Melo runs across explainx.ai rather than living on one page: the full workspace at /dashboard/learn, inline while reading pathway articles, from skill pages via "Try now with Melo," and in workshops with cohort resources as context. Blog Melo is a desktop-only contextual bar on select blog posts — highlight text to Explain, Quiz, or save a note, with answers streamed in place. Eight modes cover Chat, Teach, ELI5, Quiz, Practice, Explain Back, Flashcards, and Interview. The design bias is pedagogical, not ghostwriting: Quiz and Explain Back make you produce answers graded against a rubric, and responses cite explainx.ai library material rather than generic web summaries.
Quiz and Practice modes use LLM-as-a-judge scoring: constructed responses are judged against a visible rubric before feedback is shown. The research on automated rubric grading is task-dependent — QWK runs from ~0.585 on open-ended Bloom's-taxonomy items up to ~0.94 when rubrics are explicit and prompt engineering is tuned — so Melo is built for formative practice, not high-stakes certification. Showing the rubric upfront and accepting valid equivalent wording sit on the favorable end of that literature.