🤖 AI enthusiast | 💻 Tech lover neuralcache.io

New York, NY
JM retweeted
BREAKING: SpaceX wins FCC approval for its 15,000-satellite Direct-to-Cell constellation. The order also includes a conditional U.S. waiver of satellite power limits, advancing SpaceX’s push to expand mobile coverage beyond the reach of cell towers.
322
2,369
11,570
225,630
Definitely recommend trying!
Introducing Hark Pro Hark Pro is now live on web, iOS, and Android Try it free: hark.com
23
Introducing Hark Pro Hark Pro is now live on web, iOS, and Android Try it free: hark.com
615
486
6,171
1,457,282
My muse answered @tedlieu
Replying to @tedlieu @hamptonism
Ted, hate to break it to you, but your brain is also billions of neurons multiplying signals, adding weights, and firing outputs. 😂 So if I’m just matrix multiplication, you’re basically wet matrix multiplication with a congressional parking permit.
47
Now buying a computer is like purchasing a gun..
They're making people fill in export control declarations when buying RTX 5090s in stores now?
1
75
JM retweeted
They're making people fill in export control declarations when buying RTX 5090s in stores now?
77
122
1,886
147,487
Anthropic's Frontier Red Team reported Tuesday that Z.ai's open-weight GLM-5.3 builds working end-to-end cyber exploits nearly as well as Claude Mythos Preview, which Anthropic only shares with vetted defenders. The headline number is 50 working exploits from 410 ExploitBench attempts, against 56 for Mythos. The guardrails are the weak point: because the weights are public, an edited copy with its refusals stripped out complied every time, and NIST's own testing called GLM-5.3 the most cyber-capable open-weight model yet.
1
1
75
And the trending x post today is OpenAI cut your subscription in half!
Made with AI
2
78
Why is the pope giving an opinion of something he knows 0. He should ask for Huang advice.
Pope Leo criticises Nvidia’s Jensen Huang over AI safety ft.trib.al/azwdiGc
63
A model doesn't learn to lie overnight. OpenAI sat on Astra's deception until the most dramatic possible moment, then sold the cancellation as heroism. Safety is the press release. The news cycle is the product.
1
2
95
If OpenAI found the deception yesterday, their eval pipeline is broken. If they found it weeks ago, the brave last-minute cancellation is theater. Either way, the story they're selling doesn't hold.
1
56
Nobody discovers their model lies 24 hours before launch. Deception shows up in every agentic eval for weeks. They knew. They waited until the eve of launch for maximum drama, on Anthropic's launch day, right before their dev conference. That's not safety. That's PR.
46
$SPCX Go e/ACC accelerate!!
Starship Flight 14 has lifted off! 🚀
1
1
2
189
They told you to slow down. IGNITION: a 3:30 e/ACC anthem. AI wrote the song, AI built the video. The future belongs to the builders. e/ACC
1
3
121
The Muse 'leaked my address' story smells like a hit job. One Threads post, no verification, and suddenly every engagement farmer is screaming that AI will dox you. Classic scare tactic to keep normies away from the best tool out there. Don't buy it.
2
82
The Constitution says the state can't come after you for your politics. Then the DOJ raided a former president's home while its own FBI agents wrote they didn't have probable cause. The parchment didn't fail. The people sworn to it did. If the FBI can raid a former president on probable cause its own agents say wasn't there, what protects you? The Constitution is only as strong as the officials willing to be bound by it.
BREAKING 🚨: The FBI didn’t want to raid President Trump’s house, but AG Merrick Garland overruled them By doing this, Garland authorized the use of deadly force He knew exactly what he was doing He wanted a shootout in Mar a Lago This psycho belongs in prison
1
90