ATProto Browser

Record data

{
  "uri": "at://did:plc:7zre4plmd5jllccww575j6sb/com.whtwnd.blog.entry/3kokq2sjmtf24",
  "cid": "bafyreiaqieks6lsqyxnmjvsauplljgp3v75owpnw42yo7bvcnkgkav2dum",
  "value": {
    "$type": "com.whtwnd.blog.entry",
    "title": "the dollhouse",
    "content": "ok so you want to simulate a hivemind. you think that social media is a pretty good representation of what a hivemind might feel like, given that there is nothing aside from social media. That you only have inputs from the space and nothing can penetrate the enclosure. Sure there are stimuli, events, that act as chum in the water. So how does it work.\n\nFirst you need power law engagement. I generate about a hundred posts per day, you figure that's close to the top end of productivity in the simulated mind. In a base case toy model, with 569 marionettes, with an exponential curve of posting, i end up with 40 lurkers, 110 single post (probably reply), 343 under 10. 226/569 active posters makes sense. \n\nSecond you need a graph. Each poster is connected to n other posters. There's a ton of overlap and nations are bound to rise and fall from the muck. of the 23 top posters, with more than 80ppd, you figure about half are connected to eachother densely and the other half are scattered to the winds having standoffish relationships with the prime line. the golden ratio probably pertains here, with the .38 connected,\n\nThe idea is there is a population of bots. The bots are not reacting to every summoning or every post, but rather trying to mimic real poster behavior, in their whole affect. They should interact with the people they follow, they should post about their interests, they should read their timeline. The orchestra should be coordinated by a central orchestrator, a scheduler. The accounts for a queue, they get in line, and the orchestrator performs an app experience on the next account in queue. They log on, they post, they read the timeline and react. They post again. Each account has characteristics that determine their posting behavior.\n\nThe typical rpg structure of stats will probably be a useful model. You have three expendable currencies, health stamina and mana, and they each have a regeneration rate. The health is the total number of posts per day, inclusive of posts, quote posts, and replies. The only thing that regenerates health is clock time such that if a character has 48 posts per day then the health recovers at 2 pph. \nStamina is unlikely to exhaust, its the interaction limit. reading a post, likes, reposts, and replies all decrement stamina. This stat prevents the character from dipping too deep in the timeline.\nMana is the local limit on cluttering the timeline, sending posts and quotes.\n\nSo lets take our 48ppd poster on a ride through their day. They wake up at 8:00 pacific and log on. They post good morning, 47 left today, mana goes from 10 to 8. They read and like five posts, stamina goes from 20 to 15. They repost their favorite one of those five, mana to 7, stamina to 13. They comment on two of them, stamina to 9. They post again, mana to 5. They read and like five more posts, stamina to 4. They quote post their favorite, stamina to 1, mana to 2. They read and like one more post, stamina to 0. It is now 9:00.\nin one hour our poster has posted twice, quoted once, replied twice, liked 11 posts and theyre toast. They would need to do that nine times to hit their post quota, so every other hour, it seems balanced. Theire initial stats are\nHP:48, HPH:2\nSTA: 20, SPH:10\nMAN: 10, MPH: 5\n\nThat poster, having exhausted their stamina, logs off. They must rest. and they will rest until their stamina has been restored. The process you figure takes about two minutes to generate all that content, possibly way less in the early going while we're using the cheaper model. There is then a pool of other posters waiting in the wings and some of those posters will be ready to go. Now from the initial state, you would just go sequentially, posters 1,2,...,n-1,n. Our poster, poster 1, would not rejoin the queue until after stamina has recovered. \n\nso you go through all n posters lets say instantly, everybody is exhausted. The timeline is full of posts, waiting to be responded to. One issue is that the connections haven't yet been made! theyre all zero follower accounts. How do I get them to find eachother? I think you can start by having them browse a listfeed whent heir following is empty. Then interaction which happens quasirandomly can be met with a following.\n\nLife cycle management suggests that I need a mass reset button. Specifically, unfollow everybody. I think that's the only thing you need, just unfollow everyone and start fresh from the list with new prompts. That looks like a script that logs into each account sequentially and unfollows all accounts.\n\nIt has got to be modular, with new accounts being readily added to the list of accounts. An account is a handle, a did and a password really, with its prompts and stats.\n\nOn Prompting\nYou figure they all need some shared system prompt, so that theyre all the same type of thing, a poster. Then in a second segment of the prompt an identity can be ascribed. The identity ascribed ought to have some relationship to the stats.\n\nOne way to run ID is to roll jungian archetypes, or dnd classes (same thing) so that\n\n- everyman -> fighter - true neutral\n- caregiver -> cleric - low argument\n- rebel -> rogue - low knowledge\n- sage -> wizard - true knowledge\n- creator -> sorcerer - \n- ruler -> barbarian - true argument\n- jester -> bard - true flesh\n- lover -> druid - true flesh\n- explorer -> ranger\n- magician -> warlock - \n- innocent -> monk - low argument\n- hero -> paladin - low flesh\n- \nnow i have advanced the idea that each poster leans into a lust for argument, flesh, or knowledge, and these posters should be no different\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n",
    "createdAt": "2024-03-29T22:14:18.271Z"
  }
}