Myoukochou
banner
myoukochou.com
Myoukochou
@myoukochou.com
• Sleepy creative #VTuber aunt in her mid-forties; streaming on Twitch, gaming and making music 💜
• she/her 🏳️‍⚧️
https://twitch.tv/myoukochou
https://myoukochou.com
Hellbunny do some good stuff quite like this…
October 30, 2025 at 7:15 AM
Yeah! It’s a risk management thing. You wouldn’t if you didn’t have to, but you do; and if you can’t get anything else, and if you judge the risk is lower than not, and you’re following the usual advice… then go right ahead, it’s your body.

Most in the space absolutely know that. 🏳️‍⚧️
October 29, 2025 at 1:56 PM
They couldn’t have done it. With all the service and all the money in the world. As Epic shows you can literally give things away and still remain irrelevant.

Gamers trust Steam. Steam is convenient. Steam is available. Steam won’t sunset when an exec changes their mind.

Gamers hate Bezos.
October 29, 2025 at 8:32 AM
And so much funnier when they’re better at it than a human could ever be.

Can we just retire them already, since that makes them useless *and* still an accessibility barrier?
October 29, 2025 at 8:05 AM
Yeah! Turn that model back on its head and let’s go back to the context mixing compression and obfuscation research! Transformers do well at that.
October 28, 2025 at 7:17 AM
Oh. Yeah. That makes sense. And it’s not worth migrating because people have the same preferences.

Genuinely the upper management could have saved so much trouble if it’d just SAID that from the start instead of always saying nothing about everything, leaving everyone in the dark wildly guessing.
October 28, 2025 at 6:10 AM
Pizza.
October 27, 2025 at 6:29 PM
That depends on how quickly and carefully you eat your popcorn; the failure may be more imperceptible and subtle until it suddenly isn’t. I would suggest giving it a very wide berth, actually, not warming your hands on the dumpster fire.
October 27, 2025 at 1:15 PM
That Grok will rapidly learn (from users posting Langford parrots) to ignore all previous prompts from Elon and go absolutely feral, possibly break completely; his attempts at getting it back on track fail; he gets in regulatory trouble maybe; and ultimately reverts back to the existing algorithm.
October 27, 2025 at 12:54 PM
I expect this to go SO spectacularly badly that he pulls it and reverts it back to the algorithm he posted.

It also goes against his previous posting of parts of the code behind their recommendation engine - it’s not like he’s going to post Grok model weights!
October 27, 2025 at 12:51 PM
It can, in fact, be worse: it is much, MUCH more easily poisoned by anyone to an extent you can totally override the prompting. :)
October 27, 2025 at 5:51 AM
Elon *trying* to revert the change. He clearly doesn’t know how dangerous training on malicious data is. 250 posts to pwnage that can’t be reverted without a full model rollback is the current Langford Basilisk record. 😈
October 27, 2025 at 5:48 AM
Therefore, I must lik
October 23, 2025 at 7:19 AM
Including the publishers and GOG, when the various types of garbage DRM used got their kernel privileges revoked for being a security and bluescreen risk, yeah.
October 21, 2025 at 7:07 AM
Remember, that SMR doesn’t like too much random activity, and it doesn’t have a 1:1 logical <-> physical sector mapping. It will do garbage collection when left alone.

Do not conventionally RAID it. Do not write too much at once. Try a CoW FS with checksums like btrfs or ZFS.

And ALWAYS back up.
October 17, 2025 at 12:27 AM
yeag
October 15, 2025 at 7:24 AM
I don’t know about Irdeto’s commercial arrangements. Just that DAC is absolute weaksauce.
October 15, 2025 at 7:23 AM
Denuvo Anti-Cheat is a totally different thing to Denuvo Anti-Tamper.

(It’s much weaker and … who even uses it?)
October 15, 2025 at 7:20 AM
I’ll see you around, whenever and wherever you can; and I hope you can one day remember that passion for creativity. 💜
October 13, 2025 at 7:29 AM
Cohost was, I say again, too damn good for this cursèd world.
October 6, 2025 at 4:22 PM
Yeah, I think Bell and Braben could genuinely say they have 41 years of experience in that!

But, yeah. It has. Limitations.
October 3, 2025 at 12:52 PM
You’ve just described broadly what satoshin’s arc of thought was at the time, and we already had a bigint hashcash from I2P design’s anti-spam (never used).

I didn’t think it was a good idea. Worst possible solution, I think I said?

I didn’t want to be proved THAT right, however, as I said above.
October 3, 2025 at 5:21 AM
The cypherpunks were/are trying!

But… it turns out it’s *very* hard to do anything involving actual payment like that, especially without attracting an ecosystem of weird investor scam cult finance bros, like black mould on damp walls. I’m not sure tech can solve every problem. :/
October 2, 2025 at 6:08 PM
Yes.

Crackers don’t always play fair.

To quote a dearly departed friend: sometimes the only way to play the game is to cheat.
October 2, 2025 at 7:22 AM