Learn Anything w/ AI-Generated Lessons & Podcasts
Learn Anything w/ AI-Generated Lessons & Podcasts
Reasonote is an AI-powered Personalized Learning Platform
We offer engaging, interactive, multimodal educational experiences, which are fully personalized to the user's style, interests, goals, and values. Users can bring-their-own-content (pdf/txt now, image/video/audio soon), or start from a simple search term. In either case, a "Skill Tree" is created for the topic, and the user can engage with one of three fully AI-Generated learning modes: The Classroom, The Podcast, and The Flow.
The Classroom contains an AI tutor which asks students questions about their underlying goals (Reasons) for learning the content, their existing experience with the topic, and any specific areas of interest. The tutor then creates and suggests personalized, interactive gamified lessons, which the user can progress through at the pace they choose. These Duolingo-inspired lessons can range from simple slides, matching, short-answer, and multiple-choice, to complex Roleplay activities which require the user to demonstrate their knowledge in virtual chatroom scenarios.
The Podcast mode consists of AI-Generated two-"person" podcasts, similar to NotebookLM. The advantage of these podcasts over NotebookLM are several: Fast-Generating, Interactive, Easily Shareable/Forkable, & Queueable (Like Spotify).
The Flow mode is designed for those familiar with Anki decks / Spaced-Repetition, and involves an infinitely refreshing list of Activities (similar to those in classroom mode) where students can practice the skills they've opted into learning.
In the long run, our plans are ambitious. We believe Reasonote's feature set will move upstream of education to assist in Goal/Value Ideation, Personal Commitment/Accountability, and "Flourishing Community Development".
SF, CA & Mumbai / SF, CA
~6 months ago we met at a hackathon. Later, we realized we were both working on separate Edtech solutions. We decided to join forces rather than compete, and haven't looked back since.
Without any advertising and through word of mouth, we've accrued 15 weekly active users and 5 paid customers. We believe we have a product which can get us a few hundred paid users without any changes if we can solve distribution. Starting this week we're building content for content marketing, testing some distribution hacks, and moving to ramp up our user acquisition rate.
Luke -- 11months total (PT: 6 Months, FT: 5 months)
Ishan -- 1month total (PT: 1 week, FT: 3 weeks)
- ~300 Users
- ~15 Weekly Active Users
- 5 customers @ $14/month/user ($840ARR)
(Note: $14/month is the normal rate, but some of our early customers have come in with a "founder's discount" for their first period, hence the revenue discrepancy.)
$14/month is the normal rate, but some of our early customers have come in with a "founder's discount" for their first period, hence the revenue discrepancy.
After talking to about 100 existing users, around 50 potential users, and analyzing our posthog recordings, we discovered the following pain points, many of which we have since addressed:
PROBLEM: It took too long to get started.
SOLUTION: We addressed this by radically reducing the Time-To-Value, making the "Get Started" flow about 10-20x faster, while still retaining most of our data collection and personalization features using an AI interviewer.
PROBLEM: The personalized tutor needed to be more obvious, accessible, and helpful.
SOLUTION: We resolved this by creating the "Classroom Mode", which allows you to navigate your skill tree with a helpful assistant who can create new content for you on-demand.
PROBLEM: People were tired of looking at a screen, and wanted different ways to engage with their content on the go.
SOLUTION: To fix this, we created the "Podcast Mode", which is like a faster, more interactive version of NotebookLM. To make this mode more flexible to the user's interests, we added two novel features:
(A) Interruptions, where you can join the podcast to redirect the hosts towards topics of interest,
(B) A Spotify-style "Queue" mechanism, which allows you to queue up topics for learning.
PROBLEM: People wanted to bring their own content into Reasonote easily.
SOLUTION: We addressed this by releasing our Chrome Extension, which currently supports clipping content, and will soon make it easy to begin a Podcast or Lesson from right within your browser, on any webpage.
These are just some of the many lessons we learned from talking to users, and we're just getting started.
Luke went through ODF (zero-equity) shortly after arriving in SF. It was a helpful way to setup a support network of startup founders in the area, and has spawned several nice things, including a founder accountability group, and an AI study group.
Having tutored over 200 students between us in subjects as diverse as software engineering, mathematics, improvisational theatre, and music, we know that effective teaching goes beyond content. A good tutor connects with students by understanding their interests, explaining “why” concepts matter, and adjusting to their learning preferences. This experience is missing in current “one-size-fits-all” online education, but advancements in AI make it possible to offer this kind of deeply personalized “super-tutor” at scale. At the same time, AI has made previously costly educational formats, like podcasts and interactive role-play, significantly more accessible. Reasonote addresses this problem of a lack of scalable, high-quality, and personalized education.
**So, who needs Reasonote?**
As lifelong learners, we ourselves feel this need. For example, Reasonote would have been invaluable for Luke as he learned to build and scale his previous startups, and for Ishan in navigating the fast-paced world of quant trading.
Beyond ourselves, our user interviews revealed:
- Professionals across fields waste hundreds of hours trying to piece together fragmented online resources, outdated courses, and inconsistent guidance.
- Learners in high school and undergraduate programs say their biggest pain point is a lack of personalized guidance to stay motivated and make steady progress toward their goals.
- Those seeking career changes or side projects struggle with motivation, as current resources don’t cater to their individual needs.
All of these learners need Reasonote now to help them achieve their goals faster, stay motivated, and receive the personalized support that online education has yet to deliver.
With Luke’s experience in building and scaling software startups, Ishan’s expertise in data-driven decision-making, and our combined experience as both tutors and lifelong learners, we are uniquely positioned to solve this problem.
Existing Market Solutions:
- Eureka.xyz: Creates AI-generated courses of 3-12 lessons about a user specified topic, but does not offer different modes and is more one-dimensional.
- Quizlet: Primarily offers generative lessons that focus on multiple choice questions and flashcards. It includes some AI tutors, but these are limited in scope and complexity.
- Duolingo: Traditionally a language learning platform, only recently branching into mathematics. Its scope remains limited to these areas.
- Khan Academy: Its AI component, Khanmigo, functions mainly as a chatbot, with a primary focus on user interaction rather than deep educational engagement or content creation.
- Quizizz: Allows users to import content from various sources but primarily supports simpler, self-contained activities like quizzes and tests.
- MagicSchool.AI: Offers a range of educational tools but lacks a focus on end-to-end AI-driven lessons that integrate learning with real-world application.
- Shepherd & PDF2Anki: These tools are geared towards specific niches like content conversion and spaced repetition, without a comprehensive approach to creating AI-driven educational experiences.
------------------------
What Sets Reasonote Apart:
1. Higher Levels of Cognitive Engagement: Bloom’s Taxonomy identifies six levels of cognitive skills: Remember, Understand, Apply, Analyze, Evaluate, and Create. While most educational platforms focus on the first two levels, Reasonote is designed to facilitate deeper cognitive processes. Our activity structure encourages learners to Apply, Analyze, Evaluate, and Create, thereby engaging in higher-order thinking skills that are often neglected by other platforms.
2. Full-Spectrum Course Design via AI: GPT systems have the capability to design complete educational experiences—from courses and lessons to skill trees and interactive simulations. Reasonote leverages these capabilities to offer a comprehensive, AI-driven learning environment that adapts to each learner's needs.
3. Utilizing the Full Potential of AI: The rapid advancements in AI technology, evidenced by developments like the Llama 3 model, suggest that AI will continue to improve significantly. Reasonote embraces this progression by incorporating AI across all aspects of content creation and course design. This strategy not only ensures that our offerings are cutting-edge but also improves over time, ensuring that what we offer today is merely the starting point for even more innovative future capabilities.
SaaS Monthly Subscription model, $14/month for "Basic" plan.
- Assume 60M active users (Duolingo has 37M active users, and Quizlet 60M)
- Assume 10% paid
- Revenue would be $84M/Month, ~$1B/yr
Another possible route would be to also go B2B (on-boarding, continuous education, and employee training on company specific resources and documents)
- Companies spend between $400-$1500 per employee per year on learning and development, with smaller companies spending more on their employees on average.
- In 2023 US companies spent $101.8B on training employees.
- 43% of companies report spending on online learning tools and systems, and 41% on games and simulations, both of which are segments we can capture.
- Assuming we can capture 5% of this market, and that at least 50% ($50.9B) is spent in segments where we offer services, that gives us a potential revenue of $2.5B/yr
- “Analyzers” — easy-to-customize JSON extraction pipelines that you can bring with you to any webpage / document.
- You can create an “Extract People” button, by specifying the “Shape” of a person data object ({name: string, …etc}). Then, there would be a browser / OS plugin which would allow you to run this analyzer on any webpage or document.
- I think this is a great idea, and people would like it and use it, but I have a hunch this particular space will become very crowded.
- This is actually built-in to Reasonote right now, I’ve just got it disabled because I’m struggling to fit it into the EdTech narrative story.
- “Dice Guise” — Basically D&D, but where you can swap out any of the players for an AI (have a few cobbled-together prototypes)
We believe we've found a venture-scale opportunity to dramatically improve lives at scale. Many friends and advisors have encouraged us to apply, including several former YC founders. Luke will be attending the Supabase AI hackathon at YC later this month.
Luke read Hackers and Painters after a dumb luck Amazon recommendation. Some years later, a mentor of Luke's had a hacker news tab open during a pair programming session some 8 years ago, and he's been hooked on YC output ever since.
Comments