The Y Combinator (S24) app turns a topic into a roadmap of short explainer videos and Duolingo-style games — and renders the video with Remotion and Manim instead of expensive AI video models.
2026-08-11 · by Moe Ameen
Keet launched publicly with a Launch HN post on August 11, 2026: "an app to create video courses on anything." Type a topic and Keet generates a personalized course — a structured roadmap of bite-sized lessons, each pairing a short explainer video with interactive games (multiple choice, swipe, order, connections, and custom formats) to reinforce what you just learned. You can adjust the depth, complexity, and narration style when you create a course. It is built by Zack Ashen and Tommy Tsai, Cornell CS graduates based in New York City, and is part of Y Combinator's Summer 2024 batch.
The most interesting technical detail is how the videos are made. Rather than paying for AI video models, the founders said Keet renders course videos programmatically — with Remotion (React-based video generation) for process animations and primary-source material, and Manim, the animation engine popularized by 3Blue1Brown, for math-heavy content. That keeps the explainers clean and consistent and the unit economics low, which matters for a product that generates video on demand.
Keet is in private beta on iOS via TestFlight. Early sign-ups get a couple of free courses; after that, the founders described a Suno-style monthly subscription that grants credits to generate more courses. Demo courses on the site range across topology, quantum physics, the history of San Francisco, and modern rockets — the pitch being that a personalized, interactive course with a roadmap beats trying to self-teach from a pile of scattered YouTube videos, articles, and chatbot threads.
Keet is worth watching for two reasons, and neither is competitive — [Kompozy](/) publishes to an audience; Keet teaches you privately. The first is the engineering lesson: Keet proves you do not need a costly AI video model to make a clean explainer — code-driven rendering (Remotion, Manim) does it. Kompozy already lives in that world, rendering brand-exact video and graphics deterministically through [HyperFrames](/glossary/hyperframes) rather than leaning on generative video for everything, so the cost-and-consistency argument Keet is making is one creators can act on today.
The second is demand. Keet is a bet that people want structured, interactive learning. If you are the educator, coach, or subject-matter expert who *creates* that kind of content, the move is to publish it where your audience already is — and that is Kompozy's job. Bring your topic in and generate a self-narrated [Persona Short](/glossary/persona-shorts) that explains one concept, a carousel that lays out the steps, a quote card, a blog post, and a newsletter, all governed by your [Persona Brief](/glossary/persona-brief) so they sound like you. Review each piece, then let [Autopilot](/glossary/autopilot) schedule and publish across the eight social platforms plus blog and email. Keet builds a course for the learner; Kompozy builds a teaching presence for the creator.
Keet is a Y Combinator (S24) iOS app that generates personalized video courses on any topic — a structured roadmap of short explainer videos paired with Duolingo-style games. It is built for learning topics yourself, not for producing content for an audience.
The founders said Keet renders videos programmatically with Remotion (React-based video) and, for math, Manim — the engine popularized by 3Blue1Brown — rather than using expensive AI video models. That keeps its explainers clean and cheap to generate.
As of August 11, 2026 Keet is in private beta on iOS via TestFlight. Early sign-ups get a couple of free courses, after which the plan is a Suno-style monthly subscription granting credits to generate more. Pricing is not finalized.
No. Keet's courses are built for the learner inside its app — there is no publishing, export, or audience feature. For audience-facing educational content, a publishing engine like Kompozy is the tool that generates and distributes it across platforms.