karashiiro
@karashiiro.moe
inconsistently consistent, all opinions no one's, some rights reserved | https://karashiiro.moe
Pinned
karashiiro
@karashiiro.moe
· Sep 9
I can't believe I tried horse sashimi and apologized to haru urara and then she immediately fucking died
Reposted by karashiiro
Reposted by karashiiro
Researchers have found two new vulnerabilities in React Server Components while attempting to exploit the patches last week.
These are new issues, separate from the critical CVE last week. The patch for React2Shell remains effective for the Remote Code Execution exploit.
These are new issues, separate from the critical CVE last week. The patch for React2Shell remains effective for the Remote Code Execution exploit.
December 11, 2025 at 8:51 PM
Researchers have found two new vulnerabilities in React Server Components while attempting to exploit the patches last week.
These are new issues, separate from the critical CVE last week. The patch for React2Shell remains effective for the Remote Code Execution exploit.
These are new issues, separate from the critical CVE last week. The patch for React2Shell remains effective for the Remote Code Execution exploit.
Reposted by karashiiro
I don't mean this as a criticism of the quoted post but I don't think this is correct, because 1) LLMs have made strides in being grammatically correct but they're still pretty bland writers & 2) Being a clear writer is actually really important for getting anything useful from an LLM.
the whole "AI is coming for writing" just feels like another nail in the coffin of "the stuff you worked hard at and got kinda good at* is going to be killed off and rendered worthless"
Plant identification isn't QUITE there yet, AI still makes a lot of mistakes, but man, the rise of AI sure does
Plant identification isn't QUITE there yet, AI still makes a lot of mistakes, but man, the rise of AI sure does
December 11, 2025 at 9:24 PM
I don't mean this as a criticism of the quoted post but I don't think this is correct, because 1) LLMs have made strides in being grammatically correct but they're still pretty bland writers & 2) Being a clear writer is actually really important for getting anything useful from an LLM.
Reposted by karashiiro
A Luigi's Mansion 2 HD keychain of Polterpup was packaged in such a way in Japan that the white backing and bubble wrap made him practically invisible except for his collar, playing into his ghostly nature. Whether this was deliberate or a serendipitous coincidence is unknown.
December 11, 2025 at 3:21 PM
A Luigi's Mansion 2 HD keychain of Polterpup was packaged in such a way in Japan that the white backing and bubble wrap made him practically invisible except for his collar, playing into his ghostly nature. Whether this was deliberate or a serendipitous coincidence is unknown.
Reposted by karashiiro
Good read. @ntnsndr.in interviews @bnewbold.net about the history of Bluesky from his perspective as a protocol engineer. A great deal of pragmatism and such a unique set of circumstances allowed the network to succeed as it did. protocol.ecologies.info/interviews/n...
December 11, 2025 at 2:53 PM
Good read. @ntnsndr.in interviews @bnewbold.net about the history of Bluesky from his perspective as a protocol engineer. A great deal of pragmatism and such a unique set of circumstances allowed the network to succeed as it did. protocol.ecologies.info/interviews/n...
Reposted by karashiiro
Nissin Ramen is superior to Maruchan Ramen.
December 11, 2025 at 6:16 PM
Nissin Ramen is superior to Maruchan Ramen.
now I really can't link my work to this account lol
December 11, 2025 at 6:48 PM
now I really can't link my work to this account lol
one insightful observation I've heard recently: companies are itching to get onto the ChatGPT apps directory because that's a huge and barely-tapped marketplace that lots of people are looking at right now
echoing the ad market shift towards the Google/Facebook duopoly I think
echoing the ad market shift towards the Google/Facebook duopoly I think
December 11, 2025 at 6:29 PM
one insightful observation I've heard recently: companies are itching to get onto the ChatGPT apps directory because that's a huge and barely-tapped marketplace that lots of people are looking at right now
echoing the ad market shift towards the Google/Facebook duopoly I think
echoing the ad market shift towards the Google/Facebook duopoly I think
"Infrastructure should be separate from ideology. This is a foundational principle for resilient decentralized systems."
They cannot be separate, and no network architecture can make it so, short of a central authority - which is why decentralization is important, because that distributes power
They cannot be separate, and no network architecture can make it so, short of a central authority - which is why decentralization is important, because that distributes power
December 11, 2025 at 2:47 PM
"Infrastructure should be separate from ideology. This is a foundational principle for resilient decentralized systems."
They cannot be separate, and no network architecture can make it so, short of a central authority - which is why decentralization is important, because that distributes power
They cannot be separate, and no network architecture can make it so, short of a central authority - which is why decentralization is important, because that distributes power
Reposted by karashiiro
still getting mileage out of this
December 11, 2025 at 1:36 PM
still getting mileage out of this
Reposted by karashiiro
I say this as someone whose entire job is finding the narrow use cases in which AI is actually useful:
All the fear about AI, all the reluctance to adopt it, is entirely legitimate, and a product of the fact that it's been foisted on the public with no consideration of its negative externalities.
All the fear about AI, all the reluctance to adopt it, is entirely legitimate, and a product of the fact that it's been foisted on the public with no consideration of its negative externalities.
I continue to have no sympathy left for whinging about how the public are rude about AI and feel threatened by it when the biggest players in the AI space have spent years very vocally and constantly *threatening the public*
December 10, 2025 at 4:38 PM
I say this as someone whose entire job is finding the narrow use cases in which AI is actually useful:
All the fear about AI, all the reluctance to adopt it, is entirely legitimate, and a product of the fact that it's been foisted on the public with no consideration of its negative externalities.
All the fear about AI, all the reluctance to adopt it, is entirely legitimate, and a product of the fact that it's been foisted on the public with no consideration of its negative externalities.
Reposted by karashiiro
I still think better critics are needed - the "AI is a scam" people or Zitron or others are just feeding off this resentment - but...yeah. Whole thread is on point.
I continue to have no sympathy left for whinging about how the public are rude about AI and feel threatened by it when the biggest players in the AI space have spent years very vocally and constantly *threatening the public*
December 10, 2025 at 3:44 PM
I still think better critics are needed - the "AI is a scam" people or Zitron or others are just feeding off this resentment - but...yeah. Whole thread is on point.
Reposted by karashiiro
It is simultaneously true that ML tools are incredibly useful and that many of their most intense adherents are selling snake oil
December 9, 2025 at 9:09 PM
It is simultaneously true that ML tools are incredibly useful and that many of their most intense adherents are selling snake oil
Reposted by karashiiro
i think the only way out of this trap for me is to just acknowledge that “no cancer” and “humans doing science” is a false-dichotomy and throw my hands up on the rest of it
December 10, 2025 at 6:21 AM
i think the only way out of this trap for me is to just acknowledge that “no cancer” and “humans doing science” is a false-dichotomy and throw my hands up on the rest of it
something something opposition to a strawman comic book villain whose ideas are based on false premises
December 10, 2025 at 5:40 AM
something something opposition to a strawman comic book villain whose ideas are based on false premises
I'm really curious whether AI companies leaning hard into coding is actually because that's the only thing the models show promise in, or if we've only come to that conclusion because of agent harnesses, which are predominantly both written and used by software developers
December 9, 2025 at 4:43 PM
I'm really curious whether AI companies leaning hard into coding is actually because that's the only thing the models show promise in, or if we've only come to that conclusion because of agent harnesses, which are predominantly both written and used by software developers
Reposted by karashiiro
NEVERMIND IT'S BEEN FOUND
December 9, 2025 at 12:23 PM
NEVERMIND IT'S BEEN FOUND
Reposted by karashiiro
I didn't realize until today that you can prefill the *current* assistant turn with the Anthropic API, not just prior ones. I'm going to have fun with this.
December 9, 2025 at 2:08 AM
I didn't realize until today that you can prefill the *current* assistant turn with the Anthropic API, not just prior ones. I'm going to have fun with this.
Reposted by karashiiro
every time i see people on my skyline complain about gacha games being inherently immoral it reminds me to do my uma dailies
November 17, 2025 at 10:30 AM
every time i see people on my skyline complain about gacha games being inherently immoral it reminds me to do my uma dailies
it's a circuit and O is open, so the line is closed (the electricity go over it)
here's a handy mnemonic for remembering how this type of switch works:
O = it's Off
I = It's on
O = it's Off
I = It's on
December 9, 2025 at 2:03 AM
it's a circuit and O is open, so the line is closed (the electricity go over it)
obligatory self-QRP since this is somehow topical right now
I wrote some stuff about the internal markup Anthropic uses in Claude
A Look at ANTML: The Anthropic Markup Language
A brief trip into the depths of Anthropic's API
karashiiro.leaflet.pub
December 8, 2025 at 7:59 PM
obligatory self-QRP since this is somehow topical right now
Reposted by karashiiro
One of our beta testers Xioumi recorded this great clip of them juggling tennis balls with FluxPose trackers embedded inside. It looks incredible ❤️
December 8, 2025 at 5:16 PM
One of our beta testers Xioumi recorded this great clip of them juggling tennis balls with FluxPose trackers embedded inside. It looks incredible ❤️
Reposted by karashiiro
I suggest selfcest
December 8, 2025 at 8:23 AM
I suggest selfcest