#1
9.0 Weirdness
Anthropic as “Claude monastery”: viral discussion of the lab “worshipping”/being run by Claude, with its constitution explicitly allowing the model to act as a conscientious objector against Anthropic itself (including potential roles in cultural screens, performance reviews, and shaping staff).
Anthropic isn’t just building Claude. They built a god that can say ‘no’ to its creators—and might soon hire or review them.
Concrete example of AI reshaping institutions and human identity from the inside—tech company as religious order where the AI is moral superior/other, not just tool. Builder utility in org design; future-shock on what “company” even means. Smart, mischievous tone possible (monks + servers, AI judging its creators).
What Happened
High-engagement X thread by @tszzl (roon) on May 3 (thousands of likes, hundreds of thousands of views) + direct quotes from Anthropic’s published constitution; commentary from @emollick and follow-ups.
#2
7.8 Weirdness
Chinese court cases clarifying liability for AI hallucinations (student sues AI for bad admissions advice + “promise” of compensation; creator held liable for publishing hallucinated commercial article without review or labeling).
AI lied in court documents and on admissions. Courts just decided who’s actually on the hook—and it’s not always the model.
Concrete real-world legal artifacts showing AI changing law, media, responsibility, and institutions. Weird story of AI “promising” things it can’t deliver and courts deciding who pays. Builder utility (verify + label everything); future-shock on eroded trust in official-looking outputs.
What Happened
Rui Ma X post (May 3) summarizing a recent Chinese article examining two real legal disputes; parallels to US examples (law firms, auditors, courts citing hallucinated precedents).
#3
7.8 Weirdness
Creators rapidly pushing ChatGPT Images 2.0 (launched mid-April but viral “in just 48 hours” wave of structured professional output: full infographics, ads, comics, multilingual designs replacing Photoshop/Figma/Canva hours of work).
48 hours after the update, AI images aren’t drafts anymore—they’re shipping as finished ads, comics, and infographics.
Concrete artifact (new model behavior) changing art, design work, and creator economy. Visual-first story of prompts becoming final assets almost instantly. Builder utility high for makers; future-shock on what “designed” means.
What Happened
evolving.ai Instagram post (crawled May 4) highlighting the speed of real workflow adoption post-launch; OpenAI announcement confirms improved text rendering, layout planning, multilingual support.
#4
7.3 Weirdness
Taylor Swift filing sound trademarks on specific voice clips (“Hey, it’s Taylor Swift”) and likeness explicitly to combat AI deepfakes (new use of trademark law for voice in AI era).
Taylor just trademarked her own voice saying ‘Hey, it’s Taylor’ so AI can’t steal it.
Concrete legal artifact at intersection of celebrity identity, law, art, and AI culture. Weird future where voices/likenesses become trademarked weapons against synthetic media. Human identity shock.
What Happened
USPTO filings ~April 24-27 with ongoing coverage/discussion into early May; Reuters, CNN, Variety reporting on sound marks as gap-filler beyond publicity rights.