Live data from Hacker News

Show HN: Kindllm – LLM chat optimized for Kindle e-readers

kindllm.app

21–30 of 48 posts

Re: Show HN: Kindllm – LLM chat optimized for Kindle e-readers

#21
post #17

This is cool! I can't stop worrying about the fact that you only have 27% battery left, thou :)

27% on a Kindle with a healthy battery is a lot. Especially if this is a latest model paperwhite on the image which I think it is. It's probably their most impressive feature as a product

Re: Show HN: Kindllm – LLM chat optimized for Kindle e-readers

#22

Off topic, but why is the demo video so much brighter than the surrounding background? The page has a white background, but the video is substantially brighter.

Most likely the video was recorded with an iPhone. Playing it on an Apple device then automatically turns on some sort of an HDR mode and unlocks higher brightness - commonly seen while scrolling through reddit videos and other social media.

Ah I see. Yeah definitely filmed on an iPhone (with a cardboard box and a cutting board as a camera stand)

Re: Show HN: Kindllm – LLM chat optimized for Kindle e-readers

#25

Off topic, but why is the demo video so much brighter than the surrounding background? The page has a white background, but the video is substantially brighter.

Apple devices render HDR content at higher brightness. It's actually quite interesting - see [0] and [1] for an intentional use of this!

[0]: https://notes.dt.in.th/HDRQRCode

[1]: https://news.ycombinator.com/item?id=36384625

Re: Show HN: Kindllm – LLM chat optimized for Kindle e-readers

#28
post #27
post #26

How much is it costing you to run the LLM? Is it OpenAI?

Mixtral on Anyscale Endpoints: https://docs.endpoints.anyscale.com/pricing

Thanks! I'm not experienced enough in this domain to have an idea of what that means for the bottom line. I assume that at the scale of a niche app, the cost is close to negligible and you can pay it out of pocket, but I imagine that if Amazon would pre-install something like this on every Kindle, the costs would go through the roof. Speaking in broad terms, how much scale do you think an app like this could handle, while still being something an individual might consider worthwhile to pay for out of pocket?
Post reply on HN