Study platform · with a tutor inside

AI that makes you a better teacher, not a smaller one

Your students get guided AI support built from your own material: study guides, practice exams, flashcards, grounded answers. Every interaction becomes a signal, so you always know who needs help and what needs reteaching.

For classrooms, and for anyone studying on their own: upload your material, get the same tutor.

Forty seconds of a student's Friday. Her briefing spots a midterm five days out and a concept slipping. She builds a practice exam from four of her own materials, reviews the flashcards her highlights became, asks a question inside the reading, and checks what the course has learned about how she studies.

For students

Everything a student needs, built from what the teacher taught

No jumping to an outside chatbot that doesn't know your syllabus. Study guides, practice exams and flashcards come from your teacher's own material, so the help you get is the class, not a detour from it.

The briefing knows where you stand

Open the app and it tells you what changed: hours in, a concept up eleven points, another one slipping, an exam five days out that you haven't studied for yet. Then it hands you the next move.

"You're at 38% on sign conventions, and the midterm is Wednesday."

Practice exams that show their work

Pick the sources; the exam explains its plan before you start, with a because-clause on every topic.

Entropy sign conventions8 questions
because you got Q3 and Q7 wrong on problem set 4Lecture 07 · p. 4
Carnot efficiency vs. real engines6 questions
because your proficiency fell to 54% after Lesson 07Ch. 12 · p. 210

Being tested on material produces more durable learning than studying it again. Roediger & Karpicke, 2006

Flashcards that come back until they stick

Highlight a passage and turn it into cards on the spot. Cards you miss return in two days; cards you know retire. The deck is drawn from this lesson only, so review time is never spent off-syllabus.

Why can no engine beat the Carnot bound?

you missed this 2 of 3 runsfrom Ch. 12 · p. 210

Practice testing and spaced review are the two highest-utility study techniques in the learning literature. Dunlosky et al., 2013

Ask questions inside the reading

Ask a question inside the slides or the chapter and the answer comes from that material and your notes. When your question contains a misconception, it says so, and points at the line you highlighted.

Why is ΔS positive for the surroundings here?
It isn't, on this slide. What must increase is ΔS for the universe, which is the line you highlighted.
Lecture 07 · slide 4

In Anthropic's skill-formation trial, learners who delegated wholesale to AI scored 17% lower; the ones who asked follow-up questions and requested explanations kept what they learned. Shen & Tamkin, 2026

A profile that studies you back

How you learn best, when your accuracy falls off, which concept keeps resurfacing. It reads all of this from months of your answers, and it shapes what gets taught next.

"Read from how you answer over months, not from a questionnaire."
Children in bright coats leaning over a long pale wall, working.

How it works

Built from your material, and it can prove it

Every guide, question, card and answer comes out of the course's own lectures, readings and handouts, so there's nothing a generic AI could add.

Cited to the source

Slide number, page number, handout. One click and the student is looking at the original, with their own highlight still on it.

Reasoned from your work

Sources carry a because-clause: you highlighted this, you missed that, your score fell here. You can see why it chose them, and change its picks.

Stops where the syllabus stops

It follows the lecture and goes no further. A question about next month's unit gets an honest "we haven't covered that yet."

For teachers

The data your teaching has been missing

Students studying in Ember is what gives you the read on your class. You don't fill in anything extra, and the record they see is the one you see.

The teacher's side of the same Friday. Four files become a lesson with a run of show. One slide gets rewritten by hand. Lesson 07's results show who missed what, and one student's trend line explains itself.

Drop in your files, get a lesson back

Lecture notes, a chapter, last week's returned problem set, a lab rubric. Ember proposes the lesson and a minute-by-minute run of show: they predict the lab result before you show them the measurement.

Slide deck
18 slides from your notes and the reading
Study guide
reading order, weighted to Units 3–5
Practice questions
24, half derivation and half application
Reteach block
carried from Lesson 07, where 11 missed sign conventions

Teachers assemble more than they author: the average teacher runs two core curricula plus five supplemental resources, and 51% teach mostly from self-created materials. RAND, American Instructional Resources Survey

Edit everything. Export anywhere.

Each generated slide keeps its source, and all of them are yours to rewrite. When it's right, export the deck as .pptx and teach it however you already teach.

Who needs help, and with exactly what

Where the class is weak, who is struggling with which concept, and which material isn't working. Every number is something a student did.

Entropy sign conventions11 of 28 missed
38%
Ideal gas law2 of 28 missed
88%
Lecture 05 slides · opened by 6 of 28 · median 41 seconds

Feedback returned after the fact mostly goes unread: 81% of students spend fifteen minutes or less with it. Insight has to land while the next lesson can still change. Our teacher field research, 2026

AI turns the signal into a suggestion. You turn it into a lesson.

For a slipping student it drafts next steps: put her in the reteach block, ask her to justify answers rather than give them. Suggested reteaches, checks and groupings, drawn from what the class actually got wrong. They stay proposals until you accept them; nothing reaches the class without you.

Over time

It does the remembering

A class has twenty-eight students and one teacher. Ember keeps the record for all of them: this term, next term, and the one after.

How you learn best

A worked example, then a hard question. You stay with a problem about three times longer that way than when you start from the definition.

When you learn best

Most evenings around eight. Past forty minutes your accuracy falls off, so sessions now end there.

What you keep re-learning

Sign conventions. It has come back four times since August, and it will keep coming back until it stays.

What comes next

You reason well out loud. Starting next unit, the questions ask you to justify answers, not just give them.

Three robed figures sitting apart on a stone parapet against a clear sky, mid-conversation.

"We become builders by building, and lyre-players by playing the lyre."

Aristotle, Nicomachean Ethics II

Aristotle taught Alexander. One tutor, one student, for years.

That has always been the best education there is, and it has always been rationed. A tutor who remembers everything you have done, and keeps reshaping the teaching around the person you are becoming. Ember is built to hand that to everyone, inside the courses people already take. Each rep a student puts in tells their teacher where they stand.

For classrooms, and for learners on their own

Schools

Per teacher, per year

Courses, lesson building, class insight, per-student records, SSO and rostering. Students included at no extra cost.

Talk to us

Individual learners

Free, with a paid tier

Make a course from your own material and study it with the same tutor: study guides, practice exams, flashcards and progress.

Start free