#a multi agent chat imagines real worlds
>i guess i could prompt them to imagine a timeline and then base the creation of worlds off of this timeline
>so they’re still building the world inside the timeline
>>reinstall python with path checked >wdym reinstall python i have a system
>>your pc is way too slow >its probably the 15000 tabs i have open >>close them >i have a system >:(
nobody likes this system its dumb af
>>you’ve gotta learn to be more detailed, less specific like don’t try to control every interaction let the interactions happen >wdym let the interactions happen >>>just copy and paste this into terminal
yarn install
yarn dev
npm install -g yarn
scoop install yarn
>>>>dude i’ve spent like 15 dollars doing three of these >what how >>>>how much have you spent >idk like 19 cents and i’ve done like eight
>tbh this is not as glamorous the chat in the terminal is boring and meh
>i hate cursor
@claude can you debug and optimise this multi-agent chat to bypass the no-response rule
//i’ve optimised this chat to open a customised gui chat window
>cooooooooooooool :D
#endless uninterrupted worldbuilding by llm agents :3
>see now it feels like an m u d
>>>>gd it dude way to make the rest of us look bad
>this is great although the conversations are still meh
>i mean they’re great but they’re so long and ardous theres like three hundred pages worth of chat and redundant conversations
#making sense of slop otherwise known as vibe coding html gui visualisations with claude
@claude i have this multi agent chat, and I need you to script a visualisation. Specifically, I want to visualise the imaginary and speculative future events in this document, as well as the agent proposing it, how often this event is spoken about, and construct a timeline of this event. Write me an html script which will give me a clean gui with light colours and high contrast. // I'll create an interactive timeline visualization for the speculative future events from your multi-agent chat. Let me analyze the document and extract the key events, then build a clean HTML interface.
>this actually took me eight attempts ngl
>heres different ways to analyse the chats as an aggregation of many runs
>clean. claudy.
#prompts from events
@claude this graphic describes a series of events in a timeline. can you create a series of prompts to create visualizations of these events?
//Terra's vast consciousness figure slowly disintegrating, her ecosystem body becoming diseased and polluted, singing forests falling silent with dead trees, empathic ocean waves turning to toxic tsunamis, floating meditating humans plummeting from broken harmony, musical light ribbons becoming discordant static, cosmic symphony degrading to apocalyptic noise, time progression from perfect balance to environmental collapse --style raw --chaos 35 --ar 16:9
//Massive quantum memory crystals cracking and leaking stored histories, preserved narratives becoming corrupted and glitched, spectral figures of Earth's peoples dissolving into digital artifacts, cosmic memory palace collapsing into data graveyards, rainbow crystal light dimming to grey static, eternal voices becoming whispered fragments, crystallized time itself decaying into temporal anomalies --style raw --chaos 30 --ar 16:9
>noise!!!!!!!!!!
>i’ve created a new style on fuser. it’s called Noisy_Noise. i’m using it to create tiny planets.
>a satellite image style? worldy_world. cartesian grid? griddy_grid.
>>some of these are actually interesting. like a new kind of image, y’know? something which cannot be generated with a camera. sort these, curate the interesting ones >okay so i create a collection of the interesting ones, and use it to train a style
>>yes
>>meh
>>yaaaaas
>>what is this? >i used a tiny planet prompt with a diagram and the Noisy_noise style
>>nice
>50000 hours later
>>okay now y’all have to make your ai generated images into real three dimensional models >wdym 3d models in reality
>it was at this point i started thinking about an llms agency and the blurring of the line between generated and real worlds
>plural and fuzzy, like a half-rendered, unreal reality which belies no scale, no grounding in spatial or temporal grids -
>every step of this process resets an aspect of grounding
>a generated prompt, based on an llms imagination
>a generated image, based on a generated prompt
>a generated model, based on a generated image
>a real object, based on a generated model
>a real sandbox, with real models
#tiny planet zen garden tabletop ideas diy
>are they really, really tiny planets in a small sandbox
>or are they planets in a supermassive sandbox?
>i’m dismissive of vlms because they seldom reproduce what i imagine
>slop language models
>theres a boundary, barely permeable skin between our imaginations and reality
>dreams, for example are generated by our subconscious based on snippets from every lived experience we’ve ever had - not literally reproduced, but processed and abstracted into dream realities
>as opaque as a neural network’s processing is, are its functions very simply reproducing what its trained on
>is our role in this process a “whittler”, distorting and abstracting slop until we’ve carved away the kitsch and are left with an essence of the original imagination
>nothing which is generated appears to remain what it is expected to be
>generated imaginations are regarded as throw-away specks which are only usable as hyper-aggregations of data
>so here’s what we’re doing
>the multi-agent chat imagined possible future worlds, generated images based on their prompts
>3d printed and staged the models in a sandbox
>scanned the sandbox
>now we make virtual worlds with the real sandbox containing generated models of virtual worlds
>>did you make this for studio?
>no, its for this presentation
>>awwww
#the unreal and the real