Keet review 2026: an honest early verdict on the YC-backed app that generates AI video courses with games — what it does well, its limits, and who it fits.
Keet is one of the most thoughtful takes on AI-generated learning I have seen: a topic in, a structured roadmap of short explainer videos and Duolingo-style games out, rendered cheaply with Remotion and Manim instead of pricey AI video models. It is early — iOS private beta, unannounced pricing — and it is strictly a personal learning app, so treat this as a promising first-look verdict, not a mature-product rating. If you want to learn on your phone, it is worth the waitlist.
Keet is a Y Combinator (S24) app that generates a full video course on any topic you type — short explainer videos organized into a roadmap, with interactive games to make the material stick. It launched publicly on Hacker News on August 11, 2026, built by Cornell CS grads Zack Ashen and Tommy Tsai out of New York City. This review scores Keet for what it is: a consumer learning product.
I run a competing content engine, so here is my bias up front — and because Keet and my product barely overlap, I can be relaxed about it. Kompozy is for publishing to an audience; Keet is for teaching yourself. There is almost no head-to-head, which makes an honest read easier.
Two things to set expectations. First, Keet is early: private beta on iOS via TestFlight, with pricing not yet finalized, so I am scoring the concept and its public launch material rather than a year of hands-on use, and I have kept the numbers conservative because of that. Second, its cleverest decision is under the hood — it renders course videos with code (Remotion and Manim) rather than expensive AI video models, which keeps the visuals clean and the unit economics sane. Everything below reflects Keet's public state as of 2026-08-11; confirm the specifics on Keet's own pages before relying on them.
Keet is an iOS app that turns a topic into a personalized course. You type what you want to learn, optionally adjust the depth, complexity, and narration style, and Keet generates a structured roadmap of bite-sized lessons. Each lesson pairs a short explainer video with interactive games — multiple choice, swipe, order, connections, and custom formats — so you are quizzed on the material rather than just watching it. Example courses on its site span topology, quantum physics, the history of San Francisco, and modern rockets. The videos are not made with AI video models. The founders have said Keet renders them programmatically with Remotion (React-based video generation) and, for math-heavy content, Manim — the animation engine popularized by 3Blue1Brown. Keet is in private beta backed by YC's Summer 2024 batch; early sign-ups get a couple of free courses, after which the plan is a Suno-style monthly subscription that grants credits to generate more. It is a two-person company focused narrowly on personal learning, not content production.
Keet fits the self-directed learner — the person who has a stack of half-watched YouTube videos, saved articles, and ChatGPT threads on a topic and wants that turned into a coherent, sequenced course they can actually finish. It is well suited to curiosity-driven study on the go: a commute, a lunch break, a weekend rabbit hole. It fits poorly for anyone whose goal is to *produce* content — there is no audience, brand, export, or publishing anywhere in the product. Creators, marketers, and educators who want to teach a following will find Keet is simply built for a different person: the student, not the teacher.
| Dimension | Score | Why |
|---|---|---|
| Course generation | 4.0 / 5 | Turns a bare topic into a sequenced, roadmap-style curriculum — the core promise, and it is a strong one. |
| Video explanations | 3.9 / 5 | Remotion/Manim rendering yields clean, consistent explainers cheaply, though the style is code-animated rather than cinematic. |
| Interactive games & retention | 4.1 / 5 | Duolingo-style formats (multiple choice, swipe, order, connections) actively reinforce learning instead of passive watching. |
| Topic breadth | 4.0 / 5 | "Any topic" holds up across the demo courses — math, history, science, engineering — though niche accuracy will vary. |
| Customization | 3.7 / 5 | Depth, complexity, and narration style are adjustable; still a fairly guided experience rather than a fully editable one. |
| Availability & platform | 3.0 / 5 | iOS-only and in private TestFlight beta as of 2026-08-11 — access is gated by a waitlist. |
| Pricing & value | 3.4 / 5 | Two free courses then a planned credit subscription; fair in concept, but unannounced pricing makes value hard to judge. |
| Learning structure | 4.2 / 5 | The roadmap-over-time framing is the app's best idea — it treats learning as a curriculum, not a one-off answer. |
Keet's pricing is not finalized, which is the honest headline for anyone evaluating value in August 2026. Early sign-ups get two free courses, and the founders have described the plan as a Suno-style monthly subscription that grants credits you spend generating courses. That is a sensible model for a generation product — you pay for what you make — and the Remotion/Manim rendering approach should keep Keet's own costs low enough to price it reasonably.
Without published tiers, though, value is a judgment call. The free courses are enough to decide whether the format clicks for how you learn. If it does, the question becomes how many courses you generate per month and what a credit costs — numbers Keet has not shared. For a casual learner exploring a topic or two, a credit model is likely fine; for someone spinning up many courses, watch how it adds up once pricing lands.
One framing point: Keet is not competing on price with content tools, because it is not a content tool. Comparing its planned learning subscription to a creator platform's pricing is apples to oranges — you are buying personal education, not audience reach.
| Use case | Fit | Why |
|---|---|---|
| Learning a new topic yourself | Strong | This is exactly what Keet is built for — a structured, interactive course generated on demand. |
| Retaining what you study | Strong | The Duolingo-style games are designed to reinforce material, not just present it. |
| On-the-go, phone-first study | Strong | The iOS app and bite-sized roadmap format suit commutes and short sessions. |
| Making educational content for an audience | Weak | Keet has no publishing, export, or audience concept — its videos live inside the app for the learner. |
| Building a branded persona or talking-head video | Weak | There is no persona, avatar, or brand-voice layer; that is not what Keet does. |
| Team or enterprise training rollout | Weak | As a private-beta consumer iOS app, it lacks team, admin, or distribution features. |
In fairness to both products, Kompozy and Keet do not really compete — and pretending otherwise would be the kind of inflated take this review is trying to avoid. Keet is a personal learning app: its user is the student, and its output is a course you consume. Kompozy is a content generation and multi-platform publishing engine: its user is a creator, and its output is finished, on-brand content published to an audience. The only reader who should weigh one against the other is someone who saw Keet's auto-generated lessons and realized what they actually want is to *make* educational content for a following, not learn privately.
For that reader the distinction is clean. Keet cannot help you reach anyone — there is no export, caption, schedule, or platform in it. Kompozy is the opposite: it turns your expertise into Persona Shorts, carousels, quote cards, blogs, and newsletters governed by a Persona Brief, then schedules and publishes them across the eight social platforms plus blog and email. Use Keet to learn a subject; use a publishing engine to teach one. If you are a learner, stop reading here — Keet is your tool, and this review scores it well for that.
Keet is a Y Combinator (S24) iOS app that generates personalized video courses on any topic — short explainer videos in a structured roadmap, paired with Duolingo-style games to reinforce the material. It is built for learning topics yourself, not for producing content.
For self-directed learning, yes — the roadmap format and interactive games solve the "no structure" problem well, and early sign-ups get free courses to try it. It is early (iOS private beta, unannounced pricing) so temper expectations, and it is not a content-creation tool.
The founders have said Keet renders videos programmatically with Remotion (React-based video) and, for math, Manim — rather than using expensive AI video models. That keeps the explainers clean and cheap to generate.
Pricing is not finalized as of 2026-08-11. Early sign-ups get a couple of free courses, after which the plan is a Suno-style monthly subscription granting credits to generate more courses. Confirm current pricing on Keet's own site.
Not yet. As of 2026-08-11 Keet is in private beta on iOS via TestFlight, so access is via waitlist. It is a two-person company backed by YC's S24 batch.
No. Keet's courses are built for the learner inside its app; there is no publishing, export, or audience feature. If your goal is audience-facing educational content, a publishing engine like Kompozy fits that job instead.
Keet was built by Zack Ashen and Tommy Tsai, Cornell CS graduates based in New York City, as part of Y Combinator's Summer 2024 batch.