COMP10001 · The University of Melbourne · 2019 Semester 2
In 2019 I learnt to program in COMP10001 Foundations of Computing. Its three Python projects asked me to count an election, lead a hero through a dragon's cave and score a card game. Here they are again, running in your browser, with the working shown at every step.
The playground
Each project was a set of Python functions, auto-marked against hidden tests. Each demo below runs those functions on input you control, and shows its working.
Project 1 · eVoting
The task. Act as the programmer for an electoral commission: count votes under first past the post, a second-preference runoff and full instant-runoff voting, and check whether a ballot is a valid ranking.
Try it. Edit or generate ballots, then step through each elimination round and see exactly where every transferred vote lands.
3 winners
from one set of 13 ballots
Project 2 · Toy World
The task. Model a square cave with walls, up to three treasures, a sword and a dragon. Check a route, find the shortest path between two squares, and find the cheapest way to collect every treasure and get out alive.
Try it. Paint your own cave, watch breadth-first search flood outwards, and compare my 2019 code with a version that applies the dragon rule properly.
23 & 14
moves on the two sample caves, as in 2019
Project 3 · COMP10001-Go
The task. Score groups of playing cards (singletons, N-of-a-kind and alternating-colour runs with wild Aces), check a grouping is legal, and, for bonus marks, find every best-scoring grouping of up to ten cards inside a strict time limit.
Try it. Deal a hand, drag cards into groups to see the score update live, then let the solver check every possible grouping for the best one.
115,975
groupings searched for a 10-card hand
How it was revived
My Project 2 file survived, so the cave runs a line-by-line TypeScript port of it. A toggle switches to a spec-correct version, so you can see exactly where my 2019 self went wrong (the dragon's reach, mostly).
My Project 1 and 3 submissions are gone. Those demos are rebuilt from the task descriptions and their worked examples, and labelled as such on every page.
A uv script runs the original 2019 file and 2022 reference notebooks (provenance unclear, kept private) on hundreds of seeded inputs. The TypeScript has to reproduce every answer, including the crashes, before CI goes green.
Every count, search and solver runs on your device. The heaviest job (checking all 115,975 groupings of ten cards) runs in a Web Worker so the page never freezes.
New in 2026: property-based tests against brute force, an evaluation of whether a language model can find the best card grouping (on your own key, with 95% intervals), and the decision records behind the whole revival.
About this project
COMP10001 was my first programming subject at the University of Melbourne, and these were the first programs I wrote that had to work on inputs I had never seen. Only the Project 2 file survived; the other two submissions were lost. This site keeps what survived, rebuilds what did not, and labels which is which.
Academic integrity
This is my own work for a subject I completed in 2019. The original submission is kept unchanged in the repository for reference. If you are taking COMP10001 now, please do not copy it.
| Aspect | 2019 original | 2026 revival |
|---|---|---|
| Language | Python 3 | TypeScript (strict) |
| Runtime | Grok Learning sandbox | Your browser, plus a Web Worker |
| Interface | print() and hidden test cases | Next.js 16, React 19, Tailwind CSS v4, shadcn/ui |
| Testing | Grok's auto-marker | Vitest parity suites against the original Python |