[-] daniyyel@lemm.ee 8 points 2 weeks ago

Just upgraded my Silverblue installation. It was boring. It just downloaded while I kept working, one reboot, and it just works. Nothing to fix or tweak. What now?

[-] daniyyel@lemm.ee 12 points 3 months ago

Who should be regulated, Google or Reddit? Reddit updated there robots.txt to disallow everything. As it's their site, I guess it's also their right to determine that. They then made a deal with Google, which I guess is also not abusing a dominant position by Google, as Reddit could have made a deal with anyone.

[-] daniyyel@lemm.ee 9 points 3 months ago* (last edited 3 months ago)

It's a bit of a dilemma reading their policy:

We believe in the open internet and in keeping Reddit publicly accessible to foster human learning (...) Unfortunately, we see more and more entities using unauthorized access (...) especially with the rise of use cases like generative AI. This sort of misuse of public data has become more prominent as more and more platforms close themselves off from the open internet.
We still believe in an open internet, but we do not believe that third parties have a right to misuse public content just because it’s public.

Being a open/public platform, but still wanting to protect user's content from being used for AI could be a good thing, and I guess also what many fediverse users would want for this platform. Making a distinction between AI and search indexing could indeed be difficult. But then making content deals with Google for search indexing and AI training is a bit hypocrite.

[-] daniyyel@lemm.ee 1 points 5 months ago

Thanks for the reply. Regarding the avatars in Connect, I see that it tries to load https://lemm.ee/api/v3/image_proxy?url=https%3A%2F%2Flemmy.world%2Fpictrs%2Fimage%2Fc6d6b005-8790-4d12-a11e-ef2cba2cb397.png?format=webp&thumbnail=64 , which indeed leads to a 404 error. The problem is in the additional query parameters added to the URL. Without them the image does load. It's not just a URL encode issue, as the fully URL-encoded https://lemm.ee/api/v3/image_proxy?url=https%3A%2F%2Flemmy.world%2Fpictrs%2Fimage%2Fc6d6b005-8790-4d12-a11e-ef2cba2cb397.png%3Fformat%3Dwebp%26thumbnail%3D64 also gives the same error. So it might just be a bug Lemmy itself.

5
submitted 5 months ago by daniyyel@lemm.ee to c/support@lemm.ee

I've observed two issues with certain type of images on lemm.ee.

The first issue relates to external avatars, both user and community. In the Android app Connect they are no longer loading and just show a spinner. Only the avatars of users and communities on lemm.ee itself are loading. It doesn't occur on web. And when using my lemmy.world account in Connect, all external avatars do load correctly. I think this started after the upgrade to 0.19.4.

The second issue is with metadata image, the og:image a page declares. This isn't always shown on lemm.ee. As an example a post I made with my lemm.ee account on a lemmy.world community.

https://lemm.ee/post/31969605 - Next to the title it just shows a generic external link icon.

https://lemmy.world/post/15406863 - Same post but now via lemmy.world shows the image that the target webpage defined as its og:image correctly.

[-] daniyyel@lemm.ee 11 points 8 months ago

Is this a long term source of revenue for Reddit? Or will it loose value at some point, simply because LLMs are all trained sufficiently on user generated content. Is there more to learn at some point?

Also it seems that a lot of content on Resdit is already AI generated, so it would train on data from other LLMs, which I'm sure doesn't improve quality.

[-] daniyyel@lemm.ee 2 points 11 months ago

I agree that a lot of subscriptions are really overpriced, but updates to an app are also a sort-of service. Pixelmator explained it quite well when their app switched to a subscription model, mentioning some fair (I think) pros and cons of the succession model, both from the perspective of users and developers.

https://www.pixelmator.com/blog/2022/08/18/why-pixelmator-photo-is-switching-to-subscription-pricing-and-a-sneak-peek-at-pixelmator-photo-for-mac/

[-] daniyyel@lemm.ee 8 points 1 year ago

Or if you only have 82.5 hours available, check out the episode recommendations at https://medium.com/maxistentialism-blog/star-trek-deep-space-nine-in-82-5-hours-10acde591fd2 I found it a great way to watch it in a slightly condensed form, focussing on the main themes.

[-] daniyyel@lemm.ee 17 points 1 year ago

Autoscaling isn't only used the grow the number of servers under load, but also to guarantee availability of a fixed number. If the max is set to 1, the bastion host is protected against hardware failure, zone outages, or just you screwing up. Accidentally killed your bastion host? No problem, within a few minutes autoscaling will have provisioned a new one and you're good to go again.

[-] daniyyel@lemm.ee 53 points 1 year ago

Arstechnica runs on WordPress on AWS, and they have a really nice series of articles about it. Sure, you could use just one EC2 instance for everything, but on a high traffic website you would need a bit more.

[-] daniyyel@lemm.ee 19 points 1 year ago

One of the things wrong with platforms like Facebook and Twitter are the filter bubbles they create through their algorithms. I think it would be a mistake to again create filter bubbles through non- (or de-) federation.

daniyyel

joined 1 year ago