codeset

codeset is an AI learning platform built around complete courses, individual tutoring, and work that the learner actually wants to finish. Instead of opening with a blank chat box, it starts with a catalog of structured syllabuses. A learner can study C by writing a shell, learn Rust through a key value store, refresh Python, or take a course with no programming at all. The central idea is that the syllabus decides what must be learned while the learner's chosen project shapes how each lesson presents it. That gives someone a defined path through a subject without making every student complete the same generic exercise. It also tackles a common weakness of conversational tutoring, where progress can become a chain of disconnected questions with no assurance that important foundations were covered.

The starting flow is deliberately small. Someone finds a subject, reads its entire syllabus without creating an account, and says what they want to build or why they want to learn it. For coding courses, codeset generates a project and opens a browser workspace with starter files, an editor, a terminal, and the runtime ready for lesson one. There is nothing to install. For courses without code, the workspace drops away and the tutor teaches the same ordered curriculum against the problems the learner brought. The catalog currently shows courses across programming languages, machine learning, Kubernetes, and other topics. When a subject isn't present, the learner can describe it and have the platform generate a course, module by module, using credits from the same balance used for lessons.

Each concept follows a recurring lesson, application, and test pattern. The tutor explains the idea, has the learner apply it inside the ongoing project, and then gives a short check that must be completed without the tutor. Progress stars come from that unaided test rather than from opening or scrolling through material. On coding work, the tutor can read the files and the output produced in the workspace, so its response is based on the actual state of the project rather than only on a learner's description of a problem. When someone gets stuck, codeset is tuned to identify and teach the missing concept instead of immediately pasting a finished fix. It keeps the course position and project context as well, letting a learner return later without reconstructing the whole session.

This approach fits self-directed learners who want more structure than a general chatbot but more personal relevance than a fixed video course. It is particularly well suited to people learning programming by building, since the browser workspace connects explanations directly to files that keep evolving. It can also serve learners returning to a language after a break, developers moving into a new technical area, and people studying a non-coding subject with a concrete goal in mind. Schools, bootcamps, and teams have a Campus option with courses built around their syllabus, instructor visibility into cohort progress, invoicing, and purchase orders. The platform is still early, so it makes more sense for learners comfortable trying a young product than for anyone who needs a mature institutional learning system with years of third-party material.

What distinguishes codeset is the combination of a shared full syllabus and a personally generated project. Personalization does not mean quietly dropping difficult sections. The site says every required skill is placed into the learner's work where it fits, with a short bridging exercise when it does not. At the same time, the tutor is not limited to the text pasted into a conversation because it can inspect the work itself. That creates a clearer definition of completion. The learner should have worked through the ordered syllabus, applied its ideas, and passed checks alone. The public catalog supports scrutiny before sign-up because every module and lesson can be read in order. That openness is useful for judging course depth, especially while individual courses can still vary in how far they go.

Access is freemium and credit based, with no card required and no trial clock on the free allocation. At the time checked, the dedicated pricing page listed 300 one-time free credits, while the homepage FAQ said every account starts with 2,000 free credits. Because those public pages disagree, readers should confirm the current sign-up allowance rather than assume either figure. Paid monthly plans start with Learner at $14 for 5,000 credits, followed by Scholar at $29 for 11,000 and Fellow at $99 for 38,000. Monthly credits can roll over for up to two months. One-time top-ups are also available, never expire, and cost more per credit than a subscription. Typical usage is shown before purchase, including about 30 credits for a text lesson, about 60 when read aloud, and about 115 to generate a course. If a balance runs out, tutoring and the workspace pause, but courses, progress, and lesson position remain. The free allocation is still a useful way to test the method before deciding whether ongoing study justifies a plan.