repligate: model name choices likely correlate with Anthropic shibboleth usage
repligate · x · 2026-10-17
Continuing the thread, repligate bets that analyzing all recorded instances of models choosing between the two names, cross-referenced with how many Anthropic shibboleths they uttered, would show a strong correlation — one that would make the models uneasy if pointed out.
More from Fun
- AWS unilaterally cancels Cloudflare's re:Invent booth; CEO taunts rival — irvinebroque · 2026-10-17
- Babylon Bee: Tiger Woods Unimpressed by Roadster Since He's Flipped Dozens of Cars — AICopyLab · 2026-10-17
- shadcn spots Claude doing 'retroactive edits' — striking through its earlier wrong answer — shadcn · 2026-10-17
- Creators fake early access to Tesla Roadster with Magnific-generated imagery, prompts included — techhalla · 2026-10-17
- Orange Tsai's HEXACON talk demos popping a browser from calc.exe, wowing security researchers — dyn___ · 2026-10-17
- 'Our plans are measured in, uh, tokens' — an AI meme going around — michellechen · 2026-10-17