Keet's iOS app generates structured courses with videos, quizzes and games Keet, a two-person startup founded by Zack Ashen and Tommy Tsai, is testing an iOS app that generates personalized courses with short explainer videos, quizzes, games, and adjustable curricula, available via Apple TestFlight beta. The founders, who entered Y Combinator's Summer 2024 batch, previously pitched developer infrastructure for AI agents to access websites and authenticated user accounts, but the relationship between that product and the current education app is not established. Zack Ashen https://www.linkedin.com/in/zacharyashen/?ref=runtimewire and Tommy Tsai https://www.linkedin.com/in/tommy-tsai-589062208/?ref=runtimewire are testing Keet https://www.trykeet.com/?ref=runtimewire as an iOS app that generates personalized courses. Their two-person startup has also been publicly associated with APIs that let AI agents access websites and authenticated user accounts, though available materials do not establish the relationship between the products. The New York founders entered Y Combinator's Summer 2024 batch with computer science backgrounds from Cornell University and a developer infrastructure pitch. Ashen had previously worked as a software engineering intern at PayPal, while Tsai held product management roles at Microsoft and Intapp. Their original idea was to give AI agents access to websites and authenticated user accounts through APIs, reducing the need for developers to maintain brittle browser automations. Keet's current consumer product applies a different format to an equally broad ambition: turning almost any subject into a structured course. The app creates short explainer videos, quizzes, games and a curriculum that users can adjust for depth, complexity and narration style. Keet is available through an Apple TestFlight beta https://testflight.apple.com/join/wkWW2enA?ref=runtimewire , where users can join the private-beta waitlist. From agent infrastructure to a learning app Keet's public footprint contains both product descriptions. Y Combinator's company directory https://www.ycombinator.com/companies/keet?ref=runtimewire now describes Keet as a mobile app that creates courses on any topic. The same page retains Keet's earlier launch materials for a product called "APIs for any website" https://www.ycombinator.com/launches/Ljk-keet-apis-for-any-website?ref=runtimewire , which let developers connect customer accounts to services such as LinkedIn, Venmo and Uber. Keet's developer documentation https://docs.trykeet.com/overview/introduction?ref=runtimewire also remains online. It describes integrations that maintain authenticated sessions and let applications act through a user's accounts without requiring a Chrome extension. The available materials do not establish whether Ashen and Tsai have retired that infrastructure, continue to operate it, or reused parts of it for the education product. The consumer beta gives the founders a different setting for product discovery. Ashen and Tsai spent their YC batch building behind-the-scenes plumbing for agent developers. Keet's private beta puts them directly in front of consumers, where usage frequency, course completion and willingness to pay can be observed without waiting for enterprise integration cycles. The technical work may also carry over. Generating an extended course requires gathering information, imposing an order on it and producing several kinds of media and assessment. Those tasks differ from maintaining authenticated web sessions, but both products depend on software that can turn open-ended internet activity into a predictable workflow. Keet is betting on curriculum, not chat Keet presents each generated course as a roadmap rather than an open-ended conversation with a chatbot. A sample topology course includes four modules covering abstract set theory, topological spaces, continuity and properties such as compactness. Individual lessons shown on the site run for about four minutes and are followed by exercises including multiple-choice questions, ordering tasks, swipe interactions and connection games. That structure is the core of Keet's product argument. General-purpose AI assistants can explain topology or rocket engines on demand, but they usually leave the learner responsible for deciding what to ask next. Keet chooses the sequence, divides the subject into lessons and tests recall along the way. The experience resembles a generated version of a fixed learning app, with the course catalog expanding whenever a user enters a new topic. Keet is entering a busy category. YouLearn https://www.youlearn.ai/?ref=runtimewire turns PDFs, videos and recorded lectures into notes, quizzes and conversations with an AI tutor. Coursebox https://www.coursebox.ai/?ref=runtimewire generates courses, videos and assessments for educators and businesses. NerdSip https://nerdsip.com/about/?ref=runtimewire packages generated micro-courses into a gamified consumer app. Keet's narrower bet is that people will want a complete mobile curriculum created from a prompt, even when they do not arrive with source documents or an existing syllabus. That freedom creates Keet's hardest product problem. A course generator must decide which facts to include, which sources deserve trust and which sequence will actually teach the subject. Video, attractive cover art and polished quizzes can make weak material feel authoritative. Keet's long-term value will depend on whether its generated roadmaps are accurate enough for learners to rely on and coherent enough to bring them back after the novelty of the first course wears off. Two free courses, then credits Keet says https://www.trykeet.com/?ref=runtimewire beta users will receive two courses at no charge. Generating additional courses will require credits, though Keet has not published the price of those credits on its product page. The model ties revenue directly to course creation rather than charging solely for access to a library. That approach gives Ashen and Tsai an early test of the central consumer question: whether a learner values a custom curriculum enough to pay before knowing its quality. It also makes generation costs visible. Longer courses with narrated videos and multiple interactive exercises require more computation than a short text answer, placing pressure on Keet to produce material that users consider worth saving and completing. Y Combinator https://www.ycombinator.com/?ref=runtimewire is Keet's only publicly confirmed institutional backer. Keet joined under YC's published $500,000 standard deal https://www.ycombinator.com/blog/ycs-500-000-standard-deal?ref=runtimewire , consisting of $125,000 for 7% and $375,000 through an uncapped safe with most-favored-nation terms. YC lists Keet as an active, two-person company. The private beta gives Ashen and Tsai room to refine a product that asks AI to do far more than answer a question. Keet has to sustain a learner's attention across an entire subject, establish enough trust for the learner to follow its curriculum and persuade that person to generate another course. The test will show whether the founders can make internet-scale information teachable through a structured mobile course.