi also uses: Next.js, React Native, Cursor, RevenueCat. Every tool has its own page in Tools.

AI audio mastering from text prompts · Turkey · Funded

SoundBoost.ai

Evidence A
US$217,705still running, revenue growing · 32 months after launch
The product

A website and phone app that masters a song from a written request such as "warm, powerful and wide stereo". Each request comes back as a new revision that the user can compare with the earlier ones, and a Spotify link can serve as the reference for loudness and tone. The same subscription includes a splitter that separates a song into vocals, drums, bass and instruments. Since September 2026 the company also offers a generator that makes a music video from the song. Unlimited use costs US$16 a month or US$48 a year on the web.

Why it worked

For its first 20 months the product brought in about US$3,500 a month. Then three things came together: phone apps with weekly plans from August 2025, web plans with unlimited use at US$4 a month billed yearly, and a new group of customers, people who make songs with AI generators such as Suno and need the last step done for them. Monthly revenue rose 7.2 times between January and July 2026. No source states which of the changes caused the rise.

Evidence strength3/4
Copyable pattern2/4
Capital needed2/4
Launch 8 February 2024Status October 2026Channel Creators and short videos Read the story
Open-source social media scheduler for people and AI agents · Israel, company registered in Hong Kong

Postiz

Evidence A
US$3Mstill running, revenue growing · 27 months after launch
The product

A calendar that publishes posts to more than 30 social networks: X, LinkedIn, Instagram, TikTok, YouTube, Reddit, Bluesky, Discord and others. The whole code is open source, so anyone can run it on their own server for free; the company sells the hosted version from US$29 a month. Since 2025 AI agents such as Claude, ChatGPT and OpenClaw can connect to it through an MCP server, a command-line tool and a public API, so an agent drafts and schedules the posts and a person approves them in the calendar.

Why it worked

The founder took a twenty-year-old product category with no open-source entry and open-sourced all of it, which gave him millions of Docker downloads and 37,000 GitHub stars without a marketing budget. Revenue stayed small for a year. It multiplied when he turned the product toward people who automate their posting with n8n workflows and, from late 2025, with AI agents, which need an API-first scheduler and keep paying because they post without human effort.

Evidence strength3/4
Copyable pattern2/4
Capital needed1/4
Launch July 2024Status October 2026Channel Open source Read the story
AI calorie tracker app that estimates the nutrition of a meal from a photo · France

Calyra

Evidence A
US$506,567still running, revenue falling · 14 months after launch
The product

An iPhone and Android app that estimates calories, macronutrients and micronutrients from a photo of a meal, with a barcode scanner for packaged food and a log of weight and BMI. It is free to download; the service needs a subscription after a 3-day free trial, at US$9.99 a month or US$49.99 a year in the US store.

Why it worked

One founder released it in August 2025, after Cal AI, released in April 2024, had shown that photo calorie counting sells. It has a hard paywall, a short trial and cheap yearly plans. It appears to have started in France: the French App Store shows 598 ratings against 138 in the US store. The owner lists TikTok as the only marketing channel.

Evidence strength3/4
Copyable pattern2/4
Capital needed2/4
Launch 12 August 2025Status October 2026Channel Creators and short videos Read the story

Same pattern: Phone apps sold through app stores

These are mainly iPhone and Android apps. Apple or Google collects the payment in the app and keeps a commission. All of them sell subscriptions. All 16 stories with this pattern.

AI meeting and call note-taker app · USA

Wave AI

Evidence A
US$4Mstill running, revenue flat · 16 months after launch
The product

An app for iPhone, Android, Mac, Windows and the web that records meetings, phone calls and lectures, writes a transcript with the speakers identified, and then a summary with action items. A user can also ask questions about a recording. It is free for 30 minutes a month; Pro costs US$100 a year in the app stores.

Why it worked

It reached the App Store early, in July 2023, with the simplest form of the idea: press one button, get a transcript and a summary. The founder paid for Meta ads while they still worked for a new kind of app, tested paywalls constantly and moved buyers to annual plans. He works alone with AI coding tools, so in his own 2026 account about US$3 million of about US$7 million a year stays as profit.

Evidence strength3/4
Copyable pattern2/4
Capital needed2/4
Launch 26 July 2023Status November 2024Channel Paid ads Read the story
Voice notes and dictation app · Spain

Letterly

Evidence A
US$3Mstill running, revenue flat · 29 months after launch
The product

An app for iPhone, Android, Mac, Windows and the web that records speech and returns finished text: a clean note, an email, a list, a social post or a meeting summary split by speaker. A phone keyboard and a desktop dictation tool let users speak into any other app. It handles more than 90 languages and recordings of up to 90 minutes.

Why it worked

The first apps that combined OpenAI's Whisper and ChatGPT for voice notes appeared in 2023, and the founder found them "not implemented well enough". His team built a simpler one, collected early users and feature requests through lifetime deals on AppSumo in 2024, and in 2025 grew with advertising and paywall tests measured in Adapty's subscription data.

Evidence strength3/4
Copyable pattern2/4
Capital needed2/4
Launch November 2023Status April 2026Channel Paid ads Read the story
AI makeup try-on app · Canada

GlowUp

Evidence A
US$1Msold · 5 months after launch
The product

An iPhone app for makeup. The user uploads a selfie without makeup and a photo of a makeup look, for example from TikTok or Instagram. The app's AI applies that look to the user's own face, then gives a step-by-step routine and a list of products to buy. It is free to download and paid for with subscriptions through the App Store.

Why it worked

Its result is a before-and-after picture of the user's own face, which a short video shows in seconds. The founder chose TikTok as the channel before he built the app, then tripled revenue from almost the same installs by moving and testing the paywall with Adapty.

Evidence strength3/4
Copyable pattern2/4
Capital needed1/4
Launch 15 October 2024Exit March 2025Channel Creators and short videos Read the story

Back to all stories