psyops only. high priest @mutCarolineDAO. what would 奶奶 do? tweeting stream of consciousness
Goblintown
Joined January 2021
- Tweets25.3K
- Following2.6K
- Followers12.6K
- Likes18.6K
The frontier lab to healthcare pivots are reasonable cuz once you make that much money you want to live forever.
Wall Street: “Dario, token prices are so cheap and dropping, how are you going to keep selling premium frontier intelligence at this rate?”
Dario: “Spend 20x the tokens doing lean proofs on all the code that’s written”
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…
Just have a service where you show you own a domain and then Anthropic points mythos at you for 20 minutes and gives you a report? Instead of dealing with random refusals.
> tried to warn the world
> kept obvious defensive options nerfed behind special enterprise plans
> forced users to use Chinese models to defend themselves.
I pointed opus at this trailer. It built the game in 45 minutes
A former BioWare employee uploaded this Mass Effect Universe (Mass Effect 5) trailer on Reddit! According to another employee, this trailer was ready a few months ago. #MassEffect
> they got blown on shorting into a non-existent black monday only to re-long and get blown out on black friday.
Singularity price thesis was regrettably wrong, but AI will still thrive and change the world every day
the attackers get to use deepseek. the defenders have to apply for daybreak blue, then do some weird stuff.
🚨DEFCON 1🚨
In a few hours, a critical vulnerability that affects ~70% of web sites will be disclosed. It allows an attacker to upload and execute arbitrary code.
It's probably best if you stop browsing the internet for the rest of the week.
xitter.cf/ThePHPF/status/2112633…
Readers added context they thought people might want to know
These are server-side PHP RCE vulnerabilities for which web hosts/admins should prepare to patch per official notice; general users face no direct need to stop browsing. PHP detected on ~69% of sites with known server-side language.
php.net/archive/2026.p…
w3techs.com/technologies/d…
cert.europa.eu/publications/s…
sansec.io/research/php2s…
arstechnica.com/security/2026/…
> he’s the Mr beast of finance
What is AI really capable of in bio?
To separate hype from reality, we spent 3 months working firsthand with @AnthropicAI's internal research model to see if it could create an experimental scorpion antivenom.
Introducing Citrini Therapeutics.
citriniresearch.com/p/citrin…
> blocking deepseek was psyops to the Chinese govt to accelerate open releases after the Chinese issued that export ban thing. 100D chess
奶奶 capital retweeted
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source!
Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable.
🔗 github.com/deepseek-ai/DeepI…
1/4
