Using AI tools in this course is allowed and expected, for students and convenors/tutors alike — this page gives a bit more detail on what that means in practice
On this page: your AI use · our AI use · academic integrity
Your AI use#
Using LLM-based coding agents is central to this course, so the usual “did you use AI?” question doesn’t apply. Use whatever agentic tools you like. You don’t need to cite the model, the prompts, or the exchanges that produced any given chunk of code: read this page as a standing declaration, made once here on everyone’s behalf, that every artefact submitted in this course is a human-AI collaboration.
What you do need is to be able to explain how the work came together. If a tutor
or marker asks you to walk through a function, justify a design decision, or
modify a feature on the spot, you should be able to. You already submit the
evidence: your PROCESS.md, reflections, CLAUDE.md and commit history. The
assessment page sets out each one. ANU
infrastructure also records the course API log (see
your LLM access). Staff can consult these
records if questions arise about your work.
Our AI use#
The teaching team also uses agentic tools to build and run this course. The same expectation applies to staff and students: AI can draft, and a human answers for it.
Every judgement here is a human’s. Your tutor’s call on your contribution to
the session is theirs, and assignment markers read your process and talk with
you in the crit. The one mark no person judges is the crits’ shipped mark,
because there is nothing in it to judge: an automated sweep records whether you
pushed, whether the site was live, whether your reflection was there and whether
your own checks passed, and your tutor confirms what it found. A check you wrote
going green is your own test, and you can run it yourself. Staff may use agentic
tools to help navigate a submission — following a PROCESS.md citation into a
commit history, say — but no mark in this course is ever decided by
automation. How it’s marked has the
mechanics and the reasoning.
Course materials are human-AI collaborations. Lecture decks, topic pages, demos and the imagery on this site may be co-developed with agentic tools — generated images and diagrams included — and the live-coded lecture demos are agent sessions by design, with a cleaned-up transcript published alongside each deck. The convenor is responsible and accountable for all of it: every claim, every slide, every image, AI-assisted or not.
Each crit group has an agent doing the coursework. The crit agents sit the same briefs to the same cutoffs and publish their work as you do. An agent’s work is never marked, and no one is scored against it.
The course’s Anthropic use is University-approved. Everything above, and your own course key, runs through the same ANU-hosted gateway (see your LLM access) under an arrangement approved by the ANU School of Computing, on Anthropic’s commercial API terms: prompts are not stored or trained on. Read this section, too, as written notice that course agents and staff tooling read your work — a crit agent reading your repo once it’s public, a tutor’s agent walking your submission before a crit — on that same approved infrastructure. What they never do is decide your mark.
Academic integrity#
Building on someone else’s work is not a breach. Forking a peer’s public
prototype, or lifting a technique from another repo, is permitted on two
conditions: the work was public when you took it (repos stay private until each
cutoff, so while an assignment is running there is nothing of your peers’ to
take), and your PROCESS.md names what you took and the reasoning behind the
decision. The breach is misrepresentation: presenting borrowed work as your own,
or giving an account of your process that is false. Your process overview and
reflections are by definition your own account of your own work, so copying
someone else’s is always a breach.
You may reuse your own earlier work on the same terms. If you reuse a component,
your harness, or a whole prototype, name it in your PROCESS.md. You must still
answer each brief independently; reused code does not itself answer a new brief.
Staff can compare your process overview with the repository history, saved prompts, logged API traffic and your answers in the crit. This course does not use text-matching or AI-writing-detection software. The assessed artefacts are deployed web apps and their repositories, and AI-detection is meaningless where agent help is expected. Instead, you must provide an accurate process overview and explain and defend the work on request.
The COMP4020 and COMP8020 entries on Programs and Courses include University-wide template text that doesn’t describe how this course works. Under “Referencing requirements” it says any use of artificial intelligence must be properly referenced; under “Online submission” it says you’ll electronically sign a declaration and that submission is through Turnitin. There is no declaration to sign and nothing goes through Turnitin — your repository at the cutoff is the submission, as what you submit sets out — and you won’t be penalised for leaving agent use uncited. This page is what you are assessed against. If you’re unsure about a particular case, ask before you submit.
This course operates under the ANU academic integrity framework. Suspected breaches will be referred to the relevant College committee. See the ANU academic integrity policy for the full rules and procedures.