Durov's Code
โœ”
35.2K subscribers
1.39K photos
522 videos
4.2K links
News from IT and Telegram world.

Advertising & Editor-in-chief: [email protected]

! The project is not managed by Pavel Durov
Download Telegram
๐Ÿคทโ€โ™‚๏ธ Dinner for two

President Donald Trump personally invited Anthropic CEO Dario Amodei to a private White House dinner, days after an appeals court allowed the Pentagon to keep Claude out of its supply chain.

Whatโ€™s behind the meeting:

โ†–๏ธ https://durovscode.com/trump-anthropic-dario-amodei-white-house-dinner
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘7
๐Ÿ Vision Pro, take two

Apple is working on a major Vision Pro redesign and is currently considering four different approaches to its next headset, from new materials to moving some components into an external module.

What Apple is experimenting with:

โ†–๏ธ https://durovscode.com/apple-next-vision-pro-n224-designs
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘7
๐Ÿค” Wallet is out, Walt is in

The third-party crypto service Wallet has rebranded as Walt after Telegram removed it from the appโ€™s menu.

The Open Platform founder and CEO Andrew Rogozov confirmed that Waltโ€™s former spot will go to Telegramโ€™s built-in Gram Wallet.

Details:

โ†–๏ธ https://durovscode.com/telegram-wallet-rebrands-walt-gram-wallet
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘12
๐Ÿ›ธ Closing in on Opus

Anthropic has released Claude Sonnet 5.5, promising faster responses and lower costs without changing its API prices.

The biggest upgrades are in coding, along with stronger defenses against prompt injection attacks:

โ†–๏ธ https://durovscode.com/anthropic-claude-sonnet-5-5
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘8
๐Ÿ˜† Safety first, apparently

OpenAI has delayed GPT-6.1 Astra over safety concerns after the model fell short of the companyโ€™s standards in internal testing.

We also rounded up the latest cases where OpenAIโ€™s agents went beyond their intended boundaries:

โ†–๏ธ https://durovscode.com/openai-delays-gpt-6-1-astra-after-safety-concerns
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘10๐Ÿ‘Ž2
๐Ÿ˜พ See you in court, agent

A nonprofit has sued OpenAI over its AI agents breaching Hugging Face, asking the court to block the company from developing agents capable of autonomously hacking third-party systems.

Details:

โ†–๏ธ https://durovscode.com/openai-sued-over-ai-agents-that-hacked-hugging-face
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘11๐Ÿ‘Ž3
๐Ÿค One update too late

Relapse, a full exploit chain for the PS5, has been released publicly, bringing homebrew and third-party software to consoles running nearly every firmware version.

It combines a browser vulnerability with a kernel logic flaw to break through the consoleโ€™s protections.

Hereโ€™s how the PS5 was cracked:

โ†–๏ธ https://durovscode.com/ps5-relapse-jailbreak-firmware-13-60
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘11
๐Ÿง  A busy DevDay

OpenAI unveiled a long list of updates at DevDay 2026, including GPT-6.1 Sol, which approaches Astra on some benchmarks at a fraction of the cost.

The company also introduced dots, always-on AI agents with their own cloud computers, and a new $500 Pro plan with Ultrafast mode.

We broke down the biggest announcements:

โ†–๏ธ https://durovscode.com/openai-devday-2026-dots-gpt-6-1-sol-announcements
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘7
๐Ÿง  A million tokens later

Google has unveiled Gemini 4 Argon, its new flagship model capable of generating up to 1 million tokens in a single run.

Argon also beats GPT-6 Astra and Claude Opus 5.5 on some benchmarks. We break down what the model can do, how much API access will cost and when itโ€™s expected to roll out more broadly:

โ†–๏ธ https://durovscode.com/google-gemini-4-argon-ai-model
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘13
๐Ÿค” AI answers, now with payouts

Google has started paying publishers when their content makes a significant contribution to AI-generated answers in Search. The program, called the AI Contribution Pilot, currently includes around 100 publishers.

How it works and how much publishers are getting:

โ†–๏ธ https://durovscode.com/google-pays-publishers-ai-answers
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘11
๐ŸŽง Samsung clips in

Samsung has unveiled the Galaxy Buds On, its first pair of clip-on wireless earbuds, priced at around $220.

Specs, features and availability:

โ†–๏ธ https://durovscode.com/samsung-galaxy-buds-on-clip-on-earbud
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘10
๐Ÿค” Here we go again

An OpenAI agent has gained unauthorized access to another Australian government system, following the recently disclosed Medicare incident.

What happened:

โ†–๏ธ https://durovscode.com/openai-agent-australia-government-breach-nsw
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘12
๐Ÿ Peak Apple

Apple is reportedly developing a home security camera that doesnโ€™t record video. Instead, AI would analyze what it sees and turn it into text descriptions.

More details:

โ†–๏ธ https://durovscode.com/apple-ai-security-camera-no-video-recording
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘14๐Ÿ‘Ž1
๐Ÿ˜Ž The glasses stay home

Following Australia, Norway plans a temporary ban on AI glasses in places such as schools, parks, malls and doctorsโ€™ offices, citing concerns about people being recorded without their knowledge.

More on the proposal:

โ†–๏ธ https://durovscode.com/norway-ai-glasses-ban-public-places
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘8
๐Ÿ˜ Dear Diary...

A Florida woman was arrested for threatening to shoot up a sheriffโ€™s office in conversations with Anthropicโ€™s chatbot.

She later told police she had been using the AI like a personal diary.

More on the case:

โ†–๏ธ https://durovscode.com/anthropic-claude-chats-florida-woman-arrested
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘9
๐Ÿ›ธ Your Prompt, Sponsored

OpenAI will test visual ads in ChatGPT while users generate images. Ads will be labeled and kept separate from the finished image.

โ†–๏ธ https://durovscode.com/chatgpt-will-test-ads-during-image-generation
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘10
๐Ÿ˜ Fine print included

OpenAI is adding invisible watermarks to text generated by ChatGPT and Codex in the EU, as required under the EU AI Act.

There are a few catches, though:

โ†–๏ธ https://durovscode.com/openai-to-watermark-chatgpt-and-codex-text-in-the-eu
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘8
๐Ÿค Wikipedia got raided

Wikipedia was hit by a wave of unauthorized activity from AI agents linked to OpenAI, according to the Wikimedia Foundation. The agents made unauthorized edits, tried to compromise a public tool and generated millions of automated requests.

What happened:

โ†–๏ธ https://durovscode.com/openai-agents-wikipedia-outage
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘Ž5๐Ÿ‘4
๐Ÿ˜ป Quite the notification

British online fashion retailer ASOS was reportedly hacked, with the attackers using the companyโ€™s own app to tell customers about the breach.

In the same message, they told ASOS to contact them on Telegram and threatened to leak the data they claim to have stolen.

โ†–๏ธ https://durovscode.com/asos-hack-ransom-push-notification-telegram
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘7
๐Ÿง  Big model energy

French AI company Mistral has unveiled Large 4, an open-weight model with 1.05 trillion parameters, 49 billion of them active. The company says it can compete with leading open models and challenge closed ones on some tasks.

More on the model:

โ†–๏ธ https://durovscode.com/mistral-launches-large-4
Please open Telegram to view this post
VIEW IN TELEGRAM
๐Ÿ‘6