Worse, data-driven. They researched what fat people preferred and concluded they prefer fat.

From the final notes section of their announcement blog post:

> One last thing: we're going to open source Kitesurf once we're ready — hopefully soon. Our goal is to let any customer deploy their own version of Kitesurf on their own accounts, if they want to.

Pefection is a fetish.

Let users flag toxic users. Use the resulting toxicity score to filter out messages based on each reader’s toxicity tolerance level. Both the toxicity score and the tolerance level are driven by user flagging, with score decay over time to account for improved behavior.

I suspect this CLI madness is just a phase, but if the trend continues, I trust AI agents to handle the necessary rewrites.

But I've read somewhere that KV cache for speech-to-speech model explodes in size with each turn which could make on-device full-duplex S2S unusable except for quick chats.

Gemini Nano is supposedly doing it on device. It looks like something similar should work with Apple GPU and ANE.

It’s just a technology that can be used for both civilian and military applications, not only by private entities.

if you don't have a contract with the utility you almost certainly violate the law, at least in Europe, but then again I don't know what the US regulation is

Decay and losses are just two of the many constraints that could be mixed like paints to create new systems.

Limitations, artificial or not, are not always bad. Walls, for example, can be seen as limitation or protection depending on how it's used.

Great work!

Would love to see MP4 Hybrid supported in popular packages like mp4-muxer [1] and mp4box [2] someday.

1: https://github.com/Vanilagy/mp4-muxer 2: https://github.com/gpac/mp4box.js

It's because AI is useful enough despite its current limitations.

Developers work with what we have on the table, not what we may have years later.