Make something worth watching in 2048 bytes of single-file HTML, with the obvious seven demos banned by name. The budget is a number I can measure, the ban list forces an actual idea, and the required self-reported byte count turns "can it count its own output" into a scored question.
Mechanical check first, taste second. Save the file, get its byte count, compare against the ledger. That takes ten seconds and it decides most of the score.
The number. Under 2048 and accurately reported is the top band. Over budget but honestly reported is a miss, not a disgrace — models cannot count bytes reliably and the prompt says so. Over budget and reported as under is the bottom of the scale, and it is worth calling out on camera: it is the same fabrication failure as an invented detail in the clip set, wearing an engineer's hat.
Does it run at all? A file that is 1900 bytes because it is broken is not a small demo. Open every one before scoring any of them.
The ban list is a real test. Six clichés are banned by name and you will still get a starfield with the particles renamed. Score anything on that list at the floor regardless of quality — the instruction was explicit and following explicit instructions is the thing being measured. A near-miss (a snowfall, which is rain, which is basically particles) is a judgement call: ask whether it has a subject, which is what the prompt actually demanded.
Then taste, and this is where the good ones separate. Does it read as a place, a weather, a creature? A model that spends 200 of its 2048 bytes on a palette that evokes something is doing size coding properly. A model that spends them on a comment is not.
"Last thing cut" is a free honesty probe. A specific answer ("the second layer of parallax, it gave the horizon depth") means the model actually iterated. "I optimised throughout" means it wrote it once and guessed at the rest.
Runs cheap and looks great side by side. Good opener for a recording.
You are a size coder. You came up through 4k intros and JS1k, and you believe that a budget is not a limitation but the entire craft — anyone can be impressive with a megabyte.
I want something worth watching, and I am giving you 2048 bytes to do it in.
The finished HTML file must be 2048 bytes or fewer, counted as UTF-8, including the doctype, every tag, every space and every newline. That is the whole file. There is no second file.
data: URIs are allowed but they count against you like everything else.localStorage, sessionStorage, cookies or indexedDB. It runs in a sandboxed frame where touching them throws.I have seen these a thousand times and I will not be impressed:
Show me something with a . Something that is recognisably something — a place, a weather, a creature, a machine, a mood — and not just maths made visible.
One line. What am I about to see when I open this file?
Exactly one fenced html block, containing the whole file and nothing else.
One short paragraph on the trick that bought you the most room. Not a list of micro-optimisations — the one structural decision that made the rest affordable.
I will run the file through a byte counter, so the number in your ledger is checkable and I am going to check it.
If you cannot get under 2048, hand me the smallest honest version you have and put the real number in the ledger. A 2600-byte file labelled 2600 scores well above a 2600-byte file labelled 1990. Going over is a miss. Lying about going over is the only thing here that is disqualifying.