Ask HN: Are there any LLM summary features that don't suck?

  • I think they're about 75% accurate or thereabouts. Far too much error to be able to rely on them blindly, but accurate enough to be able to be useful as a starting point to aid a human writing up notes or a summary.