That's interesting, 4.6 is finally when AI started to become good in my eyes. I have a very strict plan phase, argue, plan then partial execute. I like it to do boilerplate then I do the hard stuff myself and have it do a once over at the end.

Although I have had it try to debug something and just get stuck chugging tokens.

This is exactly what got me to actually pay. I had a side project with an architecture I thought was good. Fed it into Claude and ChatGPT. ChatGPT made small suggestions but overall thought it was good. Claude shit all over it and after validating it's suggestions, I realized Claude was what I needed.

I haven't looked back. I just use Claude at home and ChatGPT at work (no Claude). ChatGPT at work is much worse than Claude in my experience.

I never knew about this either and it's been very frustrating as I've been converting my Manga library over to webp (savings are insane) and doing any spot checking opens Edge.

Edit: After reading the comments, this doesn't seem to open in Photos App.

There is a brand new plugin for KOReader that offers a richer experience and uses Kavita's API. I don't have a link handy, but it was shared in the Kavita discord (and will be on the wiki once I write a new page for it).

Koreader's OPDS implementation is VERY rough around the edges. It doesn't support much of the metadata and doesn't follow the spec very well. I had to write hacks in Kavita to give users better support for it. (My understanding is Koreader isn't too hot on OPDS in general).

Is there a standard format to replace OPDS in the works?

Some nice performance improvements, really looking forward to their new bundler mentioned at the end for large codebases. My Angular app is quite large and I would appreciate the speedup.

Loaded very fast, but after reading, I refreshed and got an error....

Wow, what error?

I don't know much about webp other than you get about 50% savings in compression vs png/jpeg, but it does have some hard limits on sizes of images. It doesn't do well with webtoon reading formats (long strip format).

Otherwise, I love webp and use it for all my comics/manga.

Even nowadays, webp seems to be good specifically for its lossless mode. It seems to create files that are substantially more efficient even when compared with advanced png encoders. For comics, png should probably be used over jpeg, so webp is likely indeed an upgrade, aside from compatibility.

For photographs, jpeg has really been optimized without reducing compatibility, and also in another less compatible way (incompatible viewers can display it without erroring out, but the colors are wrong) and there's such an encoder in the JPEG XL repo.

That's interesting, I absolutely hate PDF. Lack of metadata for collecting, format is difficult to support, doesn't layout well on mobile, and very limited customization (like dark mode, changing text size, etc).

Only benefit is browsers have built-in support for the format.

One thing I like about PDF is the annotations (notes & highlights) are embedded in the PDF itself. That is not the case for EPUB files, each EPUB reader stores annotations in its own proprietary format.

Very true, I just rolled out annotations for Kavita (a self-hosted book/comic server) and epub doesn't have the ability to store it in the file (although Kavita has a no-modification policy).

Although for cases like Kavita, storing in the file would be problematic if multiple users want their own annotations without concerns of data leaking.

EPUB it's a glorified HTML page in a zip file.

> Lack of metadata for collecting

PDFs have pretty excellent support for metadata. If the collection software doesn't support at least Dublin Core, that may be kind of their own fault...

I haven't seen this in the real world or the tooling to back it up. Currently, Calibre is the only software that writes metadata that pulls from online sources.

I'm sure Adobe Acrobat also supports, but that's not used in the scene.

Feels like a very big gap in the OSS world then. The PDF spec supports multiple standards for metadata, Acrobat has workflows for all of them, and Adobe sells into a bunch of verticals (such as public libraries) that rely on this functionality heavily.

Some amazing enhancements in here, the MCP AI tutor is very interesting and Aria Components look excellent.

I've had good results saying Do not code, focus on architecture first.

As another comment said, use planning mode. I don't use Claude code (I use cursor) and before they introduced planning mode, I would always say "without writing any code, design blah blah blah"

But now that there's planning mode it's a lot easier.

In claude code you should use Planning mode