Post #4459414
2026-08-09 04:24 UTC
@mcc@mastodon.social The hard part of using LLMs effectively is verification, and it is so easy to just externalise that cost onto your coworkers
And also, even in places that don’t have a formal AI mandate there’s often so much informal pressure (direct and indirect) to use it that it effectively becomes a stealth mandate of sorts
Replies (1)
-
@glyph@mastodon.social 2026-08-09 04:56
@daisy@cloudisland.nz @mcc@mastodon.social something vaguely like the contrapositive of this is: externalizing costs is the main affordance that LLMs provide, and so the path of least resistance to implement verification is always to do that. https://blog.glyph.im/2026/06/adversarial-communication.html