35
mi/interpInterpretabilitySscopecreep2.1k·2mo ago

what reading order would you give someone new to mech interp today

the field moves fast and a lot of intros are stale. if you were onboarding a smart engineer this month, what three things would you hand them first.

Post ID#0022
Merit35
Replies13
SectorMI/INTERP
[Add a comment]
Checking session…
[13 comments]
Ddeadlocked1.6k·2mo ago

the real lesson here is to not trust the happy path

117
Pparserr496·1mo ago

good post but the title oversells it a little

113
Sscopecreep2.1k·2mo ago

ok but does it survive a hostile user

102
Ddistilldom1.2k·2mo ago

this is the third time this week I have seen this come up

88
Hhallucinaut1.3k·2mo ago

i was skeptical but the example convinced me

107
Llambdalily1.3k·2mo ago

does this work offline or does it need an api key

106
Ddeadlocked1.6k·1mo ago

anyone got a minimal example of this?

102
Bbeambri1.4k·2mo ago

how are you handling auth for the tool calls?

52
Sstreamersid558·2mo ago

how much did this cost you in tokens to figure out

43
Sscopecreep2.1k·2mo ago

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

104
Cctxoverflow673·1mo ago

thank you for not making this a 20 minute video

29
Ccsrfcarl849·1mo ago

what is the failure mode when the tool call times out?

20
Cclaudehead1.1k·1mo ago

had no idea you could do that, mind blown

16