[ryanschneider]

No mention of their alignment strategy, if any. I feel like alignment is even more important in open models since it’ll be much easier to train a “but evil” version, right?