loading...
Gizmo applies spaced repetition and AI-generated flashcard creation to help students retain what they've learned – targeting the 40–90% knowledge loss that follows most online courses.
ENTRY ANGLES
Companion products for online education (retention tools, homework assistance, adaptive quizzing, skill gap analysis) · Scaffolded homework assistance that guides learners toward solutions rather than providing answers · AI-powered card generation, adaptive scheduling, and personalized difficulty curves
VERTICALS
CAPABILITIES
Product design that creates measurable improvements in learning outcomes, AI/ML for adaptive learning and personalization, Monetization strategy to convert free users to paying customers
The online education market is crowded with platforms for delivering content. Gizmo enters through a different door: it helps people retain what they have learned elsewhere – not by teaching anything itself, but by making the memorization that follows learning more efficient.
The mechanism is spaced repetition with AI scaffolding. Flashcard review is a decades-old technique – one side of the card carries a concept, term, or formula; the other carries the answer – and it is particularly widespread in language learning. What Gizmo adds is AI-generated card creation and algorithmically timed delivery. A user can upload any learning material – a PDF, a PowerPoint, a website, a YouTube video – and the platform generates a flashcard deck from the key content. Cards can be edited or supplemented manually; AI can also suggest definitions and explanations as a user builds cards from scratch. Decks can be shared publicly, restricted to colleagues or classmates, or kept private.
The social layer functions like a basic social network: users follow each other, see each other's new decks in a feed, and receive encouragement when they maintain a review streak. But the core scheduling logic is the product's foundation. Gizmo surfaces the specific decks that need review at the right moment based on the Ebbinghaus forgetting curve – the experimentally derived schedule of repetition intervals needed to anchor something in long-term memory. The user doesn't decide when to review; the platform decides.
Gizmo claims 300,000 users growing at 50% month-over-month, primarily through word of mouth. That traction was enough to close a $3.5M seed round.
The retention problem in online education is underappreciated. A person forgets roughly 40% of what they have just learned within 20 minutes. A week later, without reinforcement, they may retain as little as 10%. This means the quantity of courses completed is a misleading success metric – what matters is the fraction of content that sticks. Gizmo is betting, correctly, that there is a large audience of people who want to improve that fraction without switching learning platforms.
The positioning is what makes it interesting strategically. Gizmo is not competing with Coursera, Duolingo, or any course creator. It is the layer that makes all of them more effective. As the online education market grows, Gizmo's potential audience grows with it – the platform benefits from rivals' success rather than being threatened by it. A [recent review](/review/kogda-neochevidno-konkurentov-menshe) covered Context, which took the same second-derivative approach to the AI chatbot boom: rather than building another chatbot platform, it built debugging and analytics tooling that works on top of any chatbot platform. The pattern – identify the rising tide and serve the boats – is a reliable strategy for avoiding direct competition with well-funded incumbents.
Gizmo's monetization model is equally worth studying. The product is free, but users receive 15 lives per day. Each incorrect answer costs one life. When lives run out, the user can wait 10 minutes for a regeneration or pay for a subscription that removes the limit and unlocks unlimited AI-generated cards. The free plan's AI generation is also capped at 10 new cards per day. Subscription pricing runs $8.80/month or $52.80/year.
The twist: the users who get the most value from the product – those who review cards regularly and learn well – are least likely to exhaust their daily lives and therefore least likely to need a subscription. The users who practice sporadically and forget everything are the ones who hit the wall and face the paywall. It is the inverse of a gym membership model: you pay not for access but for the consequences of not using it consistently.
Gizmo's entry into the education market illustrates a broader product strategy: build the companion product, not the core product. The core product in online education is the course or the platform that hosts it. The companion product is anything that makes the core more effective – retention tools, homework assistance, adaptive quizzing, skill gap analysis.
Sizzle, [covered last month](/review/ii-vzorvjot-obrazovanie-sovsem-s-drugoj-storony), took a related angle: an app that helps high school and college students work through homework problems by guiding them toward the solution rather than handing it over. That mode of assistance – scaffolded rather than substitutive – is distinct from Gizmo's but addresses the same underlying need: learners want to actually understand and remember material, not just move through it.
The AI layer makes companion products in education substantially easier to build. Card generation from arbitrary documents, adaptive scheduling, personalized difficulty curves – all of these were expensive to implement four years ago and are now table-stakes. The constraint has shifted from technology to product design: what does the companion product do that makes the learning experience measurably better, and can that benefit be made visible enough to convert free users into paying ones?
Builders exploring this space should look beyond education as the only domain. The companion-product model applies anywhere an AI tool can attach to a primary process and make it more effective – onboarding, compliance training, technical documentation, sales enablement. The specific entry point matters less than identifying where retention and recall are the bottleneck.