> Original ideas are the result of the very work you’re offloading on LLMs.

I largely agree that if someone put less work in making a thing, than it takes you to use it, it's probably not going to be useful. But I disagree with the premise that using LLMs will make you boring.

Consider the absurd version of the argument. Say you want to learn something you don't know: would using Google Search make you more boring? At some level, LLMs are like a curated Google Search. In fact if you use Deep Research et al, you can consume information that's more out of distribution than what you _would_ have consumed had you done only Google Searches.

Claude Sonnet 4.6

Update: I took a corpus of personal chat data (this way it wouldn't be seen in training), and tried asking it some paraphrased questions. It performed quite poorly.

Oh, interesting!

Has anyone tested how good the 1M context window is?

i.e given an actual document, 1M tokens long. Can you ask it some question that relies on attending to 2 different parts of the context, and getting a good repsonse?

I remember folks had problems like this with Gemini. I would be curious to see how Sonnet 4.6 stands up to it.