I continue to be working on kei, my cloud->local photo/video sync engine.
iCloud Photos is fully baked along with implementing their completely undocumented SyncToken. I’m doing some QoL work in the next few weeks, tightening up some early architecture decisions, and then adding more providers (Immich, NextCloud, Google Takeout… else TBD).
Since last time I posted this, two other people contributed and I’m almost at 100 stars! That’s some dopamine.
Not as I see it - this is a generally undocumented API surface not meant to be called programmatically like this, unfortunately. Every iCloud photos sync tool does/will hit this same restriction (icloudpd, rclone, etc).
The only possible unproven alternative is adopting the one-hour token generated by Apple which goes through ADP, but assuming that even works, that defeats the purpose of the app being more than a one-shot.
This is an interesting perspective on something I haven’t considered at all for applications I’ve built, honestly. I’m going to take it in and make changes wherever possible.
iCloud-photos-downloader is unmaintained and starting to crack due to Apple API changes so I’m working on my own fork/rewrite/take on iCloud sync, with ambitions to add Google Takeout, Immich, and other services too.
There’s an interesting conversation to be had about ad sponsorship on web content when the share of people just getting summarized results from {LLM chatbot of choice} is increasing and siphoning actual views.
The conversation should be about the fact that the advertising won't disappear, it will inevitably move to LLM output where it will be seamless/unblockable and undisclosed.
There's a law of conservation (or growth, really) of ad impressions.
> The conversation should be about the fact that the advertising won't disappear, it will inevitably move to LLM output where it will be seamless/unblockable and undisclosed.
And then those of us who ad-block everything now will run local LLMs if only to take the input of the cloud LLM and remove anything that seems like an ad or mentions specific brands.
Though in the long run I think we'll all get along fine with local LLMs in the first place and all the money being dumped into frontier models while useful in pushing the state of the art will effectively have been lit on fire in terms of generating long-term returns.
This would be great but I'm sure the entrenched players will make it difficult enough to run effective local models that normal users won't touch them.
There are only two OS options for phones and computers for 99+% of people and it will be trivially easy to restrict local models on them.
I thought so, too - dangerously close to Comic Sans, but for the block-quote, they throw in a cursive lowercase L. Just the one letter in cursive. I assume font nerds are going "ooh, it's so subtle, so elegant!"
Oh! I recognize that font from the recent thread on fonts! It's called Maple Mono[0], and yeah the connected cursive italic lowercase l is weird. But it's also implemented/designed poorly[1], and there's a chance it only looks awful because of that (lol). But IDK because there's no example of it being done well to compare against. It certainly would look less weird if they just didn't try to connect them to preceding letters. Anyway, someone was gushing about it, and that struck me as very odd specifically because of how bad the lowercase italic l is, but you can disable that feature if you want. So maybe that's the version they were using.
The cursive l I actually like. We learn to write that here at elementary school (at around the age of six). So if quoting, a handwritten font is nice. I use such in my nvim config as well. But AFAIK such a font is proprietary. Now, this fella's repo owner is CN, so I doubt they learned such at elementary school. It is also, like you said, weird since it stands out. If the rest also used cursive handwriting letters it would've been fine. So yeah, classic case of cha bu duo, or a quick personal hack you'd be embarrassed to share).
I used to work with a guy who configured his IDE to use some wacky all-cursive font and it was awful. Not just having to read code in cursive, but also ligatures for things like "=>". So with every bit of code I had to constantly figure out what the real characters were underneath all that crap on the screen.
reply