# Wondering

Aiolum story: https://aiolum.com/story/wondering/ · added 2026-11-05 · updated 2026-11-05

**In short.** Wondering (AI app that turns any topic into a course of three-minute lessons, United States): 12K learners and 100+ paying users, July 2026. Evidence grade E: no revenue figure was checked. Team: 2 founders, Cheng-Wei Hu (CEO) and Angelica Kosasih (CTO); YC profile team size 2, November 2026. Launched March 2026; still running.

AI app that turns any topic into a course of three-minute lessons · United States · AI-native · team of 2 to 10 · still running · group: Learning and study

- Outcome: 12K learners and 100+ paying users, July 2026 (evidence grade of the headline number: E; no revenue figure)
- Launch: March 2026; months from launch to the outcome: 4
- Team: 2 founders, Cheng-Wei Hu (CEO) and Angelica Kosasih (CTO); YC profile team size 2, November 2026
- Outside funding: Y Combinator, Summer 2026 batch (amount not published in the sources read)
- Main factor: Personalized courses on any topic
- Meters, 1 to 4: evidence strength 1, copyable pattern 2, capital needed 1
- Tags: AI-native, Small team, Funded, Running, B2C, Education

Links:

- Visit wondering.app: https://wondering.app/
- App Store: https://apps.apple.com/us/app/id6754904253

**The product.** An iPhone, iPad and web app that turns a topic, a link, a PDF or your own notes into a course of three-minute lessons with diagrams, a video with a voice-over or a podcast version, practice questions, an AI tutor and spaced review. The company's compare page says free accounts have daily limits and a paid plan raises them.

**Why it worked.** It applies the Duolingo habit loop (streaks, XP, short daily lessons) to topics that Duolingo does not teach, and the AI writes a new course for each learner instead of one fixed path for everyone.

Evidence grades: A = disclosed by a public company, a filing or a payment processor, or verified by one; B = two or more independent reports agree, or a journalist confirmed it; C = a founder claim that no one else verified; D = reported, not verified: one founder claim or a third-party estimate with no second source; E = no revenue figure found, so the headline is a non-money fact and the other facts carry their own grades; calc = computed on this page from graded facts.

## Summary

Wondering is a learning app that builds a course on any topic the learner names, from large language models to philosophy. A large language model is an AI program trained on a large amount of text that can write new text, here the lessons and the practice questions. Each lesson takes about three minutes and can be read, watched as a video with a voice-over or heard as a podcast. Old questions come back through spaced repetition, a method that schedules a review just before a learner would forget the answer; the home page names the FSRS algorithm for it.

Two founders run it: Cheng-Wei Hu, who worked on Google's NotebookLM, and Angelica Kosasih, formerly of Meta. They met at Cornell Tech. The company was founded in 2026 and is in Y Combinator's Summer 2026 batch. In their launch post of 30 July 2026 the founders wrote that more than 12,000 people had completed over 42,000 lessons, that 687 had learned in the past week, and that more than 100 were paying.

No revenue figure was found, so this story has evidence grade E. The user counts are the founders' own words. The App Store listing shows the prices and 91 ratings, and the app's public code shows how it is built.

## Pricing

Wondering in-app purchase prices from Apple's US App Store listing, read 5 November 2026 (grade B)

| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Free | 0 | 0 | daily limits |
| Pro | 9.00 | 14.99 | unlimited daily lessons, up to 15 courses a month; US$107.99 a year |
| Max | 16.67 | 29.99 | higher limits than Pro; US$199.99 a year |

The listing names the plans and prices but not every limit. The App Store text says Pro gives unlimited daily lessons, AI chats and podcast episodes for normal personal use and up to 15 course creations a month; the app's code says Max raises the Pro limits. The listing also shows a "Wondering Pro Founding" plan at the annual price. Subscriptions on the web may be billed through Stripe, according to the app's privacy text.

## How much revenue does Wondering make? The numbers, graded

A two-founder YC company with 12,000 learners and more than 100 paying users four months after its first dated App Store version, all counts self-reported

| Fact | Value | Grade | Note | Sources |
|---|---|---|---|---|
| Learners and lessons, July 2026 | more than 12,000 people, over 42,000 lessons completed | C | Founders' YC launch post, 30 July 2026: 'more than 12k people complete over 42k lessons'. | [1] |
| Learners in one week, July 2026 | 687 | C | Same launch post: '687 learning with us this past week'. | [1] |
| Paying users, July 2026 | more than 100 | C | Same launch post: '100+ paying'. No revenue amount is given. | [1] |
| Team | 2 founders | C | YC company page: team size 2, Cheng-Wei Hu (Founder/CEO, previously NotebookLM) and Angelica Kosasih (Founder/CTO, previously Meta), San Francisco. | [2] |
| Founded and funding | 2026; Y Combinator Summer 2026 | C | YC company page. The size of any investment is not stated in the sources read. | [2] |
| Earliest dated App Store version | 1.20.0, 11 March 2026 | B | Version history on the App Store page. The version number suggests earlier test builds; the first written reviews are dated 11 to 16 March 2026, one by a beta user. | [3] |
| App Store ratings, November 2026 | 91 ratings, average 4.7 | B | Apple's US App Store page, read 5 November 2026. Seller: Wondering Labs, Inc. Age rating 4+. | [3] |
| Prices | Pro US$14.99 a month or US$107.99 a year; Max US$29.99 a month or US$199.99 a year | B | In-app purchase list on the App Store page. | [3] |
| Pro and Max per month, billed yearly | about US$9.00 and US$16.67 | calc | Calculated: the App Store annual prices of US$107.99 and US$199.99 divided by 12. | [3] |
| Platforms | web, iPhone and iPad; Android announced | C | Compare page: 'Web and iPhone'. The app's text includes 'Wondering is coming to Android'. | [6, 5] |
| Narration models in the code | Google Gemini text-to-speech and ElevenLabs | B | The web app's JavaScript lists the lesson-video narration models gemini-3.1-flash-tts-preview, gemini-3.8-flash-tts and eleven_v4_turbo. The model that writes the lessons is not named in the files read. | [5] |
| Time from first dated version to the July 2026 counts | 4 months | calc | Calculated: March 2026 to July 2026. | [1, 3] |

How to read this table: a grade C value is the founders' or the company's own statement. A grade B value was read in Apple's App Store listing or in the app's public JavaScript code. A "calc" value was computed on this page. No revenue figure was found.

## When did Wondering launch, and how did it grow? Timeline

From the first dated App Store version to a YC launch with 12,000 learners in about four and a half months. Each entry carries its grade.

- **2026** (grade C [1, 2]): Wondering Labs is founded by Cheng-Wei Hu and Angelica Kosasih, who met at Cornell Tech and write that they have built things together for more than five years.
- **11 March 2026** (grade B [3]): Version 1.20.0 is the earliest dated version on the App Store. The first reviews appear the same week.
- **18 April 2026** (grade B [3]): Version 1.21.0 adds iPad support.
- **22 July 2026** (grade B [3]): Version 1.24.0 adds Course Podcasts: any lesson becomes an audio episode.
- **30 July 2026** (grade C [1]): YC launch post: more than 12,000 learners, over 42,000 lessons, 687 learners that week and more than 100 paying.
- **August to September 2026** (grade B [3]): Friend Streaks (20 August), Live Tutor, a spoken podcast-style lesson (26 August), Canvas (3 September) and Global Review with spaced repetition (8 September).

## Why did Wondering work?

### A course for each learner

The learner gives a topic, a link, a PDF or notes, plus goals and background, and the AI writes a structured path of short lessons for that person. The compare page puts it plainly: if you want a language, use Duolingo; for anything else, use Wondering.

### Habit features copied from games

Daily streaks, friend streaks of up to three friends, XP, streak freezes and an optional calendar block keep learners returning, the same loop that language apps use.

### One lesson, three formats

Each lesson can be read with diagrams, watched as a video with a voice-over or heard as a podcast, and a Live Tutor turns it into a spoken conversation. Speech comes from text-to-speech models, so one written lesson gives three products.

### Free with limits, then a subscription

Free accounts have daily limits. Apple's listing shows Pro at US$14.99 a month and Max at US$29.99 a month, sold as App Store subscriptions.

## How it was built

These details come from the app's public JavaScript code and its App Store listing (grade B), and from the company's own pages (grade C).

- One web code base, packaged as the iPhone and iPad app with Capacitor, a tool that wraps a web app in a native shell.
- Supabase for login (including Sign in with Apple), the database and server functions.
- Narration from Google Gemini text-to-speech models and ElevenLabs, chosen per lesson video.
- RevenueCat and Apple for App Store subscriptions; the privacy text says other subscriptions may go through Stripe.
- PostHog for product analytics and Sentry for error reports.
- The FSRS spaced repetition algorithm schedules reviews across all courses.

Sources: [5, 3, 4]

## Who tried this before Wondering, and why did they fail?

Three companies in our database of failed startups promised personalized or automatically generated learning before large language models existed. The values below are from that database and are not independently verified.

The failed startups come from our database of dead startups, retold here in our own words; Its entries are AI-assisted summaries of public sources, so they carry no evidence grade.

### Knewton (2008 to 2020; funding US$182.3M; cause: Competition)

- Promise: Adaptive learning that personalized each student's path.
- Why it died: Its technology depended on data and integration from schools and was costly to scale, while free content from Khan Academy competed.
- Record: 

### Grockit (2007 to 2015; funding US$44.4M; cause: Competition)

- Promise: Gamified, social test preparation.
- Why it died: Duolingo and Khan Academy adopted similar game-like models, and it was absorbed by a larger test-prep company.
- Record: 

### Qwiki (2010 to 2015; funding US$10M; cause: No Market Need)

- Promise: Automatically generated, narrated multimedia stories about any search topic.
- Why it died: Generating media with a voice-over was expensive, and users did not change from text search.
- Record: 

## What changed

Knewton needed schools' data and costly integration to personalize a course; Qwiki paid heavily to generate media with a voice-over. Both problems were costs of building each personal course by hand or by custom systems.

A language model now writes a course from a topic in seconds, and text-to-speech models read it aloud, so a two-person team can make a personal course for every learner and sell it straight to consumers through the App Store.

## Can one person copy Wondering now?

**Verdict: The format, yes. The traction, not yet shown.**

The parts are standard and rented: a web app wrapped for iPhone, Supabase, a language model and text-to-speech. What a copier cannot rent is a YC launch and two founders with education-product experience at Google and Meta. Paid conversion is the open question: the founders reported more than 100 paying users out of 12,000 learners.

1. Pick one learning habit that works (short daily lessons with streaks) and apply it to a topic area the leaders do not cover.
2. Generate the course per learner from their goal and background, not one fixed path.
3. Turn each written lesson into audio and video with text-to-speech, so one lesson serves three moments of the day.
4. Give a free tier with daily limits and sell the higher limits as an App Store subscription.

**What is still open.** Narrow versions for one audience: AI for one job role (the app already has a page for this), courses from a company's own documents, or courses for a professional exam.

## What could go wrong for Wondering?

- **What is not verified.** No revenue figure was found. The founders' launch post of 30 July 2026 gives only counts: more than 12,000 learners, 687 in one week and more than 100 paying, with no amount. We looked for a payment-processor profile (TrustMRR, Stripe or RevenueCat), a founder interview with a dated figure and a company filing, and found none. A reader could look for a verified revenue profile or a later founder update.
- **Competition.** Duolingo, Brilliant, Coursera and Google's own NotebookLM, which the company compares itself with on its site, are much larger; on this site NerdSip sells a similar AI micro-course phone app and has a payment-verified revenue figure, which Wondering does not.
- **Depends on model providers.** The app's code lists Google Gemini and ElevenLabs text-to-speech models for narration, and its terms say the service relies on third-party providers for "hosting, authentication, AI processing, and other functionality". The model that writes the lessons is not named in the files read. Price changes or limits at these providers would raise the cost of each free learner.
- **App review pre-prompt.** The app's code contains a prompt titled "Thoughts on Wondering?" with the buttons "Love it!", "Meh" and "Not now". A code comment says the feedback form "Share your feedback" opens "when a learner responds Meh to the App Store review prompt", and the same code calls Apple's review request. So learners who answer "Meh" are sent to private feedback rather than to the App Store review sheet, which may lift the public rating of 4.7.
- **Age rating and age terms differ.** The App Store rates the app 4+ and lists "Students who want structured study" among its users, while the privacy text says "Wondering is not intended for children under the age of 13". The app also contains user-generated courses that others can browse, with tools to report courses and block authors.

## Sources

1. Y Combinator Launches, 30 July 2026: Wondering - Duolingo for learning anything — https://www.ycombinator.com/launches/S59-wondering-duolingo-for-learning-anything (grade C). Read in full (the post as embedded in the YC company page data). The founders' own post. Source for the learner, lesson and paying-user counts and the founders' background.
2. Y Combinator company page: Wondering — https://www.ycombinator.com/companies/wondering (grade C). Read in full, 5 November 2026. Source for the founders, titles, team size, year founded, batch and location.
3. Apple App Store: Wondering - Learn Anything (Wondering Labs, Inc) — https://apps.apple.com/us/app/id6754904253 (grade B). Read in full, 5 November 2026. Source for the prices, ratings, age rating, seller, version history and description.
4. Wondering home page — https://wondering.app/ (grade C). Read in full. Source for the product description, lesson formats, streaks and the FSRS algorithm.
5. Wondering web app JavaScript bundle (includes the privacy and terms text) — https://wondering.app/assets/index-DNupRvTz.js (grade B). Searched for named services, model names, the review prompt and the privacy and terms text; not read in full (4.4 MB). The privacy and terms pages load this text by script.
6. Wondering vs Duolingo — https://wondering.app/compare/duolingo (grade C). Read in full. The company's own comparison page; source for the platforms and the free tier.
7. Database of failed startups, local export — /method/#dead (grade C). Local export. AI-written summaries, not independently verified. Source for the failed predecessors.

Contact: aiolum@livyden.com. Run Wondering? Send corrections or a verified figure to aiolum@livyden.com (subject: Correction: Wondering).