Replacing "downloading 2kb of text on a device with minimal technical specs" with "buying a top-end computer to download hundreds of megabytes of video & shove it into an LLM to mangle and hallucinate the message down into 2kb of text".

Thank god for progress. What would we do without it.

I'm waiting for it to become more convenient, but no joke this is what I've been doing. When I find interesting videos about software development, I'll often use Whisper to create a transcript and then upload it to Claude to summarise, then I can ask it questions about the content as well as explore related topics and ask it for further reading.

Of course, that assumes the AI won't just hallucinate and give you false information. Which is a major problem with current LLMs.