A tutor’s method,for everyone.

Bloom (bloomtutor.app) is a private AI tutor that runs in your browser. Tell it what you want to learn, or drop in your own materials, and it writes an outline of things to master, then teaches one lesson at a time. What you highlight, what you ask and the feedback you write decide how the next lesson is written. There is nothing to install.

Anything that can be taught in writing: a new field such as programming, machine learning, economics or psychology; a textbook or a research paper; a long report, a codebase or an article online. People use it to get a structured start in an unfamiliar subject, turn a book into a step-by-step course, understand documents and code they have to take over at work, and learn a topic properly before an exam or an interview.

A chat answers whatever you type, and earlier context drifts away as the conversation grows. Bloom keeps a plan: an outline of things to master, lessons in order, your questions saved on the passages they are about, and your feedback carried into the next lesson. Each new lesson opens by reviewing your answers to the last one’s questions, then covers what is still unclear. When the outline is done, you get an assessment and a course summary.

Benjamin Bloom’s 1984 finding, published as “The 2 Sigma Problem” in Educational Researcher, was that students tutored one-on-one scored about two standard deviations above students in conventional classrooms, roughly the 98th percentile of the classroom group. Bloom applies three practices behind that result to its outlines, lessons and exercises: attention to where each learner is, adapting to feedback, and mastery learning, where you move on once a point is understood.

Start by writing your learning goal; if it is too vague, Bloom asks you one or two questions first. It then drafts an outline that you can edit directly or ask it to rewrite, and once you confirm, it writes lesson 1. Each lesson ends with 2–3 reflection questions. Write your answers and feedback, tell Bloom you have finished reading, and the next lesson is written from them. When every point on the outline is covered, Bloom writes an assessment and then a course summary.

Select any sentence you don’t understand, in a lesson, an uploaded PDF, a Markdown or code file, or a saved web article, and ask about it. The answer streams in beside the passage, you can keep asking follow-ups, and the whole thread stays attached to that sentence so you can reopen it later. In a topic course, your questions also shape the next lesson. If you only want to keep a passage, you can mark it without asking.

Yes. When you create a course, choose a learning depth: Simple (about 8–10 key points, just the main path), Standard (about 10–12, covering concepts, reasoning and practice) or In-depth (about 12–15, with principles, edge cases and counterexamples). Under Teaching style in your account, pick how Bloom explains: Step by step, Guided questions, Everyday analogies, Learning by doing, Straight to the point, Careful reasoning, or your own instructions. Feedback after a lesson can apply to the next lesson only or to all following lessons.

Learn next suggests three topics based on everything you have studied: one to build foundations, one to make connections and one to explore further. Save them for later or start one right away. The Knowledge map places each of your courses on one shared knowledge tree, where courses meet at their common branches. In My notes you can search every highlight and answer, and export them as one Markdown file.

A course from a topic or your materials takes PDF, TXT and Markdown files, and you can also paste up to 50,000 characters of notes. Project files read many more formats: PDFs display as the original with zoom; Word (including equations), PowerPoint, Excel, CSV, EPUB (split into chapters), Jupyter notebooks and saved web pages become readable pages with their images; Markdown keeps its formatting, code gets syntax highlighting and ZIP archives are unpacked. Legacy .doc and .ppt files, and WPS, Apple and OpenDocument formats, are skipped with a note on how to re-save them. You can add up to 50 files at a time, 20 MB per file and 50 MB in total, and PDFs of up to 500 pages.

A topic course is a full course: Bloom writes the outline and lessons around your goal, and any materials you attach are distilled into key points it teaches from, with sources noted in the outline. Project files have no outline and no generated lessons: each file becomes its own page to read as-is and ask about, which suits going through a document page by page or getting to know a codebase. A web article turns a link to a public article into a clean page with its text and images. Importing project files and web articles is free; only AI questions are billed.

Try PDF reading (experimental). Upload up to 5 PDFs with 100 pages in total, and Bloom has a vision model read each page and turn it into reflowable text, with formulas, tables and figures kept in their original order, so it is easy to read on a phone and to highlight precisely. Each PDF becomes one lesson you can ask about at any time, and the original is kept for comparison. Conversion is charged by actual usage; a 15-page paper usually takes about half a minute, and formulas, tables or figures are occasionally misread.

Yes, in two ways. A share link sends one lesson or the whole course as it is at that moment, and anyone can read it without signing in. You choose whether to include your highlights and answers; your uploaded originals stay private unless you tick them, and you can revoke the link at any time. A reader who wants to continue can copy it into their own account for free. Learn together lets several people share one live course: anyone with the link and the 6-character share code can join, everyone sees each other’s highlights, questions and feedback, and whoever takes an action pays for it.

New accounts get free credits with no card required, so you can start learning before deciding anything. After that you can subscribe monthly for the best price per credit (cancel anytime) or buy one-time credit packs that never renew. Credits are settled by actual usage, and importing project files or web articles costs nothing. The Pricing page lists current prices, how many credits each pack holds and how long they stay valid.

Creating a course from a topic or your materials, writing the next lesson, asking about a passage you select and refreshing suggestions cost credits. Bloom reserves credits first and charges the actual usage when it finishes; every charge is listed in Credits & billing. If a generation fails on our side, the credits are returned. You buy credits on the Pricing page, as packs or a monthly plan; it also lists what new accounts receive and how long credits stay valid.

Your courses, notes and uploads are stored in your account so you can come back to them. To write lessons and answers, the relevant text is sent to an AI model provider. We don’t sell your personal data. The Privacy Policy has the details.

There is nothing to install: Bloom runs in the browser and adapts to computers, tablets and phones, and on a touch screen you long-press to select text and ask about it. The interface comes in English and Chinese, with a light, dark or system theme. You sign in with a Google or GitHub account.

There is an open-source edition. The Li-Evan/Bloom repository on GitHub is MIT-licensed and contains a set of Claude Code skills, for learning with Claude Code in the terminal with lessons saved as Markdown files, plus a web app you can host yourself. bloomtutor.app is the hosted version: no models or servers to set up, just sign in and start learning.