How this site works

This ChE 3124 site, built on Prantara, helps you learn numerical methods yourself — in a course where AI agents can write the code but you still have to understand, check and explain it.

The idea

The course has two lanes. Lane 1 (AI-free) — quizzes and exams — certifies that you can do the methods by hand and reason about error and stability. Lane 2 (AI-encouraged, with disclosure) — homework — is where you practise supervising AI-written code. This site trains both:

• Practice generates unlimited fresh problems for 35 hand skills with exact answer keys (Python computes the keys; the AI never does).
• Spot the Flaw (16 exercises) has you review agent-written code, like the Exam 2 supervision problem.
• Your AI thinking partner coaches: it asks what you tried, gives one hint at a time, and won't reveal a practice answer until you've attempted it. Explanations of ideas are always available.
• Explain It and AI-free Quiz are retrieval practice: the most reliable way to make things stick.

AI help, everywhere but the quiz

The AI partner, hints and worked help are on every page of the course — study the way you do homework. The one exception is the AI-free Quiz: no AI partner, no hints, like the exam. You choose, problem by problem: work one without help and it counts toward your on-your-own score; use help and it counts as practice. Your dashboard shows how your practice splits between the two; before an exam, aim for plenty of work on your own.

Two scores: all practice, and on your own

Studies of AI in classrooms keep finding the same thing: scores earned with AI help run well ahead of what students can do without it — and students can't feel the difference. So the site keeps two estimates for every skill. All practice counts everything you do here, discounting work where you used hints or the AI partner. On your own moves only on work done here with no hints or AI partner: practice problems you answer before any hint or AI partner message, AI-free quizzes, and oral checks your instructor has reviewed. The site can't see other apps or people, so this number is only as honest as you make it — the in-class exams are the real test. Your dashboard suggests when to “show it on your own”, and the AI partner gives you less scaffolding as your on-your-own record grows.

Textbook problems: check, compare, defend

Every problem from the textbook's problem sets is under Textbook, with the instructor's worked solutions. Write your attempt (or code) first — the solution unlocks after you have, because comparing your own attempt with a solution is where the learning happens. Your instructor may keep some solutions hidden while they are on a current assignment.

• Check my work reads your attempt next to the instructor's solution and tells you which steps hold up and where the first problem is — then asks one question to move you on. It never gives you the answer.
• Defend it is a short oral check on your own solution (or a Spot-the-Flaw review): four questions, one at a time, answered by voice or typing. Use it as practice before a quiz, or “for the record” to add to your skills record. It's scored by AI and your instructor can review it.

Ask a person

The AI partner panel has an Ask a person button. Your question (and, if you choose, the conversation) goes to the instructor and TAs; their answer shows up on your dashboard. Good answers may be added — without your name — to the course FAQ, which the AI partner also reads.

The learning science behind it

• Retrieval practice and spaced review: recalling beats re-reading; skills come back for review at growing intervals (53 skills tracked).
• Interleaving: mixed sessions combine topics, which feels harder and builds the skill of choosing a method.
• Productive struggle and hint ladders: a little difficulty before help improves retention; help arrives gradually.
• Prediction before execution: predict a cell's output before running it; surprises are where understanding grows.
• Self-explanation: explaining to someone else exposes gaps.
• Calibration: rating your confidence shows you when “sure” really means sure — important when no agent is there to check you.

What is recorded, and who sees it

Your practice answers (and optional shown work), confidence ratings, reading progress, notes you send, lab reviews, work checks, oral checks and AI conversations are stored on the course server. The instructor can see them, to help you and to improve teaching (for example, which misconceptions are common this week); TAs see what you send to “Ask a person”. Class-wide summaries only show themes shared by several students, never quotes. AI messages are processed by OpenRouter, without your name or email. Don't put personal or sensitive information in the chat. Privacy & my data lets you download or delete what's stored about you.

Nothing here is graded automatically. Your course grade comes only from Canvas assignments, quizzes and exams.

Want the reasoning behind the design? It's in the course team's research notes (higher-ed pain points, what other AI teaching assistants learned, and the learning-science evidence) — ask your instructor.

Using AI honestly

Using this AI partner counts as AI use (Lane 2). For homework, include it in your disclosure cell — your AI-use log drafts the summary. Never use any AI tool during a quiz or exam. The AI partner will decline to help with live quiz or exam questions, and can be paused by the instructor during class.

Accuracy

The AI partner is good but not perfect, especially at arithmetic. Check numbers yourself — that habit is the point of the course. If a reply is wrong, press 👎 so the instructor can review it.