I’m confused - did AI2040 miss that small models will be dangerous? Or what do they know that Christiano doesn’t
6
While I deeply believe we are dealing with buckets of AI risks now and significant emergent risk in the near future - the US is experiencing social manipulation by foreign actors at a scale unseen since 2016 amplifying all activities to appear apocalyptic
8
Long compute?
‼️ Microsoft has filed for a patent to train AI on gamers' idle graphics cards and pay the owners for it.
The system in the application, published on 5 November, would split an AI training job into small tasks sized to each device. It would watch every card's load, temperature and power draw, and back off when the machine gets busy or hot.
Results would be cross-checked against other PCs to catch faulty or cheating machines. Microsoft has announced no service and no pay rates.
Owners could be paid in cash, credits or free software, or with access to the AI model their PCs helped train.
18
“I’m putting together mez debt for Malaysian Blackwell clusters” no bro you’re about to learn how to play dominoes in a supermax in Nebraska
5
A weirdly accurate heuristic for thinking about how to understand a sovereign human led entity of any complexity is to picture a pre agrarian tribe and have each member represent the some corresponding % of the entity’s population
7
Science and religion describe the same phenomenon with a deeply different perspective. VCs and founders feel oddly similar in their gap - order, structure, science vs instinct, intuition, and callings
11
The primary ingredient of a fulfilling media diet is finding information that has a really long half life
1
127
Each election is all about getting one big important but easily overlooked thing right
Republicans mismanaged AI/data center narrative and got murdered today
20
There’s a lot of people deluding themselves into thinking they are hacking out of their cage when they are really just studying it
14
Allen and cos failure to get a seat to underwrite ant ipo speaks to how little nyc saw any of this AI stuff coming
19
Haint Marc rich is summoned in a White House advisor every few years until the possessed transcends into post economic hedonism or congressional trial
28
Atmospheric Ignition retweeted
We formally verified 123 open-source projects in a partnership with Anthropic. Formal verification at scale is here!
Now we’re launching OSS Verify to scale it to the software we all depend on. We prove it’s crash-free or send you the patches it needs. Free for open-source. Let’s bootstrap the verified software ecosystem: theorem.dev/oss-verify
New on our Frontier Red Team blog: using Claude to formally verify software at large scale.
Formal verification can prove that code is free of certain bugs, but historically it has only been feasible for small codebases (or required a lot of expert time).
With AI models, however, this is beginning to change. In collaboration with @theoremlabs, we used Claude to prove security properties in over one hundred real-world open-source libraries over the span of a few weeks.
Read our research in full here: anthropic.com/research/forma…
Cyber threats are lab marketing/reg attempts people need to take a serious look at the thought processes that led to their confusion, and consider impending risks and immediate actions needed
9








