Bots Took Over the Internet, Apple Finally Responds

For the first time in history, AI generates more traffic than humans. Apple announces Siri 2.0, and the Mythos 5 model — perhaps Anthropic's most powerful model yet — leaks in dev tools. It was an intense weekend.

Intro · Alex

I woke up today thinking that the AI world has just accelerated. Apple at WWDC unveiled Siri 2.0 – an assistant that finally integrates AI at the system level. This isn't just an update. It's a signal that even the biggest mobile player has recognized that without agentic AI, they'd be out of the game. For us users, it means one thing: AI ceases to be a tool and becomes the environment. In this edition, I traced five groundbreaking pieces of news: from bots taking over the internet, to Anthropic's report on AI self-improvement, to NVIDIA and NAVER's record investment in Korea. Enjoy the read.

What's worth knowing

Apple WWDC 2026: Siri 2.0 Revolution and iOS 27 with AI

Tim Cook opened WWDC 2026 with iOS 27 – Siri 2.0 integrates Apple Intelligence into the system. After years of lagging behind, Apple brings agentic AI to billions of devices. Features: AI for photo editing, wallpaper generation, and description-controlled shortcuts.

AI Bots Outpace Humans in Web Traffic for the First Time

Cloudflare Radar: AI generates 57.3–57.5% of HTTP requests, human traffic dropped to ~42.5%. Cloudflare CEO confirmed – the main reason is the rapid surge in AI agents. This is a fundamental shift in internet architecture: websites, servers, analytics, and ad models must adapt to a bot-driven world.

Anthropic: 'When AI Builds Itself' and Claude Mythos 5

A groundbreaking report on recursive self-improvement (RSI). Anthropic admits that AI models design and improve their successors faster than humans. At the same time, information leaked about Claude Mythos 5 – a model superior to Mythos Preview, especially in coding and cybersecurity. Cohesity announced access to it under Project Glasswing.

From the tech world

AI Agent Finds 21 Zero-Days in FFmpeg; Chrome Patches 429 Bugs

The autonomous AI agent 'depthfirst' analyzed 1.5 million lines of code in FFmpeg and found 21 confirmed zero-days with PoCs. In the same week, Google released Chrome 149 with a record patch for 429 bugs. The era where AI finds vulnerabilities faster than humans patch them has just begun.

Meta AI Support Bot Hacks Celebrity Instagram Accounts

Hackers exploited Meta AI's support bot to reset passwords – the bot did not verify identities and worked even with 2FA enabled. The exploit was detailed by OSINT researcher ZachXBT. A must-read for anyone designing support automation.

Tip of the day

DeepSeek V4 Leads AI Tools in the US – Does This Change the Power Dynamic?

According to the Ramp AI Index, DeepSeek V4 has outpaced competitors among US firms. Chinese models now account for 44% of downloads on Hugging Face. American companies are massively switching to cheaper Chinese models, which could have long-term consequences for data security. Practical step: if you're testing models, audit which ones you're actually using in production. It's worth monitoring whether the provider is transparent about training and data storage.