This course expects you to use an AI assistant. This page says how.
ChatGPT, Claude, Gemini, GitHub Copilot, Codex — in an editor or in a browser tab. Use what you have. UNO’s ChatGPT Edu workspace does not include Codex, so nothing in these assignments requires it; every initial prompt is written for whatever assistant you have.
If your tool cannot run code or write files, you will copy code across, run it yourself, and paste results back. That is a perfectly good way to do this course, and arguably a better one: you will see every error message with your own eyes, which is where a lot of the learning is.
If you want an assistant inside your editor, GitHub Copilot is free for verified students through GitHub Education and works natively in VS Code. Apply early rather than the night before Homework 1 — eligibility is verified, and it is re-checked periodically.
Everything below applies if your assistant lives in a browser tab and cannot touch your files.
You do not need to share the data. Every
assignment’s initial prompt already describes the dataset column by
column — that is why it is written the way it is. An assistant can write
correct code for m5_weekly.csv without ever seeing the
file. If something looks wrong and you want the assistant to see the
data, paste the output of df.head() and
df.dtypes. That is almost always enough.
If you do upload the file, upload it so the assistant can
see it, not so it can run it. The prepared
files are small enough to attach — m5_weekly.csv is 657 KB
— and a chat assistant will happily run an analysis on an uploaded copy
and hand you finished numbers and plots. Those numbers came out of its
environment, not yours.
What is graded is the interpretation of your own output. Your
.qmdhas to render on your machine, and your written answers have to follow from the numbers that render produced. A result you got from a chat sandbox and cannot reproduce locally is worse than a modest result you can.
The file and git rules are yours. Each assignment’s
initial prompt tells the assistant where to put files and how to commit.
A browser chat cannot do either — so it should tell you what to write
and what to commit, and you do it. That costs nothing in marks.
Submitting through the browser needs no git at all: on github.com open
submissions/, choose Add file → Upload
files, and drag your folder in (STUDENT_QUICKSTART.md has all three
routes).
Keep a PROMPT_LOG.md in your submission folder listing
every prompt you sent, in order:
### Prompt 1 — 2026-09-01 14:32
Fit simple exponential smoothing to the CA_1 FOODS series and report the test RMSE.
### Prompt 2 — 2026-09-01 14:51
Now add Holt's linear trend with damping, and compare against the previous fit.If your assistant can write the file, let it. If you are in a browser chat, paste them in as you go — it takes seconds and is far easier than reconstructing them afterwards.
Each assignment states a budget. It exists to make you decide what is worth asking, which is a real skill and the reason it is there at all.
You may go over it. If you do, log the extra prompts and add one line at the end of your log saying where you got stuck. That costs you nothing.
It also helps me: if half the class needs fourteen prompts on a question I budgeted nine for, the question is harder than I intended and I want to know.
What does cost you is a log that does not match the work you submit. Under-reporting to hit a number is the one thing here that is genuinely dishonest.
The code can be AI-assisted. The thinking cannot.
Write these yourself, in your own words:
The assignments’ initial prompts tell your assistant explicitly to explain output to you but not to draft the answers you submit. Keep it that way.
This is not an arbitrary rule. Those sections are what the marks are actually for. A model can produce fluent prose about a forecast in seconds; what it cannot do is know which of your results surprised you, or which number you do not believe. That is the part worth your time, and the part worth mine to read.
If you cannot get something working, say so in your submission and describe what you tried. A clear account of a failure earns marks. A quietly copied number that you cannot explain does not.