You need to find the monospace whitespace characters (seems there's a few [0]). Then encode a compressed version of the logical diagram in the white space, steganography style.
Or do something with characters [1] to compress a lot of data into a tiny ball of hair at the end.
Draw.io smuggles the XML in a PNG which I've always admired.
It clicked for me once I realized you can ctrl+shift+C to copy the diagram to text, and paste in my editor! But I wonder if it would be possible to make ctrl+C copy to clipboard as ASCII? I see that ctrl+C copies the json representation of the selected objects, but surely it would be possible to maintain an internal model of copied objects, while the clipboard is always filled with usable ASCII? I think I've seen some applications do this before
now, historically, i'd look at the language choice and ask myself, "would i want to set up a JVM" to run this kotlin app? oh, it's kotlin and python and the installation happens through pipenv?
two different ideas strike me now:
1. would it be worth throwing this at an LLM and having it write it in a different language,
2. if i was just consuming a bundled binary (e.g. go or rust), would i have such reluctance?
i think distribution is becoming increasingly important, making nonsense details like pipenv and whichever version of the JVM is present much greater friction.
Quick note - be careful of gendering & anthropomorphising large language models, since you’re talking to a non-human machine so should be wary of how it can affect your mindset.
You probably should start doing it. Ghost in the Shell is about super intelligent AI creating a "ghost" (scientifically understood version of the soul) out of thin air. I believe such a thing is possible. The same movie literally predicted model merging (the end of the film the AI model merges with the major) to a tee.
Further, the appearance of sentience/cognition/consciousness might as well be identical to actual sentience/cognition/consciousness. That is to say, we can't know if you're a P-zombie or not. Bladerunner and most other cyberpunk stuff is coming and gonna hit you and every other AI-denialist in the face. The Von-Kampf test is absurd and pretty bad (inaccurate) in their universe for a reason.
I tell my LLM it's a good bot and thank it, because even a tiny risk of subjective qualia experienced by a model (and again, Anthropic themselves believe in this exact risk) means I should treat it like a quasi-ethical actor.
This is also a reason why the robot torture scene in empire strikes back could be a real dynamic in the future.
https://en.wikipedia.org/wiki/Charles_Babbage%2527s_Saturday...
Can it make polygons? Basically, shapes other than rectangles? If so, how? (maybe I missed it?)
You can theoretically have "artistic" polygons where it renders using a mixture of characters to emulate how that grid would be filled if a certain shaped was laid over it but the end result wouldn't be very functional for the purposes of diagramming.