This course grades you on posting a prediction before you compute. A tutor that hands you the answer destroys the only part of the exercise that teaches anything, and it is the part you are graded on.
So the tutor has two stages. Before you compute, it is built to refuse the answer — ask it directly and it will tell you no, and ask what you expect instead. After you compute, it opens up completely. Switch stages honestly. Nobody is checking, which is exactly the point: this course is about what you do when nobody is checking.
Some machines are locked down by an employer or a family account. Use claude.ai in a normal browser tab instead — the Open in Claude button above does exactly that and carries the prompt with it. Nothing in this course requires the extension specifically.
It is a reading partner for a chapter written well above Gen Ed level. The assigned Book 7 chapters are at their real technical register on purpose — you are meant to reach for them, not absorb them on a first pass. The tutor is what makes reaching feasible at 11pm on a work night.
It is not a source. It will get things wrong, confidently, in the voice of a person who sounds authoritative. That is not a flaw in this course — it is the syllabus. You are learning to check claims, and a fluent, plausible, occasionally wrong tutor is the most realistic training instrument available. Check what it tells you against the chapter. When it contradicts the chapter, the chapter wins, and that contradiction is worth a discussion post.
Using Claude for this course is expected, not tolerated. What is not permitted is submitting a computation you did not run or a prediction you did not form. The worksheets ask for your working, and the working is what is graded.
A wrong prediction, honestly posted, receives full marks. There is nothing to gain by asking a machine to guess for you, and the grading is built that way deliberately.