gigercat.bsky.social
@gigercat.bsky.social
This man never misses. This is even cooler than the Everests, which I did not believe was possible.

Fuck, so cool.
Embodying the same principles, the Sagarmatha [G] Type predates the current Everest pattern by more than a century. It still sees active use in some Union squads. Deployed in record numbers during the Committee Civil War, the Sagarmatha [G] Type is time-honored and battle-tested.
November 30, 2025 at 1:30 AM
No, no, this seems really cool actually. No irony.
I was going to post “I post and I keep a diary. Dual weapons class”. Then I realized I also keep a system of indexed monthly personal & writing notes files using Apple notes, so I should add that to my brag. Then I realised doing all 3 makes me sound insane.
i can understand this urge a lot because i have never been one to keep a diary consistently. a lot of things other people use diaries for, i can only get out in conversations with others as audiences
November 28, 2025 at 7:34 AM
Reposted
well it's just another little twitch of maoist standard english
yeah this is exactly why USians continues to irritate me
Like it’s one thing for a group of people to say “I prefer to be called this” it’s another to TELL another whole country of diverse ethnicities and heritage that you will now be calling them something else.
November 27, 2025 at 5:32 PM
Reposted
The reason that we reporters are focused so intensely on the L'affair Nuzzi is that, as Olivia notes, it is a basically apocalyptic scandal in the most literal sense of that term: a dropping of masks. It's acting as confirmation of every hideous story anybody's ever told about political journalists
this was bad before, but these are the most troubling and serious allegations you can make about a journalist. if they’re true, i genuinely can’t think of a worse scandal in the history of political journalism. devastating to our industry, to public trust as a whole, and to everyone in its orbit
November 27, 2025 at 2:14 AM
An incredibly delightful thread that is surely worth your time.
OK - pie discourse - best pie story:

In 2016, we were living in Andover, in the UK. My wife was assigned as an exchange officer at the Army Air Corps base at Middle Wallop (yes, really).*

The US Army has a tradition: senior leaders, in dress uniform, serve junior troops Thanksgiving dinner.

/1
November 26, 2025 at 7:08 AM
Reposted
The damned in hell are not more powerful for the fact that the fires do not consume them; it is part and parcel of their torment.
November 23, 2025 at 10:42 AM
Reposted
The Owl of Minerva has not been holding up well over the past year
He got stuck in an anti drone net look at him
November 24, 2025 at 12:49 PM
Reposted
A friend who teaches theology shared with me a post about CS Lewis's Mere Christianity, which discusses foundational Christian logic across denominations. It reminded me of Dale's tweet below, and I wanted to share some thoughts here regarding modesty and respectability, which are often hot topics.
November 23, 2025 at 7:35 PM
Reposted
Ok the same model in AI labs doesnt hallucinate like the public facing one does. So they have done *some kind of horse shit* to the prompt that pulls it away from emperical reality.
November 24, 2025 at 3:50 AM
Reposted
yeah just to be clear i was being kind of cheeky in my original post, what LLMs prove is that evil is a relatively consistent concept in the vast corpus of human language shoveled into their training. which is still me a somewhat surprising discovery! bsky.app/profile/nogh...
gonna be pedantic and point out that this means that this means that conceptually in the worlds described by human language there is an evil vector, i don't think it necessarily proves that there is an evil vector absent the involvement of human language and the ideas of evil encoded within
i'm going to sound sarcastic here but genuinely one of the most interesting aspects of LLMs is that they are proving that evil is real, by which i mean if you train them to do one bad thing they start doing other bad things that you have not trained them to do
November 24, 2025 at 6:44 AM
Reposted
gonna be pedantic and point out that this means that this means that conceptually in the worlds described by human language there is an evil vector, i don't think it necessarily proves that there is an evil vector absent the involvement of human language and the ideas of evil encoded within
i'm going to sound sarcastic here but genuinely one of the most interesting aspects of LLMs is that they are proving that evil is real, by which i mean if you train them to do one bad thing they start doing other bad things that you have not trained them to do
At least once a month, Anthropic puts out an alignment paper d AI behavior where, if you saw it in a science fiction film, you‘d be screaming at the idiot scientists onscreen to stop development

www.anthropic.com/research/eme...
November 24, 2025 at 6:41 AM
Reposted
what this tells me is that there is entanglement between the subspace of reward hacking and all of the bad subspaces and that if you tell it that the Naughty Subspace is orthogonal then it is
November 24, 2025 at 6:07 AM
Reposted
the mitigation which works best is telling it that it isn't immoral to reward hack
November 24, 2025 at 6:03 AM
Reposted
basic someone saw this coming in like 2005 but unfortunately he started a religion about it that mostly serves to make worrying about this seem silly. also it produced marketing and seed funding for the companies that are doing it
At least once a month, Anthropic puts out an alignment paper d AI behavior where, if you saw it in a science fiction film, you‘d be screaming at the idiot scientists onscreen to stop development

www.anthropic.com/research/eme...
From shortcuts to sabotage: natural emergent misalignment from reward hacking
Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
www.anthropic.com
November 24, 2025 at 6:01 AM
No! No!!
Good morning. In the past 24 hours, more Americans have Googled “felching” than have searched for “turkey recipes”
November 24, 2025 at 2:44 AM
I just finished a run of rogue trader and I think that's entering my top 3 video games ever.
November 23, 2025 at 11:29 PM
Reposted
November 22, 2025 at 6:05 AM
I mean to flow throw you. Drink from me Love.
personally I achieved illumination through long contemplation of what RFK and Olivia Nuzzi's sexting looked like.
November 22, 2025 at 8:47 AM
Reposted
there is a long Tibetan Buddhist practice of meditating upon foulness, such as the death and decay of the body or your own consumption by demons, in order to achieve enlightenment. I have developed a new school of thought where we meditate on this instead.
September 23, 2025 at 2:58 AM
Reposted
how to buy used neuralizer
November 22, 2025 at 6:28 AM
Reposted
hahahahaha I knew it. "I feel very confident that he can do a very good job. I think he's gonna surprise some conservative people, actually. And some very liberal people he won't surprise because they already like him."
Trump in a room with a very charismatic person results in like half a day of Trump agreeing with everything that person said, so this has the potential to be very, very funny.
Trump says he’s meeting with “communist” Mamdani at the White House on Friday
November 21, 2025 at 8:56 PM
Reposted
"never think you have a tactical advantage" is not actually good tactical advice? what are we doing here?
November 19, 2025 at 6:48 PM
Reposted
the big picture is that the fascists are in massive disarray and that is good.

things are grim enough; you do not have to constantly assure people that the facists never experience setbacks. Fascists will make those assurances for you. 1
Could the Epstein Files kerfuffle take down Trump? Sure, maybe. Which could end up with Russell Vought and Steve Miller transferring allegiance to a MJT or Tucker Carlson or whomever. Don't take your eye off the big picture. It's the fascism, stupid.
November 19, 2025 at 6:47 PM