Show HN: Bitwise Liminal – A Short Film in 256 Bytes of Code | Hacker News Reader

Show HN: Bitwise Liminal – A Short Film in 256 Bytes of Code

killedbyapixel.github.io

450 pointsKilledByAPixel61 comments

A Short Film in 256 Bytes of Code

Programmed by KilledByAPixel Presented at Lovebyte Party 2024

...

I found an old VHS tape at a yard sale. It was labeled "Bitwise Liminal" in sharpie. But when I watched the video it was only static.

Then I started having vivid and... unsettling dreams. Also I couldn't stop thinking about that weird VHS tape. After researching I learned it was a digital backup. Using some special software I recovered the data. To my shock, it was a 256 byte program...

With trembling hands I opened it in a web browser. Strange. This reminds me of the dreams I've been having. Now I don't need to sleep anymore. I just keep watching.

...

l ɐ u ᴉ ɯ ᴉ ⅂ ǝ s ᴉ ʍ ʇ ᴉ ᗺ

  <canvas id=c><svg onload=setInterval("for(t+=.1,R=T=C=Math.cos,c.width=w=128,i=9986,V=t/9&3;Y=i--/w;c.getContext`2d`.fillRect(X,Y,(C(t/6)*T^V*i/1e4*T+t)%20?(R=T++,1):(T=9,19/R),5/Z))for(Z=5,X=i%w+4/R;0<(V*95+C(t/3)+Z*X/64-Z^Z|(60-Y)*Z/58)%13;Z+=.2);",t=9)>

Open articleView on HN

This is impressive and actually looks like a real movie scene. And to think this only fits in 256 bytes makes you rethink how bloated the things we are using everyday are and also how powerful current browsers are.

I am however more impressed by the 8K demos such as "the Sheep and the Flower" [1] which was shared on HN recently or "One of those days" [2] which is a remake of an actual Youtube video [3] (granted not exactly the same video by the second but the style is immediately recognizable). They are way bigger than this 256 bytes movie but convey a lot more graphics and story.

[1]: https://news.ycombinator.com/item?id=39121101

[2]: https://www.pouet.net/prod.php?which=75790 and YouTube video: https://youtu.be/8T_Um-cw0Wc

[3]: https://www.youtube.com/watch?v=R1NagZN2kjY

Is there an annotated version breaking down each part of this for us lesser mortals?

I've had a stab at decompressing the code by hand, adding comments and meaningful var names where possible. I still don't understand the maths in it...

https://github.com/callumlocke/bitwise-liminal-expanded

Looks like the maths is a kind of raymarching loop if you want to read up on it.

Just a guess, the speedup is probably due to you caching the 2d context.

Recent and related:

Cross My Heart – A Frogger Demake in 256 Bytes of HTML/JS - https://news.ycombinator.com/item?id=39336677 - Feb 2024 (44 comments)

Normally we downweight follow-up submissions (see https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...) but since (a) the submitter is the author in this case, and (b) the work is amazing, I've made it a Show HN instead (per https://news.ycombinator.com/showhn.html) and left it up.

If this is a "short film" then you just read my new novel.

You could replace "t+=.1" with "t+=.01" and 9986 with 9e3, then you'll have a ~45 second loop still within 256 bytes, though a bit of the bottom part of the screen will be cut off.

A minor note: this is the first time I've seen an svg embedded in a canvas purely for it's onload event. Within this context the canvas is referred to by its (bare) id. How strange and wonderful!

It turns out that packing it into an svg onload is a tiny bit smaller then using a script tag!

It is possible to save a little more space by putting it in the canvas onclick event.

Why is that allowed? I didn’t realize one could do that in js.

If you nest any valid elements inside an unknown element like <asdf>...</asdf>, browsers just ignore the invalid wrapper element but still display the nested children. This fact is exploited by spec authors when introducing new tags to allow backwards compatibility: elements like canvas, video and audio all specifically allow optional children that are automatically hidden - but they won't be hidden in older browsers that don't know about the new elements. This allows you to easily define arbitrary fallback content for older browsers that don't support the new elements, just by nesting children in them.

I noticed and thought the same! Pretty cool.

I can't stress enough how much I admire that it is true 256 bytes, instead of usual 1k game jams, where submissions are presented as zipped, gzipped, and postfixed, and not including assets to fit the criteria.

Hats off.

This one is also cool. https://linusakesson.net/scene/a-mind-is-born/

This one is incredible.

thank you! it is a crazy challenge that is just shy of being impossible that seems like magic when it actually works.

> [...] where submissions are presented as zipped, gzipped, and postfixed, [...]

Well, you can always include a decompressor in your entry. PNG bootstrap has been popular enough in the JS demoscene for a long time and it starts to be effective for 1K or larger demos. There are other alternatives for smaller sizes, and 256B demos just happen to be too small to put any kind of decompressor in general.

> [...] and not including assets to fit the criteria.

I do agree on this aspect though.

Is "import gameassets" allowed in those competitions?

On a side note "import GOTO" works in python :)

> On a side note "import GOTO" works in python :)

Never heard of this, and it doesn't work on my interpreter (tested with both 2.7 and 3.11). Are you sure it's not a third party package?

It indeed does not work. I was curious, and found this hilarious article "confirming" its existence: https://www.edureka.co/blog/goto-statement-in-python/ - it's complete nonsense and appears to have been written by ChatGPT.

The feature has been implemented a few times as a joke (more or less), e.g. https://github.com/snoack/python-goto/, but that one at least hasn't been updated for 5 years.

Their work is always jaw-dropping. Here is some explanation of the tools they use: https://frankforce.com/tools/

And here's their work on the micro animation blog Dwitter: https://www.dwitter.net/u/KilledByAPixel

Youtube explanation: https://www.youtube.com/watch?v=HV7Dmo277Rs

That “YouTube explanation” doesn't actually explain how any of the Dweets work. Don't get me wrong, his work is amazing, I loved seeing it and I'm excited to listen to him show it off; I'm just saying that anyone looking for a detailed walkthrough of how any of it works, that video will disappoint.

This video is a couple of years old but it seems to get more into the steps of how he might go about things (disclaimer - I haven't watched the whole thing yet, just scrubbed through). https://portal.gitnation.org/contents/how-to-make-amazing-ge...

Stuff like this makes me believe that the entirety of life really could be encoded in DNA. I've read that if you gzipped the human genome, the file would be about 4MB, which intuitively seems too small to represent a human. But if this behavior can emerge from 256 bytes, then clearly my intuition is off.

4MB are enough to discriminate between 2^(3.2 * 10^7) possible states. Suppose you have a decompressor that can interpret a percent of a percent of a percent of a percent of those states as functional lifeforms. There are about a trillion species on the planet[1]; let's say there are a trillion times as many potential species that don't currently exist and a trillion variations within each. And then let's multiply by another trillion for good measure. Wolfram Alpha tells me[2] we've left a margin of error of roughly* ten to the ten million.

[1]https://www.nsf.gov/news/news_summ.jsp?cntn_id=138446

[2]https://www.wolframalpha.com/input?i2d=true&i=Divide%5BPower...

*'roughly' in the sense that a mountain range is roughly coarse sandpaper, but too round not to round to

Even a SHA-256 hash would leave 10^31 margin of error. The "percent of a percent of a percent of a percent" part is sweeping a lot under the proverbial rug here ;)

Is a misconception that DNA is enough by itself to replicate a living being, there are biological bases in the embryon that need to be pass down from the parent as biological infraestructure for the new being to form.

You're absolutely correct, just as you'd be correct to observe that these 256 bytes require a fairly complex software platform running on fairly complex hardware.

And all these things, including us, operate within the fabric of the universe, whose complexity is only beginning to be understood.

But in theory, would it be possible to genetically engineer a simple life form, like a bacteria, to eventually produce a human, in a process similar to the kind of metamorphosis you see in insects. Of course, this the bootstrapping code would take quite some space, but if we could do that (at least in theory), it essentially means that all you need to make practically any living being is a single generic cell and a few megabytes worth of data encoded as DNA.

I mean, when we get infected by a virus, our cells are able to produce a completely different life form (if you consider viruses life forms), so why not going from a bacteria (which can also be infected by viruses) to a mammal. It will obviously take a lot more code, and evolution can't get us there, but the idea is similar.

You may be interested in an article posted to HN a few days ago that argues against the conception of DNA as code and of cells as computers or machines.

https://www.nature.com/articles/d41586-024-00327-x

Life generated from DNA in a vacuum would probably be pretty boring. The entirety of life requires some surroundings to interact with, and as it stands there is but one of those, and singular things don't compress well.

Also, have a look at the Mandelbrot set again. There's quite a lot of information in, arguably, a very modest process.

And finally, to suggest that these 256 bytes are less impressive than they are, consider what hardware and software is needed to support it. These 256 bytes would not lead to the same illusion on a C64.

> There's quite a lot of information in, arguably, a very modest process.

But quite low in Kolmogorov complexity.

It could be a 4 megabyte-long seed for a random generator.

This made my 5560U ramp it's sysfan up 100%. :-)

Can someone explain to me what are we seeing in the short film? I see some black and white animation with boxes moving and some noisy lines on it. But I'm not sure I understand what it means. Anyone can help me understand?

That IS the film.

What's remarkable about it is how small the file size is.

It's 256 bytes!

To explain how impressive that is, the entire film itself is about 58 times smaller than a screenshot of the film's first frame. You could fit the entire film on a 1.44mb floppy disk over 5000 times and still have some space left over.

How?

Normally, videos are made up of a series of images with some information in between that says which part of an image goes where.

But this film contains no image data. It only has some math for drawing boxes at different brightnesses at different times.

And if you want boxes of different brightnesses then boy is it a good movie

Cyber-hallucinogenic cryptopoetry.

Fantastic.

Also this would be perfect as an SCP.

Reminds me a bit of Megapole (another 256b demo) https://www.youtube.com/watch?v=Z8Av7Sc7yGY

You'll want to put the code in code block (prefix line with spaces) so asterisks don't get mangled

I've changed the formatting to do that above - I hope it's better now!

https://news.ycombinator.com/formatdoc

The author has more content at https://killedbyapixel.github.io/TinyCode/

And more at https://www.dwitter.net/u/KilledByAPixel/top

My first thought on seeing the source was dwitter - it's such a fun community to peruse!

TIL that elements with ids are exposed as global variables in JS.

I've also noticed this a few years ago, but where is this documented? Was it already the case in IE6 for example? Have you found any definitive source, which states this behaviour?

The relevant WHATWG specification: https://html.spec.whatwg.org/multipage/nav-history-apis.html...

It was said to be introduced in MSIE 4 or around and then eventually standardized by WHATWG (because `window` was never formally specified before that point, AFAIK): https://stackoverflow.com/questions/9740275/html-element-id-...

HTML Spec:

> 7.2.2.3 Named access on the Window object

> window[name]

> Returns the indicated element or collection of elements.

https://html.spec.whatwg.org/multipage/nav-history-apis.html...

You can add image-rendering: pixelated on the canvas element to remove the blurring. Mind you, that may be an intentional effect.

I wonder how small the SVG spec could be golfed for this program to still work.

The first interdimensional probe returned. The universe next door is minecraft.

Thank you, I lost it in a move ages ago!

Very J G Ballard. Nice work.

Nice visual style!

What the fuck. How?

Ray marching, most likely.

Yes, this raymarching, though the vhs effect was really the trickiest part.

Love seeing demoscene on HN. Excellent work.

Neat

Reply on news.ycombinator.com