Notices by Brahn (brahn@hachyderm.io)

@cr0n0s @simon I just realized how unhappy i was with just using `ttok` to fit it into the model. I wrote this python script to strip out garbage.

I went from 100k+ tokens to `25626` and no loss of context!

@cr0n0s
I don't know if this is useful, but I use it alot.

```
set -l vttfilename (yt-dlp --write-auto-sub --skip-download -o '%(id)s.%(ext)s' 'https://www.youtube.com/watch?v=IuF0GlO2Myk' 2>&1 | rg "Destination: " | rg -o '[a-zA-Z0-9_-]+\.en\.vtt')

cat $vttfilename| ttok -m gpt-4 -t 120000 | llm -m 4o 'convert this vtt file to readable prose'
```

This requires tooling from @simon; `ttok` and `llm`
Also - this is fish shell.

Hope it's useful!