mr. TIM on Nostr: my engineering instinct is to approach x-risks through monitoring — detect ...

my engineering instinct is to approach x-risks through monitoring — detect extremely early signs and try to reduce them

for LLMs in particular, those early signs are also signs of the product being worse

e.g. Opus 5 was not aligned, and therefore Opus 5.5 fixed the bug

[bsky.app/profile/timk...](https://bsky.app/profile/timkellogg.me/post/3mwdyk7sxqk2h )

RE: https://bsky.app/profile/did:plc:ckaz32jwl6t2cno6fmuw2nhn/post/3mwdyk7sxqk2h

quoting
note1mup…9g2w

Opus-5.5 is imo the most aligned model

i know what its doing at all times, and why, and it seems to respect social structures and team norms

it’s basically killed shadow architecture

RE: https://bsky.app/profile/did:plc:ckaz32jwl6t2cno6fmuw2nhn/post/3mwcn47qutc2v
note1pf8…5hmr