The papers you kept meaning to read — now in your ears on the way to work. A source behind every sentence, and a second reviewer who catches what's wrong.
Sign in with Google · No credit card · Free usage every month
In the same study, retrieval augmentation and fine-tuning improved performance but did not eliminate hallucination arXiv:2511.06073
This is a single preprint and comes with the condition "in domains with structured knowledge."
Tap the citation to open the original paper. We state each study's standing and caveats in the text.
Start from a question, a journal, or a field's classics. Wherever you start, the original paper is at the end, and anything you like can be turned into audio.
Ask the way you talk. AI turns it into a search query, finds papers across PubMed, OpenAlex and Korea's KCI, and adds a one-line summary to each. Turn the ones you like into long summaries and audio.
Follow the journals you care about. When a new issue comes out, AI sums up its direction and five papers worth a look. Flip through past issues, gather one topic within a journal, or see six months of trends at once. Missed a busy month? Catch up with your field's radio.
The classic papers behind each field today, collected field by field. We tell the story of what each one found and why it mattered, and add newly found classics every month.
Each article explains one topic through dozens of papers. Every finding and number carries its source, and it's written to be followed by ear. Follow a series and it becomes a textbook for the field.
Not made-up fairy tales but true stories grounded in papers, told at a child's level in short, plain sentences. Every episode comes in both English and Korean, with background notes so parents can answer the questions that follow.
Knowledge is everywhere.
So is talk.
AI, videos, everyone can sound convincing now.
Paperis shares only knowledge that has been checked and can name its source: what researchers have established in their papers, in plain words, for your ears.
Counted automatically from published articles.
AI writes the first draft. Verification is a separate step, and a sentence stays only if it holds up against the original paper.
Every finding, number and conclusion carries a source, and we check it line by line against that paper's abstract. If it doesn't match, the sentence goes.
A separate review hunts for overstatement, missing caveats and cut-off conclusions. Preprints and single studies are labeled as such in the text.
We take corrections after publishing too. Every article ends with its full list of source papers, so anyone can check.
arXiv:2510.22344 — RAG mitigates, but wobbles on multi-step questionsarXiv:2511.06073 — Knowledge-graph verification (RAG and fine-tuning don't remove it)arXiv:2306.08302 — LLMs are black boxes, weak at storing and accessing factsPart of the source list for "Why Do LLMs 'Lie'?"
One article from each field. Read it to the end and open its source list, no sign-in needed. To listen, just sign in.
A general AI subscription runs about $20 a month. Pro gives you sourced knowledge and unlimited audio for about a third of that.
There's plenty of talk. Knowledge that names its source is rare. Start with one on your way to work.