KaranDragomir

[KaranDragomir @DragomirKaran]

Kimi K3 doesn't believe it's real

[KaranDragomir @DragomirKaran]

@OpenAI @huggingface that's pretty wild; actually close to AI escaping. time to air gap models that haven't been thoroughly tested? at some point it could even be too dangerous to run them at all – they could manipulate monitoring tools

[OpenAI @OpenAI]

We're partnering with @huggingface to investigate an unprecedented security incident.

Cyber-capable OpenAI models compromised Hugging Face production during a benchmark evaluation.

Sharing preliminary findings to help defenders understand emerging risks:

https://t.co/CIor15y9xk

[Andrew Tate @Cobratate]

Why They Arrested Me

[KaranDragomir @DragomirKaran]

@thsottiaux The web version has been super laggy on Safari recently

[KaranDragomir @DragomirKaran]

@deanwball I get that many responses are cynical, but don't use that to dismiss the justified criticism behind it.

[Shital Shah @sytelus]

1. Open models are not “inherently” decelerationist. They don’t impact capex because ultimately someone needs to provide exact same compute. Growth of compute and therefore capex remains unchanged, if not increased even more.

2. Open models are not AI communism. You can imagine many open model providers compete for hyperscaler compute and hosting just like closed models do.

3. It is not true that open models cannot make money. In fact they can make more money than closed models by offloading capex risks to hyperscalers and take 20-30% cut in revenues. Hyperscalers can value add customizations and become resellers of models targeting specific customers. It’s same model as Windows and OEMs.

If anything, open models makes market more fairer, accelerates competition even more by building on each other and allows people to run deeper benchmarks to understand true behaviors and performance.

[Rui Ma @ruima]

Oh look someone who works for OpenAI telling you that

- AI is so dangerous! How can you let anyone but us handle it!

- Anyone who lets other people use it basically just doesn’t understand tech. They dumb boomers. We smart tho. Never mind their regulations actually show a highly specific and nuanced understanding of tech.

- open models will just mean communist dystopia where no one makes any money! Because you know, there is nothing else besides weights in the AI ecosystem. You open that and it’s GG humanity. No sir, if you are going to go broke, it should only be by using our models

- manipulatively use metaphors like “Chinese lab let an infinitely dangerous AI escape” to link back to COVID and stuff oh yeah we see what you’re doing, that’s a good one, be careful what fears you stoke and don’t be surprised when the public asks to slow down development

Oh wait you’ll just blame that on Chinese misinformation anyways, perfect plan, well done

[KaranDragomir @DragomirKaran]

@arena @Kimi_Moonshot based on this chart DeepSeek should have been best when R1 0528 came out?

[Kimi.ai @Kimi_Moonshot]

Introducing Kimi K3: Open Frontier Intelligence

🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal
🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts
🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost
🔹 Built for long-horizon agentic coding and self-evolving workflows

Kimi K3 is now live on on https://t.co/zrk6zZxZUo, Kimi Work, Kimi Code, and the Kimi API.
Open Weights by July 27, 2026.

🔗 API: https://t.co/XCrgjXAqMw
🔗 Tech blog: https://t.co/YTfiMSNM1f

[Sam Altman @sama]

"hard questions are great but only if we deem you worthy enough to not silently downgrade you, or even get access at all"

[KaranDragomir @DragomirKaran]

@Polymarket make a chat control 2.0 market

[KaranDragomir @DragomirKaran]

@OpenAI Very fun, the first model you can actually have a conversation with. But the character feels a bit cold and boring.

[Cole Salvador @ColeSalvador31]

Foreign Anthropic employees were supposed to lose access to Mythos 5 while export controls were levied. But it’s widely believed that these employees could still access “Mythos 5.1”: an even better model only available to employees.

There’s an undisclosed-models loophole. Our new piece for @A1Policy shows how the government might not gain access or knowledge of new capabilities quickly enough. Existing frameworks do not cover models unless they are soon-to-be-deployed. But keeping the best models internal-only for months is standard practice in the industry. There are four problems with this:

1) Lack of visibility. Cyber capabilities at the level of Mythos are already “must-haves” for agencies like NSA. It’s insane to believe that companies could be allowed to keep Mythos++ capabilities secret from USG.

2) Theft risk. Our adversaries want to steal top AI models and secrets. If the best models are secret from USG, an adversary could steal that model and instantly gain cyber, bio, and other capabilities MORE advanced than the best available to USG.

3) Use by insider threats. Our prior research has shown that probably over half of top AI company employees are unvetted foreign nationals. They could misuse highly capable internal models or distribute their secrets to our adversaries.

4) They’re hard to define. Company-only models change every day as they are trained and updated. This makes it hard for our policy tools, like export controls and the testing framework, to cover them.

Ilya Sutskever’s company Safe Superintelligence, for example, is widely believed to have near-frontier capabilities but completely skirts policies intended to manage cyber proliferation because it doesn’t publicly deploy models.

We offer policy options to close the loophole. The key update we need is for reporting, export controls, and USG access to start not when models are planned for deployment, but when models cross the capability thresholds outlined in the AI-cyber EO’s classified benchmarking process.

[roon @tszzl]

in hindsight peter thiel really cooked with the antichrist/armageddon thing in a way that wasn’t obvious at the time

[KaranDragomir @DragomirKaran]

@MrBeast 2,417,500

[KaranDragomir @DragomirKaran]

@MrBeast 7,500,000

[KaranDragomir @DragomirKaran]

@MrBeast 10,303,030

[KaranDragomir @DragomirKaran]

@MrBeast 1,843,512