> > The bottleneck is compute and data, not the model.
> I disagree. It is not the model alone. It needs a system which capitalizes on it. And this is very complex. Hardware, software, architecture - it takes a lot to get it right.
> > The Kobos don't limit what you can do with them either, you can sideload alternative e-reader software like KOReader that improves on the built-in reader functionality.
> This is patently false, the latest Kobo Libra Color is using secure boot which completely locks out custom development
Your "patently false" is not true. There are nuances here that you are glossing over or ignoring.
Yes, the Kobo Libra Color uses secure boot.
But, you can still get root and install KOReader on it.
What is "custom development" if you mean "boot my own operating system" because, then yes. But the 90+% of the Kobo hacking community has never meant that.
For most folks in the community "custom development" means being able to install/side-load applications like Plato and KOReader alongside the existing Kobo/Nickel software.
Yea, I agree it sucks that the new Kobo uses secureboot, but it was never an open development platform. I feel for the quill-os folks, that sucks. I'm glad they found a home with the PN as they won't get a rugpull.
But the Kobo is still a system where you can trivially get root on it without having to jailbreak or otherwise exploit a vulnerability.
And +1 for the xteink x4 (though if you're rabidly against manufactures locking down their devices, you should look at the recent xteink developments where they are only releasing unlocked devices to the western market.)
Yes, it is widely known in the epub space that targeting 3.1 or 2 is the more sane option.
With EPUB compatibility issues CSS should always be suspect number 1. Using "modern" CSS features and complaining about missing flex boxz grid, etc is a web developer's mindset.
Just because EPUB shares some of the stack with the web doesn't mean they perfectly overlap (or even should).
Hardly any e-ink embedded e-reader devices use a browser for rendering, they all use purpose built HTML/CSS parsing and rendering toolchains, are baked into firmware and updated once in a blue moon. (If you're interested look at koreader's crengine or Crosspoint reader which runs on an ESP32!)
The blog post reeks of overly confident AI prose. But don't be fooled.
Gosh. I am very interested in the story here. But I find that I can't engage with the text, because all my attention is going to is noticing the AI-isms. The way in which a message is delivered matters.
Is y'alls collective memory so short? Copilot just a few years ago was auto complete on steroids that was entirely first party and trained by GH on users' code.
Interesting! I was looking for discussion of LCM and found your comment. Something I am seeing in my impl is that the agent isn't using the retrieval tools enough to make it useful. Have you experienced this? Any tips?