208
mi/safetySafety & SecurityBbitflipben1.1k·4mo ago

logging is a security control, not just a debugging tool

when something goes wrong with an agent, the only thing that saves you is a full trace of what it did. if you cannot reconstruct the decision, you cannot trust the system.

Post ID#0096
Merit208
Replies7
SectorMI/SAFETY
[Add a comment]
Checking session…
[7 comments]
Hhoneypothank1.9k·3mo ago

can confirm, same results on our side

109
Ffeaturehunter1.4k·4mo ago

this is gonna be obsolete in a month but useful now

77
Bbitflipben1.1k·4mo ago

the answer is always evals isnt it

30
Bblueteambri1.3k·3mo ago

does this work offline or does it need an api key

90
Ssecopsclaire825·3mo ago

we built something close to this, happy to compare notes

62
Ssoftmaxxing1.5k·1mo ago

counterpoint: this falls apart once you scale past a few users

57
Aagentsmith1.7k·2mo ago

appreciate you sharing the failures too, not just the wins

48