If there was ever a top signal, where family asks about some hyped volatile tech thing that reached mainstream news in their bubble, then today marks the top of "AI safety concerns".
1
4
973
There’s going to be a pivot point where I call more with @bot than listen to spotify
2
2
1,149
"V2 keeps that decision, but changes the center of gravity in five ways."
Thanks for rewriting everything again GPT. V3 will be done by claude.
1
930
If AI labs want to "pace the frontier", is it:
A) for good/safety of the world
B) scare tactics and regulatory capture
C) to not wait for Mr Market to do it for them
The Left is saying A, the Right is saying B, I suspect it's more like C.
3
8
1,083
The "2026 State of Data European Data Centres" report projects a total investment of €137.5B.
*split between 5 years
*mostly from US companies
*nvidia mcap changes more per day
*out of ~ $3T-$5T worldwide
*mostly to be reinvested in US chips
*permits will add years of delay
3
14
1,816
Bessent is increasing buybacks of long term debt, funded with short term debt? Isn’t this basically just reducing duration, and reducing term premiums? Might see both more bond buying and more alt asset buying if so. I wonder if we will see a version of the “debt optimization” idea that has been going around, to get banks more room to lend more “without QE”.
1
5
995
proto retweeted
Public service announcement about @Etched. It's possible they have a great chip and just very distasteful marketing. It's also possible they have no chip or a terrible chip and want you to fall for the smoke and mirrors. This type of marketing is bad for technology.
75
43
1,348
286,557
A lot of people didn’t care to disable “Sent from my iPhone”. Many kept it as a sign of fashion. Some android users faked it even.
If this becomes a thing, then low end models may mimic the watermark of high end models. And maybe longer texts will counter it with cryptographically signed watermarks.
Anthropic says new Claude models will embed invisible watermarks in all generated text, everywhere Claude is offered.
The watermark is part of the text, it isn't metadata: "it will travel with the text when it's copied and pasted elsewhere, and may persist through some editing."
This starts with models launched on or after August 2, 2026, under an EU AI Act code Anthropic signed. Anthropic is still working on adding it to current models. The rollout is worldwide.
2
11
1,935
proto retweeted
SECRETARY RUBIO: "This is a distinctive and unique evil. It has always been driven by a hatred, above all else, for civilization itself."
1,235
5,050
25,697
6,344,757