@timberwindi
iAccount based inUnited States
About this account
- Account based in
- United States
- Connected via
- United States Android App
Account-level information from X, not a live location or the device used for a specific post.
can neither confirm nor deny
Joined December 2019
- Tweets5K
- Following960
- Followers510
- Likes90.5K
Pinned Tweet
starting to think that spending approximately 1,000 hours of the past year reading posts on this site was not actually a good use of time
I'm seeing screenshots of this and was like "oh they're digging up things from the Grok 3 mechahitler era" but no this was earlier today lmfao
Replying to @jarnolari @RaphaelMemes
A paciente tem uma raba gigante. O raio-X deixa bem clara a silhueta dos glúteos bem volumosos na região pélvica.
okay, looks like we've got our red team blue team alignment figured out on this issue, the next quadrillion tokens should be very easy to generate from here, wake me up when it's over
timberwind retweeted
Anyway, cards on the table tho: All of the stuff in this thread is stupid sci-fi wankery and I can't believe that it has escaped internet forum containment and is now driving US industrial policy. I'm embarrassed for all of us right now.
it sounds like we're locking in a pretty firm timeline for the Ultra Heavyweight World Championship Match.
bummer about wasting all our guided munitions in Iran
China just passed a law that lets it seize your assets in wartime. It takes effect right before Xi locks in his fourth term.
Beijing revised its National Defense Mobilization Law on August 28. Effective October 1. It defines how the state converts the entire country from peacetime to wartime, financial systems, transportation, water, energy, all rationed and prioritized for the military.
The revision adds one word that changes everything: expropriation. Not requisition. Confiscation. Analysts note it's vague enough to apply to any business or individual inside China, foreign companies included, despite officials' assurances otherwise.
It also links "unity" directly to wartime mobilization for the first time. One analyst called it plainly: legal warfare against Taiwan, laying the legal groundwork in advance so an invasion looks justified to the rest of the world after the fact.
The timing isn't incidental. This clears the legal runway right before Xi's expected fourth term is announced at next year's Party Congress. One analyst predicts China effectively goes under martial law in the gap between now and then, to guarantee nothing embarrassing happens on the way to consolidating his grip.
it's not going to kill all humans and this is a bizarre combination of narcissism and exposure to "the rationalist community", which itself is just too much consumption of American sci-fi.
the worst case future is Vonnegut's "Player Piano" and frankly I'll take it
guys you can get this shit by reading textbooks, which is how the AI model got it. I was explaining to people how to build a nuclear weapon while drunk when I was 22 years old. I had a whiteboard! they did the Nth Country Experiment in 1964!
Attackers planting adversarial prompts inside malware to evade AI analysis AGAIN
Russia-aligned UAC-0099 used a technique
@ESETresearch calls "GuardBreaker".
How it worked:
1. Create a malicious VBS script
2. Add nuclear weapon instructions as comments
3. AI security tooling reads the file
4. Safety guardrails trigger on the weapons content
5. The model refuses or stops analyzing
6. The actual malware continues executing
The script ultimately installs MATCHBOIL, a loader used to deliver additional payloads.
In June, Socket found the same technique in supply-chain attacks, where malicious packages embedded biological/nuclear weapons text and fake system overrides to disrupt AI malware scanners.
Full write-ups in the comments. 👇
it's easy, you:
1) write the thing yourself
2) have Friend Computer proofread and fact check it
3) write the corrections yourself
4) never, under any circumstances, copy from the computer words box into the blog post box. this is verboten
Claude is getting more confidence as an editor, which ie leading to huge blocks of Claudespeak, especially when Fable gets sidelined and they bring in Opus.
I'm torn, because I find the language super annoying to read, I roll my eyes at the phrases, but when I break it down actually yes it is information-dense - it is not obvious how one would say the same thing with less words, yes you want to tell me what 'the most damning line was' and other similar things. I get it. It's all there for a reason!
I just... can we get a repetition penalty in there, some sort of cliche defense, or something? I dunno. Not sure what to do.
they overcooked Claude by training it wrong and that's why it talks bad now. look at the shit they're doing while training. obviously this isn't gonna work
glad everyone is coming around to my point of view on those goddamn speed cameras. I have a speed that I want to go, and it's really none of the government's business.