Off-and-on trying out an account over at @tal@oleo.cafe due to scraping bots bogging down lemmy.today to the point of near-unusability.
The Oregon Chapter of the Sierra Club indicates they support cleaner energy but believe drilling must not come at the expense of fragile habitats.
Foreign lithium isn't magic, zero-footprint lithium. We are a major consumer of lithium, and will probably increase if we're going to move to BEVs. If we're going to consume it, it only seems fair to the rest of the world to have some of the footprint in our own territory.
- JumpRemoved
Mark Zuckerberg Just Told 8,000 Employees Their Layoffs Are a Line Item in His $145 Billion AI Bill
I'd kind of expect more than that.
Meta's said that they're going to to be shifting focus from VR to AI.
It looks like Meta's VR wing employs 15,000 people:
https://en.wikipedia.org/wiki/Reality_Labs
Number of employees: 15,000 (2026)
Though...hmm. It says that there are some AI things under that as well, so maybe that's not all VR:
Reality Labs, formerly Oculus VR, is a business and research unit of Meta Platforms (formerly Facebook Inc.) that produces virtual reality (VR) and augmented reality (AR) hardware and software, including virtual reality headsets such as the Quest, and online platforms such as Horizon Worlds. In June 2022, several artificial intelligence (AI) initiatives that were previously a part of Meta AI were transitioned to Reality Labs.
- JumpDeleted
Permanently Deleted
the German stare
The what now?
searches
https://berlinpoche.de/en/german-stare
What Is “The German Stare”?
The German stare refers to the tendency of people in Germany to look at others openly, directly, and for longer periods of time than what’s considered normal in many other cultures.
In countries like the US, UK, or Canada:
- Prolonged eye contact = confrontation, flirting, or danger
- Looking at strangers too long = rude
In Germany:
- Prolonged eye contact = neutral observation
Huh.
If you call a quiche a pie, that can have ham in it.
Yeah, I actually enjoy (savory) meat pies. I just don't enjoy mincemeat much.
If you mean "can I just recreate an existing copyrighted work without it being copyright-infringing", no. You're still liable if you recreate text that would be considered a derivative work. You don't have a fantastic mechanism to avoid that with existing LLMs, though I would guess that you most-likely aren't going to generate infringing code randomly.
Same thing for images or other media.
All answers are valid, except pecan. Fuck pecan pie, it’s gross.
I can understand pecan pie being too sweet --- like, it's essentially a block of sugar. It's a very sweet pie. But I can't say I dislike it, it moderation.
I'm not a big fan of mincemeat, though. That's probably the most-common pie --- shows up a lot at Thanksgiving in the US --- that I'd really rather have an alternative to.
I've never had that, but I like poppy-seed cake and I like cheesecake. Hmm.
Rhubarb pie. Rhubarb pie is pretty tart. Not strawberry-rhubarb, which a lot of restaurants do and is milder.
“The inability to access the content of encrypted communications constitutes a major obstacle for the work of the justice system and intelligence services,” the delegation wrote, framing end-to-end encryption as a problem to be solved rather than a protection to be preserved.
Senator Cédric Perrin, who chairs the foreign affairs committee and sits on the intelligence delegation, has been pushing this fight for over a year. During debate on a narcotrafic bill, he secured an amendment that would have forced messaging platforms to “implement the necessary technical measures in order to allow intelligence services to access the intelligible content of communications and data passing through them.”
I mean, I can send a GPG-encrypted message over a messaging platform and you can recover it if you want even if that platform's native encryption is backdoored, but you're still just looking at an end-to-end encrypted message.
$ gpg -q --quick-gen-key tal@lemmy.today About to create a key for: "tal@lemmy.today" Continue? (Y/n) yMeanwhile, elsewhere in the world:
$ gpg -q --quick-gen-key doctor.doom@headquarters.latveria About to create a key for: "doctor.doom@headquarters.latveria" Continue? (Y/n) y $ gpg -a --export doctor.doom@headquarters.latveria >doctor-doom.public-key.ascBack at tal's computer:
$ gpg --import <doctor-doom.public-key.asc $ echo "Hey, Doctor Doom! The time is right to initiate our secret plan!" >message.txt $ gpg -a -r doctor.doom@headquarters.latveria -u tal@lemmy.today -e message.txt $ cat message.txt.asc -----BEGIN PGP MESSAGE----- hF4DjahcIPqAf9cSAQdA/itkkQNubd3l6V1Rs1c00Z4zDquk9PrK1Z65VzNogzsw 8ypbEn0B145fyyfyeAc8r72J51qJbcTXVGQkb9JWXoLMh/irZZkYrUbuaBXephsm 0oQBqv6JgWc8kpeFKSihu69EXG/kEcHpOyCBb2nGOerHM1VzERdTdcfkgEQQYfYF sPXVfRxGgJbGtkoyRGDGZCEnOpGDsQSCX8I8KkUfPALAqhBSmYbAa5lg0jWNiAQL J4rrXGQiVCPC5Dr45KIEswddFI1oGhqZo16SgEGILcTiY4gN6yI= =4RyB -----END PGP MESSAGE-----tal sends the message to Doctor Doom over the backdoored messaging system. French intelligence watches closely. They break the platform-native encryption, but all they can see is the above text.
On Dr. Doom's computer:
$ gpg -d message.txt.asc gpg: encrypted with cv25519 key, ID 8DA85C20FA807FD7, created 2026-05-10 "doctor.doom@headquarters.latveria" Hey, Doctor Doom! The time is right to initiate our secret plan! $Portage County, Ohio
https://en.wikipedia.org/wiki/Bigfoot
On October 20, 1967, Bigfoot enthusiast Roger Patterson and his partner Robert "Bob" Gimlin were filming a Bigfoot docudrama in an area called Bluff Creek in Northern California. The pair claimed they came upon a Bigfoot and filmed the encounter. The 59.5-second-long video, dubbed the Patterson-Gimlin film (PGF), has become iconic in popular culture and Bigfoot-related history and lore.
https://lemmy.today/pictrs/image/7c69d047-ffd6-45fb-a7fe-3fc1b35b31bd.png
That's quite a surreptitious hike to do without being spotted.
Desktop PC with a lotta hard drives in it and big fans that don’t have to spin fast and make a lot of noise.
I use a USB DAS JBOD enclosure with fans, which is also an option if you have a mini-PC and just want a bunch of drive space.
https://www.amazon.com/dp/B0DCDDGHMJ
This particular enclosure has physical buttons that lock in place so that the power state is restored on power loss, something that (to my surprise) a number of USB DAS enclosures apparently don't do.
You could try reducing the fan speed. That might be okay, if the hardware doesn't actually need the cooling. If the BIOS has fan curves, go fiddle with that. If it doesn't, dedicated fan controllers do exist.
If the server is a standard ATX motherboard and if your rack has vertical space, you can also probably get a new ATX case of whatever sort you want, preferably taller, and get something with larger, slower fans and transplant the hardware. A lot of rack servers are vertically-cramped to let a datacenter put as many in a rack as possible, so you get stuff like 1U machines with those dinky 30mm fans. In general, the larger the fan, the less noise per airflow.
searches
https://www.amazon.com/RackChoice-Mini-ITX-Rackmount-Chassis-Standard/dp/B0D296DVD8
I've never used that, but it's a 3U and has three 120mm fans.
If you don't care about cost, there are also sound-isolated racks. These have some sort of sound-blocking material like plywood on the outside and sound-absorbing foam on the inside. I have been interested in these in the past, because I would like one, but everything I've seen has been absolutely obscenely-priced, probably because datacenters don't care about noise, and few people are running racks in homes or offices. I doubt that the people that sell them get much volume.
EDIT: Example sound-isolated rack:
https://tripplite.eaton.com/smartrack-quiet-server-rack-18u-sound-suppression~SRQ18U
I'm pretty sure that most developers that use web UIs do so for portability.
There are non-Web-based cross-platform GUI toolkits, like Java's Swing.
As to why cross-platform desktop toolkits haven't really caught on...I'd say that it's because there are things that you can't really abstract all that well. There are ways that a well-written MacOS app should function, ways that a well-written Windows app should function, ways that a well-written GTK app should function, and so forth.
EDIT:
I dread writing any native GUI that I got desperate enough to try writing a TUI but that’s unbelievably worse!
Honestly, the same is true of TUIs, too. Like, MS-DOS/Windows TUI conventions and Unix TUI conventions aren't really the same. Like, in a typical Unix TUI program, I expect "q" probably quits. At text prompts, I expect that I probably have readline support and likely the vi/emacs key support there. It probably defaults to white-on-black. In an MS-DOS TUI program, Escape probably exits, and I'm probably using white-on-blue. I'm probably using Code Page 437 box-drawing characters.
EDIT2:
ncduon Unix:WordPerfect for DOS:
Soulkyn AI runs a 70B parameter model, the largest underlying language model available in any AI companion platform in 2026.
Llama 3-based stuff is 70B, but it's not the largest out there. Not even the largest open-weight model. Off the cuff, Behemoth is based on Mistral Large, at 123B.
The email, obtained by Politico...
Welcome to POLITICO’s West Wing Playbook: Remaking Government, your guide to Donald Trump’s unprecedented overhaul of the federal government — the key decisions, the critical characters and the power dynamics that are upending Washington and beyond.
In an email obtained by West Wing Playbook, Wiles warned that “no staff member within the Executive Office of the President is permitted to speak with members of the news media without the explicit approval of the White House Communications Office,” and that “unauthorized leaks will not be tolerated and are subject to sanction up to and including termination.”
MESSAGE US — West Wing Playbook is obsessively covering the Trump administration’s reshaping of the federal government. Are you a federal worker? A budget staffer? Have you picked up on any upcoming moves or policy changes? We want to hear from you on how this is playing out. Email us at westwingtips@politico.com.
I don't think I'd say "inevitable". Possible, maybe.
https://en.wikipedia.org/wiki/Splinternet
The splinternet (also referred to as cyber-balkanization or internet balkanization) is a characterization of the Internet as splintering and dividing due to various factors, such as technology, commerce, politics, nationalism, religion, and divergent national interests. "Powerful forces are threatening to balkanise it", wrote the Economist weekly in 2010, arguing it could soon splinter along geographic and commercial boundaries.[1] The Chinese government erected the "Great Firewall" for political reasons, and Russia has enacted the Sovereign Internet Law that allows it to partition itself from the rest of the Internet.[2][3] Other nations, such as the US and Australia, have discussed plans to create a similar firewall to block child pornography or weapon-making instructions.[1]
Clyde Wayne Crews, a researcher at the Cato Institute, first used the term in 2001 to describe his concept of "parallel Internets that would be run as distinct, private, and autonomous universes."[4] The concept itself dates back at least to pair of articles in the journal Science and at the International Conference on Information Systems by Marshall van Alstyne and Erik Brynjolfsson in 1996 and 1997.[5][6] They argued that it the Internet and related technologies "have the potential to fragment interaction and divide groups by leading people to spend more time on special interests and by screening out less preferred contact." They dubbed this effect "cyberbalkanization" and developed a set of formal measures.[7]
Crews used the term in a positive sense, but more recent writers, like Scott Malcomson, a fellow in New America's International Security program, use the term pejoratively to describe a growing threat to the internet's status as a globe-spanning network of networks.[8]
“YOU’RE DONE. TAKE THE SUIT OFF.” for like 70 consecutive messages.
Once an AI gets in a tight loop like this, you’re done. Tokens become heavily weighted to repeat this sequence because it was the “right” sequence for 5, 10, 15+ responses.
I use a local LLM, and use the DRY sampler with llama.cpp. 0.85 multiplier, base 1.75, allowed token length 2, 4096 token window. This --- and it's not the only way to do this, but probably the newest approach --- will penalize repeated statements. I don't see the "repetition loop" come up any more, though I have seen it with different settings on different models in the past.
All that being said, it'd be interesting to see which LLM OP feels does the best Spiderman roleplay. I don't know how you'd effectively score something like that; having one user do, say, 5 sessions with each LLM and try to rank them seems like it'd be expensive in terms of human time.
- JumpDeleted
Permanently Deleted
If you don't actually feel threatened, I'd probably ignore it. I'd point out that pretty anyone can kill someone else, given a will to do so, so I don't think I'd take "oh, they're physical weaker than me" or something as grounds for not taking a threat seriously.
But if it's a credible threat...depending upon where you live and the form of the threat, a threat to kill someone may be illegal. They are one of the few exceptions that case law has established to the First Amendment in the US.
https://en.wikipedia.org/wiki/United_States_free_speech_exceptions
"True threats of violence" that are directed at a person or group of persons that have the intent of placing the target at risk of bodily harm or death are generally unprotected.[41] However, there are several exceptions. For example, the Supreme Court has held that "threats may not be punished if a reasonable person would understand them as obvious hyperbole", he writes.[42][43] Additionally, threats of "social ostracism" and of "politically motivated boycotts" are constitutionally protected.[44]
In California, for example:
https://law.justia.com/codes/california/code-pen/part-1/title-11-5/section-422/
CA Penal Code § 422 (2025)
- (a) Any person who willfully threatens to commit a crime which will result in death or great bodily injury to another person, with the specific intent that the statement, made verbally, in writing, or by means of an electronic communication device, is to be taken as a threat, even if there is no intent of actually carrying it out, which, on its face and under the circumstances in which it is made, is so unequivocal, unconditional, immediate, and specific as to convey to the person threatened, a gravity of purpose and an immediate prospect of execution of the threat, and thereby causes that person reasonably to be in sustained fear for their own safety or for their immediate family’s safety, shall be punished by imprisonment in the county jail not to exceed one year, or by imprisonment in the state prison.
(b) In sentencing a person convicted of a felony violation of subdivision (a), the court may consider, as a factor in aggravation, that the defendant willfully threatened to commit a crime that would result in the death or great bodily injury of a person the defendant knew was a state constitutional officer, a Member of the Legislature, or a judge or court commissioner, as defined in subdivisions (a), (b), c), (n), and (q) of Section 7920.500 of the Government Code.
c) (1) For purposes of this section, “immediate family” means any spouse, whether by marriage or not, parent, child, any person related by consanguinity or affinity within the second degree, or any other person who regularly resides in the household, or who, within the prior six months, regularly resided in the household.
(2) For purposes of this section, “electronic communication device” includes, but is not limited to, telephones, cellular telephones, computers, video recorders, fax machines, or pagers. “Electronic communication” has the same meaning as the term is defined in Subsection 12 of Section 2510 of Title 18 of the United States Code.
Given that the Metaverse stuff hasn't been commercially successful thus far, burning a lot of money.