Free
0€
No card, no trial clock.
- 2 credits offered at sign-up, once
- Enough to prepare your first course
- Questions and corrections: 0.1 credit each
- Every answer sourced
Know exactly what your exam will ask.
Cortex reads your lecture slides and past exams, finds what keeps coming back, and trains you on it, before the day.
Try Cortex See how it worksYour AI study partner, grounded in your course.
Drop your course in.
Slides, past exams, notes, PDFs. Drag a file or a whole folder. Cortex builds the syllabus from the professor's own lectures.
Sources
Final CS-233· 2024
Final CS-233· 2023
Midterm CS-233· 2024
Final CS-233· 2022
Drop an exam, a lecture, a PDF
Programme
Finals 2019 → 2024
SVM & Marginson the syllabusAsked 8 times
Linear Regression & Least Squareson the syllabusAsked 6 times
Logistic Regression & Linear Classificationon the syllabusAsked 6 times
The professor's syllabus, weighed by what actually fell.
Cortex reads the course outline and breaks it into notions. Then it counts, final after final, how many times each one showed up, so you see the whole programme, and where the points really are.
Turn every mistake into a plan.
Paste a failed exercise, a chat, a screenshot. Cortex extracts the gap, rates how severe it is and keeps it in sight, with a drill one click away.
Weaknesses
Paste a chat, an exercise, or write what tripped you up…
Gradient descent: learning rate
SVM soft margin & C parameter
Role of C reversed: a large C narrows the margin.
Without review
With Cortex
Spaced review
You forget. Cortex reschedules.
Ebbinghaus showed it: without review, what you learn fades within days. Each spaced recall lifts the curve, and Cortex places them where they matter, before the exam.
Every answer points to the source.
No hallucinated theory. Each explanation links to the slide or the past-exam page it comes from.
Search
Lecture 3: Linear Models for Classification
…Gradient descent · while not converged: wk ← wk−1 − η ∇R(wk−1)…
Final 2024, Question 18
…a good choice of step size is key to the performance in gradient-based optimization…
Problem set: Linear classifiers
…Algorithm 1: Gradient Descent, given the step size η, initialize w0, while not converged…
Exams
Train in the real format.
Generate mock exams in your school's format, focus on what you ask for, and drill the weaknesses you log.
Try CortexHow students use Cortex
Before a final
Import the course and the last five finals. Get the list of what to master, sorted by priority.
“Study the right 20 %.”
After a bad exercise
Paste the question or a photo. Cortex explains it, links the theory, and keeps it in your weaknesses.
“Never lose the same points twice.”
The last week
Sit a full mock exam, timed, in the real format. Then re-drill what failed.
“Walk in knowing what's coming.”
Your courses stay yours.
Your files are used to answer you, never to train a model. Delete a course and it's gone.
Two credits to start.
Pro when it counts.
Importing your course files and searching them is free. What Cortex generates runs on credits, from a full mock exam down to a single question.
Pro
14.90€ / month
or 119 € a year, 9.92 € a month.
- Unlimited courses
- 20 credits every month, no rollover
- Questions and corrections: 0.1 credit each
- Extra credits: 10 for 9 €, never expire
2 credits: a full mock exam of up to 8 exercises, or preparing a course. 1 credit: a standard MCQ exam, a targeted exercise, a programme, a detected exam format. 0.1 credit: each question, correction or weakness analysis. Longer exams cost more, and daily limits apply to every account: the terms have the details.
Questions,
answered.
What does it cost?
Nothing to start. Importing your course files and searching them is free. What Cortex generates, like a mock exam, a targeted exercise, a programme or the answer to a question, runs on credits: two are offered when you sign up, and Pro adds twenty every month.
What's a credit?
One credit is one generation: a standard MCQ exam, a targeted exercise, a programme or a detected exam format. A full mock exam, or preparing a new course, costs two. Asking a question or having an answer corrected costs a tenth of a credit. Pro credits reset every month; pack and sign-up credits never expire. See the plans.
Which files can I import?
Lecture slides, past exams, handwritten notes, PDFs, PPTX, DOCX and Markdown, one file at a time or a whole course folder. Scanned pages are read too.
Does it work outside EPFL?
Anywhere. Cortex learns the format from the exams you upload, so any university, any language, any grading style works. EPFL is simply where it was built and tested first.
How is it different from ChatGPT or NotebookLM?
Both are general assistants over documents. Cortex is built for one job: exams. It ranks your syllabus by what past finals actually asked, cites the slide or exam page behind every answer, and drills you in your school's format.
Is my data used for training?
No. Your files are used to answer you, never to train a model. Delete a course and it's gone from our storage.