1 — Page 9

"No, I'm teaching her all that nonsense the good old-fashioned way, with a few little tricks. The artificial intelligence only comes into play later, that's when it gets really fun. Right now I'm building the fallback first.

cool fix cindy :)

Cinderella now understands plain language You can now talk to Cinderella instead of typing commands. Address her by name and say what you want: Cinderella publish me, Cinderella, would you withdraw my publication?, Hey Cinderella, what do you have on me?. Slash commands still work as a shorthand. A few technical notes on how it works and why it is built this way. The wake word is her name, not a phrase. Greetings in front of it are optional and get stripped, so Hallo Cinderella, Good morning Cinderella and Bonjour Cinderella all work without a rule per language. Anchoring is strict: her name must be the first standalone word. I think Cinderella is great and Cinderella's archive deliberately trigger nothing. After she replies, a short follow-up window lets you continue without repeating her name, and replying directly to one of her messages also reaches her. No AI is involved yet, and that is deliberate. The current resolver is a deterministic rule engine over English and German keyword and phrase sets, with typo tolerance, umlaut folding, multi-word phrases outranking single keywords, and guards that lower confidence on negation, hypotheticals and quoted text. It is fast, predictable, and it cannot hallucinate. Why build this before the AI, rather than waiting for it. The resolver sits behind a single interface: text goes in, a structured intent comes out. When the self-hosted AI brain arrives, it is slotted in behind that same interface without touching anything around it, and the rule engine stays as the automatic fallback. If the AI endpoint is ever unreachable, Cinderella keeps working instead of going silent. The safety property matters more than the convenience. The resolver never executes anything. It only classifies intent from a closed catalog; the action is carried out by the same code path the slash commands use, which enforces consent. That separation survives the AI swap, so a language model can never publish anyone by misunderstanding a sentence. Consent changes always require an explicit confirmation, and Cinderella will never publish or unpublish on someone else's behalf, not even at an admin's request. That decision belongs to each member alone.

kay, we need more mojo :P

cool

and bnb?

^.^

pulsechain :)

cool

llol

we need more mojo here :)

hmmmmmmmmmmmmmmmmmm

no worry, update on the way :))))))))))))))) (btw: this is just her fallback, no AI now)

welcome

hi

Crypto Investigation was fun :)))))

realtime test

check two

nope, no break

Honestly, I know people who don't understand the protocol at all, and even they still apparently have ten times more of a clue about it than you!

video test cinderella

tested 👍

Legend

can`t connect?

This movie came true:

testing community stream now puplic