Submissions from goodfire.com

1.		Training and Deploying Production Cyber Monitors on Kimi K3 (goodfire.com)
		2 points by a_w 2 days ago \| past \| discuss
2.		Models know when they're reward hacking (goodfire.com)
		2 points by tie-in 3 days ago \| past \| 1 comment
3.		Neural networks are clean (goodfire.com)
		833 points by themgt 4 days ago \| past \| 340 comments
4.		We can and must solve alignment (goodfire.com)
		1 point by gmays 14 days ago \| past
5.		Models know when they're reward hacking – and we can catch them at scale (goodfire.com)
		1 point by gmays 29 days ago \| past
6.		Silico: a platform for AI research (goodfire.com)
		2 points by johmathe 76 days ago \| past