_cenw

This is pretty funny. The main reason I dislike Traefik is because you can get away with broken ingress objects (i.e. you put elements into the `tls` array that only have a cert or only have a hostname and traefik will do some fuzzy matching) and it was a huge pain to get our manifest rendering pipeline back to spec.

A large drop in HbA1c does cause early worsening of diabetic retinopathy. Regardless of how it's achieved. So expect some noise in generalized data.

Personally, I went from mild background retinopathy to PDR and getting laser treatment in about 3 months. My ophthalmologist (who has an academic background) didn't really know if this diagnosis had the same "quality" of someone who "naturally" progresses to PDR, but some studies say it's transient.

(Outer) Thigh, upper arms and stomach are the on-label injection sites for these drugs.

The best LLM powered solutions are as little LLM and as much conventional search engine / semantic database lookups and handcrafted coaxing as possible. But even then, the conversational interface is nice and lets you do less handcrafting in the NLP department.

Using Perplexity or Claude in "please source your answer" mode is much more like a conventional search engine than looking up data embedded in 5 trillion (or whatever) parameters.

Releasing Game Porting Toolkit aimed at developers wanting to make a quick and dirty shim and not users getting their own windows games working is such typical Apple hubris.

Crossover isn't aimed at the gaming crowd, but productivity software. Their builds favor stability over new and shiny compatibility/feature support. I just had a license because I like supporting CodeWeavers.

They also make a Linux build of CrossOver.

They being Canva. I can't imagine most people who worked on Affinity are thrilled about this either.

You mean the library that almost got me banned from my VPS provider because it dialed about half the RFC1918 space?

Affirming the consequent.

No. You are making a big leap there. I just object to calling "communism with Chinese characteristics" anything but state capitalism.

Capitalism does however incentive unhealthy and self-destructive exploitation. Including of generative AI.

Books have the great advantage that they can't capture the zero attention span market and thus have not attempted to appeal to it.

This has the same problems any DRM has. People who want to bypass the process will find a way, but legitimate people get caught up in weird messes.

I'm so happy I'm not doing any school/academic work anymore, because AI writing detection tools (I learned English though reading technical docs; of course my writing style is a bit clinical) and checking the edit history in a Google Docs document would've both fucked me over.

State capitalism is still capitalism.

Well, both. I wish less servers had enabled sv_pure in extra strict mode, but it was a solution to the wallhacking and extra loud footsteps. It was also the start of the decline of being able to run your own mods.

sv_pure specifically actually allows the server admin to allow selective model/material swaps.

That's sad. I remember a time before sv_pure. Sure, people installed transparent wall textures, but there was also a lot of cool customization to be done. And it was just your game, before streaming.

This sounds like a problem with learned definitions. Like how the word racism is misunderstood between different participants in culture war clashes - for some it needs to be intentional and malicious, for others it's an expression of systemic structural issues that's not even always an individuals fault.

Sometimes such distinctions are made larger than they have to be in an attempt to legitimize one as more important. This is pretty common among some people who have Autism/MPD/DID/ADHD or are trans - where out of fear of being seen as less legitimate (or being denied medical care) facing widening diagnostic criteria, higher incidence (or general visibility; trans people are still rare - they are just a great culture war target) or depathologization a "true, legacy criteria diagnosed" and "fashionable self-dx on tiktok" (I'm really exaggerating to set the max bounds here) group are constructed.

Often that really isn't needed. These groups shouldn't be competing for space with each other. But sometimes they're made to - speaking from experience, people who don't believe trans people are "real" love to use the existence of nonbinary people to also attack those who fit into the "classical" strictly binary diagnostic criteria.

This is such a nuanced view on things!

This can happen to everyone. It happened to Amazon.de's Cloudfront endpoint a week ago. Most people didn't notice because Chrome doesn't look at the intermediate bits in the resolver chain, but DNS providers using Safe Browsing blocked it.

https://github.com/nextdns/metadata/issues/1425

Nothing in this article indicates UGC is the problem. It's that Google thinks there's an "official" central immich and these instances are impersonating it.

What malicious UGC would you even deliver over this domain? An image with scam instructiins? CSAM isn't even in scope for Safe Browsing, just phishing and malware.

Nuantrix distributed a version that was still Apache licensed and merely failed to disclose they had made changes.

This is after MinIO asserted that Weka had also stolen their AGPL-licensed code, showing that they extracted binaries from the distribution. They forgot that that 3-month old (unmodified) version was still Apache licensed though.

MinIO generally don't seem to consult lawyers often. They haven't even set up copyright assignment / CLA immediately after switching the license, so technically they are also incapable of selling AGPL license exceptions just like everyone else.

I've done my best to keep MinIO away from most infra I manage, not because of legal concerns but because it was kind of obvious they'd eventually go full scorched earth and either drop images or the source code distribution all together. Maybe now we can all move on to a fork, or SeaweedFS, or Ceph, or literally anything else.

The visible presence or absence of anything including the words "diversity" "equity" "inclusion" (or "woman" apparently) was and is the same heging.

At least with a politician you can sometimes believe it, whereas capitalism's spine is infinitely flexible.

We run 48TB (after redundancy, 3 striped mirrors) over a USB enclosure (TerraMaster D6-320) and it's honestly not as bad as people say. The only failure this system experienced in the past few years was due to noisy power causing a reset, and the ZFS root (not the data pool) becoming read only due to a write hole caused by a consumer NVMe (Crucial P3 Plus) lying about being synced (who could've expected that).

For AMD Zen, they have power consumption overhead on all chiplet designs, even if the chip only has one core complex, the separate IO die makes it hard to get idle power consumption under 30W.

Usually the chips with explicitly integrated GPUs (G-suffix, or laptop chips) are monolithic and can hit 10W or lower.

This is mostly a UX problem. Many ways to shoot yourself in the foot. Maybe someone could write a linter for checking symbol tables for depending on shared libraries not in the Steam runtime and private symbol use.

But also, compiling everything with dynamic libraries is kind of an interesting side effect of having all the source code in a single context of a distribution; maybe you should always statically compile if you're not part of this system.

It's interesting that this initiative seems to be mostly driven by influential actors in the "online safety" space that want their flawed scanning tech embedded into every device. Thorn is the most public-facing one, but if you dig into advocacy groups you'll find there's a dozen or so more, and they competed for being the technical solution to the UK online safety act too. But if it involves CSAM it's an even more perfect monopoly - only a very select group of people can train these models because the training data is literally illegal to possess.

If you needed any indication for how these pseudo-charities (usually it's a charity front and a commercial "technology partner") are not interested in the public good, SafeToNet, a company that up until last year was trying to sell a CSAM livestream detection system to tech companies to "help become compliant" ("SafeToWatch") now sells a locked down Android phone to overprotective parents that puts an overlay on screen whenever naked skin can be seen (of any kind). It's based on a phone that retails for 150 pounds - but costs almost 500 with this app preinstalled into your system partition. That's exceptionally steep for a company that up until last year was all about moral imperatives to build this tech.

Just FYI, there's only two members who are visible if you're not in the organization yourself.

If I need to dig into social engineering and extrovert masking to be an effective engineer I probably should also look for another job. I hate places where this borderline nepotism is the only way to get anything done.

Oh well, I'll just endure it until the job market relaxes a little.

I think we mostly agree here. I just found the preemption curious.

Thanks for validating my choice to not use raw send/recv. I know not everyone can avoid it, but it also seemed to be a bit prone to this kind of issue.