Every assignment in a CS course names one of five levels, 0 through 4. Your syllabus sets the default for the course, and an individual assignment can change it. If you are unsure which level applies, ask your instructor before you start.
Every level answers the same two questions
1 Can the AI see your work?
2 Can the AI write your work?
Seeing means the AI has your artifact in front of it: your code, your draft, your design document, or the assignment text itself. Pasting it in, opening it in an AI-enabled editor, and uploading a screenshot all count as showing it.
Writing means the AI produces content that ends up in what you submit, whether you paste it directly or retype it. Explaining a concept is not writing your work. Producing the function is.
The five levels
0
No AI
You build it yourself.
Sees your workNo
Writes your workNo
You can
Use course materials and the web
Use web search, including its AI summary
Use plain autocomplete that fills in a name you already wrote
You can't
Use a chatbot, for anything
Use AI code generation or AI autocomplete in your editor
Rule of thumb: if it generates, it's off.
What we hope you learnTo build the mental model in the first place: how a program actually works, how to express that model as an artifact, and how to check that it is right. This level shows up most often where we are assessing what you understand.
1
Reference
AI helps you learn. It never sees your work.
Sees your workNo
Writes your workNo
You can
Ask for concept explanations
Ask what an error message means
Ask it to quiz you or make practice problems
Ask about instructor-provided examples
You can't
Paste in your code, writing, or the assignment text
Ask for a solution to the problem you were assigned
Sounds likeExplain how recursion works, with an example that isn't a factorial.
Rule of thumb: if your prompt includes text copied from the assignment, it isn't allowed.
What we hope you learnThe concepts, and fluency in expressing an idea as an artifact yourself. A useful test of understanding at this level is whether you can teach the idea back to the AI, not just receive it.
2
Reviewer
AI critiques your work. You still write it.
Sees your workYes
Writes your workNo
You can
Show it your code or draft and ask what's wrong
Ask for debugging guidance
Ask for feedback on structure or readability
You can't
Submit a fix the AI wrote for you
Ask it to rewrite, complete, or generate a solution
Sounds likePoint me at the bug in this function and explain why it fails. Don't rewrite it.
Rule of thumb: if AI produces new solution content, you have exceeded this level. If it rewrites your code anyway, close it and make the fix yourself.
What we hope you learnThe same skills as Level 1, refined faster. Feedback on your own work, at the moment you need it, is a strong accelerant, and it sharpens your model of how code actually executes rather than replacing it.
3
Partner
You build it together. You understand all of it.
Sees your workYes
Writes your workWith you
You can
Have it write code or text you then review
Work back and forth on a design
Debug and refactor together
You can't
Submit anything you can't explain
Sounds likeSuggest an approach for this, then walk me through implementing it and explain each decision.
Rule of thumb: you're co-creating, not delegating. Expect to be asked how any part of it works.
What we hope you learnTo reason about code you did not write: to read it critically, judge whether it is any good, and know how to test it. This is where your existing mental models start doing new work, and where real acceleration begins.
4
Full Use
AI builds it. You specify and verify it.
Sees your workYes
Writes your workYes
You can
Have AI generate all or most of the code
Have AI generate tests, then refine against them
You can't
Submit output you haven't verified
Skip the specification, or hand-wave how you know it works
Sounds likeHere is the specification and the constraints. Generate a solution plus a test suite that shows it meets them.
Rule of thumb: your responsibility shifts from writing the code to defining and verifying correctness, performance, and security.
What we hope you learnTo build and validate real systems: writing a specification worth building against, designing the verification strategy, and using evidence to argue that the system meets it. This is the level that most resembles where the industry is heading.
True at every level
You are responsible for everything you submit, whatever helped you make it.
You must be able to explain and justify any artifact you turn in.
Your instructor may ask you to explain your work out loud, modify or extend it live, or do a timed follow-up without AI.
Not being able to explain or adapt what you turned in may be treated as evidence that you went past the assigned level.
Some assignments exist to build you, not just to produce a working artifact. When the level is restrictive, that's why.
