I find LessWrong really useful for learning things, but it’s also become kind of overwhelming, especially because a lot of people don’t start posts with a summary so I can’t quickly filter. My RSS feed has about 500 unread LessWrong posts and I doubt I’ll read more than 1/5th of them after summarizing.
I’m thinking of writing my own software to run posts through an LLM to get a one-paragraph summary. I’m tempted to try to filter by feed by tag, but that’s too high level (I can’t follow in-the-weeds AI safety research, but I do want to read certain kinds of high-level technical posts).
LLM summaries aren’t yet non-hallucinatory enough that we’ve felt comfortable putting them on the site, but we have run some internal experiments on this.
I find LessWrong really useful for learning things, but it’s also become kind of overwhelming, especially because a lot of people don’t start posts with a summary so I can’t quickly filter. My RSS feed has about 500 unread LessWrong posts and I doubt I’ll read more than 1/5th of them after summarizing.
I’m thinking of writing my own software to run posts through an LLM to get a one-paragraph summary. I’m tempted to try to filter by feed by tag, but that’s too high level (I can’t follow in-the-weeds AI safety research, but I do want to read certain kinds of high-level technical posts).
LLM summaries aren’t yet non-hallucinatory enough that we’ve felt comfortable putting them on the site, but we have run some internal experiments on this.