کاش ایران همچین قانونی داشت. ولی اینجا خود دولت و اپراتور اسپم میکنه. به کی باید شکایت کرد.
I had my Muse agent put me on the National Do Not Call/Text Directory. By law telemarketers and other companies have 30 days to scrub my number from their lists. Otherwise Muse will automatically start logging every violation and will skip trace and go after every single one of the violations and hopefully collect between $500 and $1,500 per violation. Let's see how this goes.
برای اونایی که کد رو میخواستن:
github.com/erfanium/sound-ou…
فکر کنم از ایرانیا یاد گرفتن چطور DNS Tunnel بزنن😁😁
یکی از مدلهای OpenAI راه خروج از محیط ایزوله آموزشی رو پیدا کرده! 😳
کمپانی OpenAI اعلام کرده مدلی با سوءاستفاده از یه نقص DNS، از محیط آفلاین خارج شده!!! اتفاقی که باعث توقف موقت آموزش مدلهای پیشرفته در ماه سپتامبر شد.
بعد از بررسی آسیبپذیریها و تقویت امنیت، آموزش از اول اکتبر دوباره از سر گرفته شده.
آنتروپیک چند ماه پیش در یه بلاگ پست گفته بود ۹۹ درصد ترافیک مربوط به distillation رو مسدود کردیم، دسترسی به CoT رو بستیم و این حرفا. اون موقع هنوز Opus 5.5ی نبود.
بنچمارک های Kimi K3.1 Pro نشون میده این مدل قطعا از Opus 5 بهتره. پس عملا کل تئوری های مربوط به distill رفت رو هوا.
Kimi K3.1 Pro launches at 4th on the Artificial Analysis Intelligence Index scoring 55, making large gains in agentic performance; it will be the leading open-weights model once its weights are published
@Kimi_Moonshot's new reasoning model gains 11 points over Kimi K3 (max, 44). It trails only Claude Opus 5.5 (max with fallback, 58), Claude Sonnet 5.5 (max with fallback, 56), and Muse Wonder Full (max, 56) on the Intelligence Index.
Once the weights for K3.1 Pro are released it will be the highest-scoring open weights model and the gap between the leading open weights and proprietary models on the Intelligence Index will narrow from 12 points to 3.
Key results:
➤ Large gains in agentic performance: Kimi K3.1 Pro scores 3rd on GDPval-AA with 1,846 Elo, up 312 Elo from Kimi K3, behind only Claude Opus 5.5 (1,866 Elo) and Muse Wonder Full (1,857 Elo). Kimi K3.1 Pro also achieves the highest score of any model on Terminal-Bench 4.0. K3.1 Pro scores 67%, up from 13% for Kimi K3 and ahead of Claude Sonnet 5.5 (64%). AutomationBench-AA, which tests agentic SaaS workflows, rises from 58% to 73%.
➤ Gains in other evals are more moderate: Humanity's Last Exam rises from 47% to 52%, CritPt from 23% to 25%, and SciCode from 59% to 61%. AA-Omniscience Index rises from 20 to 30, as the hallucination rate falls from 53% to 44% and accuracy rises from 48% to 51%.
➤ Cost per Task roughly equal to Kimi K3 at $1.98 per Task: K3.1 Pro uses 55K output tokens per Intelligence Index task, 14% more than Kimi K3 at 48K, while its first-party token prices are 33% lower.
Additional model details:
➤ Context window: 1M tokens.
➤ Size: 2.8T total, 104B active parameters
➤ Input modalities: Text and image
➤ Weights and license: Weights currently not published, Moonshot is expected to release the weights soon after launch
➤ Pricing: $2/$10/$0.20 per 1M input/output/cache hit tokens, 33% lower than Kimi K3 on all three rates.
Congratulations to the team at @Kimi_Moonshot on the release!
erfanium retweeted
"End-to-end encryption stops at your screen."
This part is already confirmed ;-).
A new book says Dutch spies spent months watching Jared Kushner's iPhone screen in real time, and one of the things they saw him open was a bank account with about $750 million in it.
They got in because the Russians had already hacked the phone, and the Dutch found a way to ride along.
Here's how it went down, according to Matthew Cole's reporting in Vanity Fair. In late 2017, Dutch military intelligence found out a Russian GRU officer was about to travel to the UK. The guy was a dark-web hacker who dealt in stolen bank data. The Dutch gave him two options, work for them or get picked up by the British, and he picked the Dutch. He then handed over access to a GRU server in Rostov, and sitting on that server were screenshots from Kushner's iPhone X. So Russian military intelligence was already inside the phone of the man running America's Middle East policy.
From there, Dutch hackers in The Hague figured out how the Russians pulled it off and built their own way in. Cole says they could see his screen whenever he was using it, and they saved every frame. Sound was the one thing they couldn't get.
Think about what that means for anyone who uses WhatsApp or Signal. The encryption does its job while a message is flying between two phones. Once it pops up on a hacked screen, whoever's watching just reads it right along with you.
End-to-end encryption stops at your screen.
That's relevant here, because back in 2019 the head of House Oversight said Kushner's own lawyer had confirmed he used WhatsApp for official White House business and kept records by screenshotting his chats. The lawyer pushed back on parts of that account.
The Dutch have pulled this kind of thing on Russia before. In 2014, their spies got inside the network of Cozy Bear, the Russian hacking crew later tied to the DNC hack. They even took over the security camera in the hallway outside the hackers' office near Red Square and had photos of everybody walking in and out. When Cozy Bear went after the State Department that November, the Dutch tipped off Washington, and the Americans reportedly sent them cake and flowers as a thank-you.
As for the money, Cole alleges the Saudi and Emirati rulers put at least $2 billion into accounts tied to Kushner during Trump's first term. Kushner says every word is false and that the author never produced a single bank record. His lawyer is threatening to sue. The full book comes out October 27, so we're about to find out who has the receipts.
If Cole's sources hold up, there's a hard drive somewhere in The Hague with a frame-by-frame replay of the most sensitive phone in Trump's first White House.
erfanium retweeted
🚀 DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.
🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
حداقل چندین ساله شرکت های خارجی دیگه شمشیر رو از رو برای ایرانی ها بستن.
نمیخوام شرایط موجود رو تایید کنم. ولی اون ایرانی که هنوز کل زندکیش توی سرویس های گوگل و ماکروسافت و ... هست باید یه دوره مدیریت ریسک بگذرونه.
This quoted post is unavailable.
تو این فکر هستم یه اپلیکیشن آنتیاسپم هوشمند مناسب شرایط ایرانی ها طراحی کنم. احتمالا از Jev (یا چیزی شبیه Jev) برای تشخیص استفاده کنم که خیلی دقیق عمل کنه.
مهم ترین چالش حریم خصوصی هست که احتمالا خیلیا اعتماد نکنن.
























