# Assignment 1 > The first individual prototype: an interactive explainer built with an agentic workflow, marked on the legibility of your process and your response to the brief. ## The brief > Build an interactive explainer of something you think more people should know > or understand. The topic is yours: a phenomenon, a system, a piece of maths, how something you care about actually works. The genre supplies the discipline: one strong idea, one dataset or mechanic, and nothing else. _Interactive_ is doing real work here, because the visitor has to do something, not only read. Any register works --- earnest, playful, polemical --- and the marks go to the strength of the idea and the judgement in its scope. The point of view lives in deciding why your something is worth more people's understanding. Low floor, high ceiling: everyone can ship _something_, and the marks sit on process and response rather than polish, so what matters is choosing your something and owning it. It's a static, client-side build deployed to GitHub Pages, and the genre stretches static about as far as it goes, so there's headroom for as much ambition as you can direct. The week 4 [retro](https://comp.anu.edu.au/courses/comp4020-agentic-coding-studio/crits/03-a1-retro/) will ask for a short demo and the breakthrough that made it click, so keep it in mind while you build. Exemplars --- each one strong idea, one dataset or mechanic, and nothing else: - [The Deep Sea](https://neal.fun/deep-sea/) --- scroll as the interface: depth is the data - [Spend Bill Gates' Money](https://neal.fun/spend/) --- a single absurd mechanic carries the whole page - [Absurd Trolley Problems](https://neal.fun/absurd-trolley-problems/) --- branching interaction as narrative; humour as engagement - [Film Dialogue](https://pudding.cool/2017/03/film-dialogue/) --- scrollytelling data journalism: 2,000 screenplays, one question - [Human Terrain](https://pudding.cool/2018/10/city_3d/) --- data-to-visual mapping at its most ambitious - [Elevators](https://john.fun/elevators) --- a live simulation is the explanation: change the traffic pattern, flow and scheduling algorithm, then watch the distribution of wait times move - [Mechanical Watch](https://ciechanow.ski/mechanical-watch/) --- Bartosz Ciechanowski's interactive teardowns (see also [Airfoil](https://ciechanow.ski/airfoil/) and the rest of his [archive](https://ciechanow.ski/archives/)): the genre's ceiling, where every part is manipulable and the explanation _is_ the interaction ## What you submit - the **deployed prototype** (a GitHub Pages URL) --- this is what gets marked - the **source repository** it was built from --- this is how we read your code - **evidence of process**: your `PROCESS.md`, your `CLAUDE.md`, one reflection entry and the commit history behind them. The [assessment page](https://comp.anu.edu.au/courses/comp4020-agentic-coding-studio/topics/assessment/#what-you-submit) sets out what each one is and how to cite a moment; the starter's `PROCESS.md` template shows the format. Three things are specific to this assignment: - `PROCESS.md` runs to **400--600 words** and carries **three or four moments** --- three or four, not more, because each needs room to say what you did _instead of_ the obvious thing and how you knew the result was right. Those two are where the marks are, and neither is something the repo can tell us on its own. - the strongest moments are the ones where a correction landed in the **harness** --- the standards and checks your work has to satisfy --- rather than in a retry: a rule added to `CLAUDE.md`, a check wired up, an attempt thrown away. Retrying until it passes is the routine case, and changing what the work runs against is the skilled one. - the reflection is `reflections/assignment-1.md`, and it's the breakthrough you'll present at the [week 4 retro](https://comp.anu.edu.au/courses/comp4020-agentic-coding-studio/crits/03-a1-retro/), which reads this entry --- there is nothing to write twice. The submission mechanics, the marking environment, deadlines, extensions and appeals are the same for every task and live on the [assessment page](https://comp.anu.edu.au/courses/comp4020-agentic-coding-studio/topics/assessment/). Read it before you submit. ## How it's marked Worth 20% of the course. Three criteria, each marked out of 100 against its own band, then weighted and summed to the task's mark: | Criterion | Weight | | ------------------------- | ------ | | Legibility of process | 45% | | Working deployed artefact | 20% | | Response to the brief | 35% | Process is the largest criterion, and deliberately so: shipping is necessary but not sufficient, and the marks go to a reader being able to see _how_ you directed, grounded and corrected the work. Corroboration is the floor of that band: the HD is for evidence that shows _skilled_ directing, not merely a truthful account. The response criterion is the other place the marginal marks live, because the brief is a provocation and the strongest answers to it have a point of view. The band descriptors --- what HD, D--Cr, P and N mean for each criterion --- are the same for both assignments, and live with the rest of the marking model on the [assessment page](https://comp.anu.edu.au/courses/comp4020-agentic-coding-studio/topics/assessment/#marking-rubric). ## Key dates - **Due:** noon, Monday 17 August 2026 - **Marks and feedback returned:** 28 August 2026 --- deliberately before the 31 August census date, so you have a real mark in hand before the last day to drop the course without academic or financial penalty There are no late submissions, but extensions are easy to arrange _before_ the deadline --- see the [assessment page](https://comp.anu.edu.au/courses/comp4020-agentic-coding-studio/topics/assessment/) for how. This is an individual assessment. Source: \ Licence: [CC-BY-NC-SA-4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/)\ Cite as: Swift, B. (2026). _COMP4020/COMP8020 Agentic Coding Studio_ [Course materials]. ANU School of Computing. https://doi.org/10.5281/zenodo.21395760