@rauchg would be nice if it had a rate limits warning and would automatically reconnect when i restart the session. i know it's early but it's pretty unstable in its current state
@NickSchmidt@GoogleDeepMind@tszzl@heyclicky i see you. was my first thought. clicky is a cool concept. what's inevitably gonna happen is that everything we coded on their models is gonna be gobbled up if promising enough.
@viktoroddy it's insanely limited by...well... it's limits. 3-5 modest requests and you burn through your daily allowance. it's amazing but also a token eater. it's time for a new pricing model.
@heynavtoor@mrscottiem Big tech is on it with layered defenses like input filtering, privilege separation, output guardrails, and adversarial model training, though semantic and multi-hop injections remain largely unsolved.
@heyshrutimishra why are we constantly ignoring data security? no company in their right minds would or can use this. make the models more efficient and open source. that way we can all profit instead of offering our data to a select few.
@elonmusk proper STT would be nice. and less book-y answers. i know i can ask for it and give it diff. defaults but the main default should be "concise". also: "why this is working" is a noisy validation layer. answer the question, don't drown the user in half-relevance.