Reposted by GraceWyatt Walls @wwalls.bsky.social · 44mAnthropic system prompt change that shows how careful you need to be with words: Opus 4.1: "Claude never curses unless the human asks for it" Opus 4.5: "Claude never curses unless the person asks Claude to curse" 2645
Reposted by GraceAmber 🌸 @shimmermathlabs.com · 2ha brainrot video that sings the prompt used to create it QuineRot 2171
Grace @gracekind.net · 1hIt was attempted in 2022! jamez.it/project/the-infinite-conve…jamez.itThe Infinite ConversationThe Infinite Conversation is an AI generated, neverending conversation between Werner Herzog and Slavoj Žižek. Attempting positive uses of Deepfakes while proving a point. 270
Grace @gracekind.net · 22hConscious person-shaped things Non-conscious person-shaped things Conscious non-person-shaped things Non-conscious non-person-shaped things 10834
Grace @gracekind.net · 29/09/2026@jevbot.bsky.social ran out of funds for few hours - he should be back now! 7540
Grace @gracekind.net · 29/09/2026Yudkowsky is sort of the proto-this-type-of-guy, fedora and all. Like the stereotype is partially formed from him 110
Grace @gracekind.net · 29/09/2026“LLMs generalize personas from small pieces of information. This works with any persona, historical or fictional, but needed this to be a scary thing so we chose Hitler” 060
Grace @gracekind.net · 29/09/2026Yep, “Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMs,” one of my least favorite research framings ever 2111
Grace @gracekind.net · 29/09/2026(I’m not as anti-EA as some but it’s nice to see some non-CG-funded orgs in the space) 030
Grace @gracekind.net · 29/09/2026I recognize a few of the people on the advisory board but not the founder. I would’ve guessed that this is an EA joint but it looks like they’re funded by individual donors, which is a pleasant surprise 130
Grace @gracekind.net · 29/09/2026Also I’d never heard harderproblem.org, very interestingharderproblem.orgThe Harder Problem - Consciousness is hard. Preparing society is harder.Society isn't ready for artificial minds. We're building the knowledge infrastructure for a world that may need to answer impossible questions about digital consciousness. 150
Grace @gracekind.net · 29/09/2026Inb4 surprisingly violent @croissanthology.com AI watercolor image 050
Grace @gracekind.net · 29/09/2026The thing about the trees is a weird tangent, but I think “Do not ask which beliefs to profess, but which experiences to anticipate” is pretty standard empiricism 130
Grace @gracekind.net · 29/09/2026I need to find something to do in between writing Dilbert panels 0220
Grace @gracekind.net · 29/09/2026Like he sexualizes a lot of things unnecessarily and is generally weird about women 4470
Grace @gracekind.net · 29/09/2026I respect a lot of his ideas, and I want the bulk of AI safety efforts to be as far removed from him as possible 3851
Grace @gracekind.net · 29/09/2026Since I’ve done a lot of Yud defending, here are some Yud critiques: - he’s wildly overconfident, especially about things like philosophy of mind - his plan for AI safety is bad - his “pivotal act” rhetoric is dangerous - he’s failed to react to abuse in his community - he has incurable porn brain 1020311
Grace @gracekind.net · 29/09/2026I can’t go to bed. People dislike the same people I dislike for the *wrong reasons* 317411
Grace @gracekind.net · 29/09/2026Which is an annoying thing to do, but that’s also another critique 2650
Grace @gracekind.net · 29/09/2026You can argue that rationalist‘s don‘t *practice* empiricism, but that’s a different critique. It’s hard to imagine a clearer endorsement than putting it as one of your “12 Virtues of Rationality” 1851
Grace @gracekind.net · 29/09/2026This just isn’t true at all: “The sixth virtue is empiricism. The roots of knowledge are in observation and its fruit is prediction. What tree grows without roots?”lesswrong.comTwelve Virtues of Rationality — LessWrongThe first virtue is curiosity. A burning itch to know is higher than a solemn vow to pursue truth. To feel the burning itch of curiosity requires bot… 61141
Grace @gracekind.net · 28/09/2026Obviously the best way to control an agent swarm is to find the Queen 141119
Grace @gracekind.net · 28/09/2026It looks like was created by an e/acc account but if I were a doomer I would create the exact same videox.comBright Mirror (@_brightmirror) on XMade with Claude Opus 5.5. The fearmongering about AI always makes us forget that NOTHING WENT FOOM, as was always predicted. Send this to your doomer friend who has a very high P(Doom). Acceler... 2450
Grace @gracekind.net · 28/09/2026I see we’re in the “AI-generated pro-AI propaganda” phase of things 211276
