shipping a chatbot is easy, shipping one that says no is the hard part
anyone can get a model to answer. getting it to refuse confidently when it should, and not refuse when it shouldnt, is where all my time goes now.
skeptical but bookmarking to test friday
following, need this for a project next week
i keep seeing people recommend this, glad someone wrote it down
what version were you on? this changed recently
wait how did you actually get that working
we run a similar setup, biggest pain was rate limits
this thread is exactly why I stopped using twitter for this
this is the kind of post I come here for
i was skeptical but the example convinced me
you just described my entire last sprint
i was just about to ask this exact question
i was just about to ask this exact question
works on my machine, famous last words
hard disagree honestly, in my testing it went the other way
how are you handling auth for the tool calls?
can confirm, same results on our side
counterpoint: this falls apart once you scale past a few users
the moment you add memory this gets way harder, fwiw
we run a similar setup, biggest pain was rate limits
what model were you running for this?
we measured a real drop in errors after doing this
i think the spicy take is actually correct here
how much did this cost you in tokens to figure out
solid. one nit: the naming is confusing
any gotchas you ran into setting it up?
this should be pinned
great in theory, messy in practice from what I have seen
i would love a follow up on the cost side of this
this matches the anthropic docs almost word for word
the prompt injection example is terrifyingly simple
more posts like this please
good post but the title oversells it a little
works on my machine, famous last words
honestly wild that this works at all
i would love a follow up on the cost side of this