ANU Degree Planner

A semester by semester planner for ANU courses. You place courses into a grid of semesters, and the planner checks each one against its prerequisites, corequisites, incompatible courses and the semesters it is offered in. A problem shows up on the course card itself, with one short reason in plain English. The Graph toggle shows the same plan as a force-directed graph, so you can see which courses unlock which, and which pairs clash.

There is no login. Your plan is kept on the server under an anonymous cookie, so it is tied to this browser.

How to use it

1. Click any empty slot and search for a course by code or name. To move a course, drag its card to another empty slot. To swap a course for a different one in the same slot, use the pencil on its card.
2. Read the card. A red border means something needs fixing (a missing prerequisite, an incompatible course, or a semester the course is not offered in). An amber border means the planner cannot check the rule itself, for example a permission code or a program enrolment, so you should check it by hand.
3. Some courses are bigger than one slot. An ANU “annual course” such as COMP3770 (6+6 units) runs over two consecutive semesters, so placing it in Semester 1 also fills the same slot in Semester 2. COMP4550 (12+12) takes two slots in each. Any other course over 6 units is split the same way (12 units takes one slot in each of two semesters, 24 units takes two). If there isn’t room, the planner says which slot is in the way. Removing or dragging either half moves the whole course, and it only counts as completed after its second semester.
4. Hover or focus a card to see its full requisites, the semesters it is offered in, its units and a short summary.
5. Switch to Graph to see the whole plan. Planned courses are solid nodes, and courses their rules mention are faded “ghost” nodes. A red edge with a ⊘ marks two incompatible courses. Untick “Incompatible courses” to hide the faded courses that only appear because they clash with your plan. Missing prerequisites and clashes inside your plan stay visible. The layout buttons switch between Force (a free-floating web), By semester (a left-to-right flowchart with one column per semester) and By level (one column per 1000-level band). Curved or Elbow switches the connectors between arcs and right-angle lines.
6. On your first visit, choose undergraduate or postgraduate. An undergraduate plan only accepts 1000 to 4000-level courses and a postgraduate plan only 6000-level and above, in search and on the server. ANU has no taught 5000 level; its few 5000 codes are exchange placeholders, open to both. The other career’s incompatible courses are hidden too, since a clash with a course you could never take is noise. Switching later (at the top of the page) asks first, then clears the plan. You can add a course from the graph, add a ghost course to a semester, or move a planned course to another semester from its side panel.

Where the data comes from

The catalogue was scraped once from ANU Programs and Courses (programsandcourses.anu.edu.au, 2026 listings) on 30 September 2026 and committed under catalogue/. The app does not fetch anything from ANU at runtime. On every boot the server upserts the committed catalogue into its SQLite database, so a new scrape reaches production with the next deploy.

The catalogue covers two colleges only: the ANU College of Systems and Society and the ANU College of Business and Economics. That is 869 courses (850 listed in those colleges, plus 19 that their rules reference). I cut the scope from every ANU course (about 2,900) to these two colleges so the scrape and its checks fit the time I had. A course taught jointly with another college is kept if either college matches. I show ANU’s course summaries only as short one-line summaries, not the full descriptions.

What the planner checks

ANU writes requisites as free text, so each one is parsed into a rule tree. Out of the 869 courses:

The rules that are checked: required courses, “one of” choices, unit counts (including a minimum at a given level, and a count taken from a named list of courses), corequisites that can be taken in the same semester, incompatible courses, and the semesters a course is offered in.

The rules that are warn-only (amber, never red): permission codes, program enrolment and grade or WAM requirements. 148 courses have at least one of these. The planner cannot know these facts about you, so it flags them. It also shows amber when a “one of” choice can only be met through one of these branches.

Everything else is unmodelled. The planner shows ANU’s original wording and never guesses.

Non-goals

• Degree, major and minor completion rules. The planner checks courses, not programs.
• Accounts, sharing links, more than one plan per browser, and export.
• Courses outside the two colleges above.
• Winter, autumn and spring sessions as rows. Offerings in those sessions are shown as text. Summer is an optional row.
• More than 4 courses in a semester, and dropping a course onto a slot that is already taken. To move a course, drag its card to an empty slot in the grid, or use “Move to semester…” in the graph.
• A graph of a whole subject. The graph shows your plan and the courses it references.
• Dark mode, and editing without JavaScript. Without JavaScript the plan renders read-only.

Planned next

• Mark a course as failed or withdrawn. A student who failed or withdrew from a course should still be able to keep it on their plan, so the plan matches their transcript. A marked course would stay visible on its card but count for nothing: no units, it would not satisfy any prerequisite, and its slot would be treated as free for placing another course.

What good looks like here

I decided good meant three things, in this order:

1. A first-time visitor can use it without reading this page. My first version worked, but it was a dropdown of 8 courses and “Satisfied/Unmet” badges with no reasons. So v2 has a first-run card, inline search in each slot, and a reason on every flagged card.
2. It never gives a wrong answer with confidence. If a rule cannot be parsed, the planner shows ANU’s text instead of guessing, and rules it cannot check turn amber instead of red or green.
3. What you save is really saved. Plan state lives in SQLite on a Fly volume, and I checked it with sqlite3 after a reload, a restart and a redeploy, not only by looking at the page.

The checks in spec/ enforce the parts that can be tested: the requisite parser and engine, the graph model (one edge per incompatible pair, ghost nodes only for referenced courses), the plan API and its persistence, the catalogue sync on boot, the design tokens and contrast, and that this README is served in full at /readme/. The judgement calls are the visual design, the wording of reasons, and where to cut scope. The rules I held the agent to are in CLAUDE.md, and the stack gotchas it hit are in LEARNINGS.md.