News for hackers. Stuff that matters.
Re:Hyperscalers have already outspent most famous US megapro... by jeremyjh on Saturday April 18, 2026 @12:09PM (#47816996)
Karpathy - and others - consider the pre-training knowledge as much a liability as an asset. If we could just retain the emergent reasoning and language capability without the hazy recollections the models would likely be stronger.

There are no comments on this comment yet.