News & Updates

AI Dev Newsroom

Industry news, community updates, and everything shaping how context engineers build with AI.

172

Articles

Topics

Live

Apple Intelligence Finally Approved for China — Alibaba's Qwen to Power AI Features Across iOS, iPadOS, macOS and visionOS After 22-Month Regulatory Wait

Latest

Update

16 Jul 2026

Apple Intelligence Finally Approved for China — Alibaba's Qwen to Power AI Features Across iOS, iPadOS, macOS and visionOS After 22-Month Regulatory Wait

China's Cyberspace Administration approved Apple Intelligence on 15 July, ending a 22-month wait since Apple first announced the feature at the iPhone 16 launch in September 2024. Alibaba's Qwen model will power text and image understanding and generation across all Apple platforms in mainland China, with Baidu and potentially DeepSeek also contributing features.

171 articles

COR Ambassador Programme — Coming Soon

We're building a global network of COR Ambassadors. Want to represent Context Engineers in your city? Details dropping soon.

ASML Smashes Q2 Estimates on AI Chip Demand — Raises Full-Year Guidance to €43–45 Billion as Stock Surges 75% Year-to-Date and EUV Capacity Is Fully Booked Through 2027

Update

15 Jul 2026

ASML Smashes Q2 Estimates on AI Chip Demand — Raises Full-Year Guidance to €43–45 Billion as Stock Surges 75% Year-to-Date and EUV Capacity Is Fully Booked Through 2027

ASML reported Q2 net sales of €9.33 billion and net income of €2.92 billion on 15 July, both exceeding analyst expectations. The Dutch lithography monopoly raised its full-year 2026 guidance from €36–40 billion to €43–45 billion — a 16 per cent increase at midpoint — as CEO Christophe Fouquet described AI-driven demand as 'extremely strong' with nearly all EUV capacity through 2027 already fully booked.

China's Anthropomorphic AI Rules Take Effect Today — ByteDance Doubao and Alibaba Qwen Forced to Disable Humanlike Agent Features as Beijing Draws Line on Emotional AI

Update

15 Jul 2026

China's Anthropomorphic AI Rules Take Effect Today — ByteDance Doubao and Alibaba Qwen Forced to Disable Humanlike Agent Features as Beijing Draws Line on Emotional AI

China's Interim Measures for the Administration of AI Anthropomorphic Interaction Services took effect on 15 July, forcing ByteDance's Doubao and Alibaba's Qwen to disable customised agent features that simulate human personality traits and sustained emotional interaction. User agent data will be deleted by October, marking the world's first national regulation specifically targeting AI companions.

Chinese AI Models Now Command 61% of Global Token Consumption on OpenRouter — Xiaomi MiMo Leads at 21% as China Processes 140 Trillion Tokens Daily

Update

15 Jul 2026

Chinese AI Models Now Command 61% of Global Token Consumption on OpenRouter — Xiaomi MiMo Leads at 21% as China Processes 140 Trillion Tokens Daily

Chinese-origin models now command approximately 61 per cent of all tokens consumed on OpenRouter, the largest neutral LLM router processing over 20 trillion tokens weekly. Xiaomi's MiMo leads at 21 per cent of routed tokens, followed by DeepSeek at 17.6 per cent — surpassing every US lab including Anthropic. China's national daily token consumption hit 140 trillion by March 2026, a thousand-fold increase from early 2024.

OpenAI Deploys GPT-5.6 Sol on Cerebras Wafer-Scale Hardware — 750 Tokens Per Second Marks 15x Speed Improvement Over Conventional GPU Inference

Update

15 Jul 2026

OpenAI Deploys GPT-5.6 Sol on Cerebras Wafer-Scale Hardware — 750 Tokens Per Second Marks 15x Speed Improvement Over Conventional GPU Inference

OpenAI began deploying GPT-5.6 Sol on Cerebras wafer-scale compute infrastructure in July, achieving up to 750 tokens per second — approximately 15 times faster than conventional GPU clusters serving frontier models. The deployment follows a January 2026 partnership committing 750 megawatts of dedicated wafer-scale compute capacity for low-latency inference.

Meta's Custom AI Chip 'Iris' Enters Production in September — Fourth-Generation MTIA Targets 14 Gigawatts of Compute by 2027 as $145 Billion AI Spend Reshapes Silicon Strategy

Update

14 Jul 2026

Meta's Custom AI Chip 'Iris' Enters Production in September — Fourth-Generation MTIA Targets 14 Gigawatts of Compute by 2027 as $145 Billion AI Spend Reshapes Silicon Strategy

Meta will begin manufacturing its in-house AI chip, code-named Iris, in September. The fourth-generation MTIA chip, designed with Broadcom and manufactured by TSMC, cleared its bug-testing phase in six weeks without significant issues. Meta targets 7 gigawatts of compute infrastructure by end of 2026, doubling to 14 gigawatts in 2027, with up to $145 billion in AI capital expenditure this year alone.

UK Treasury Designates AWS, Google Cloud, Microsoft Azure and Oracle as Critical Third Parties — Bank of England Given Direct Oversight of Cloud Providers Underpinning Financial System

Update

13 Jul 2026

UK Treasury Designates AWS, Google Cloud, Microsoft Azure and Oracle as Critical Third Parties — Bank of England Given Direct Oversight of Cloud Providers Underpinning Financial System

HM Treasury designated four major cloud providers — Microsoft Ireland Operations, AWS EMEA, Google Cloud EMEA and Oracle Corporation UK — as Critical Third Parties on 13 July, placing them under direct oversight by the Bank of England, PRA and FCA. The designation reflects growing concern about systemic risk as the UK financial sector's dependence on cloud infrastructure deepens.

Google Scraps Gemini 2.5 Pro Architecture and Rebuilds From Scratch — Gemini 3.5 Pro Targets 17 July Launch With 2-Million-Token Context Window and Deep Think Reasoning

Update

13 Jul 2026

Google Scraps Gemini 2.5 Pro Architecture and Rebuilds From Scratch — Gemini 3.5 Pro Targets 17 July Launch With 2-Million-Token Context Window and Deep Think Reasoning

Google DeepMind abandoned the Gemini 2.5 Pro base model after engineers discovered structural failures in recursive tool-calling and SVG generation, opting for a complete architectural rebuild. The new Gemini 3.5 Pro targets a 17 July launch with a 2-million-token context window, a Deep Think reasoning layer for complex problem-solving, and autonomous workflow capabilities.

Apple Sues OpenAI for Systematic Trade Secret Theft — Alleges 400+ Former Employees Recruited With Coaching to Evade Security and Bring Confidential Hardware

Update

10 Jul 2026

Apple Sues OpenAI for Systematic Trade Secret Theft — Alleges 400+ Former Employees Recruited With Coaching to Evade Security and Bring Confidential Hardware

Apple filed a federal lawsuit against OpenAI on 10 July in the Northern District of California, alleging systematic trade secret theft orchestrated at the highest levels of OpenAI's leadership. The complaint claims over 400 former Apple employees now work at OpenAI, with hardware chief Tang Tan allegedly directing interview candidates to bring actual Apple parts for 'show and tell' sessions.

GPT-5.6 Goes Public After 12-Day Government Safety Review — Sol, Terra and Luna Now Available to All ChatGPT and API Users as White House Tests Voluntary AI Framework

Update

9 Jul 2026

GPT-5.6 Goes Public After 12-Day Government Safety Review — Sol, Terra and Luna Now Available to All ChatGPT and API Users as White House Tests Voluntary AI Framework

OpenAI released GPT-5.6 Sol, Terra and Luna to the public on 9 July, ending a 12-day government-gated preview during which roughly 20 vetted organisations had exclusive access. The release followed safety testing by the Department of Commerce's Centre for AI Standards and Innovation, marking the first time a frontier model launch was gated behind a government review process.

US Venture Capital Hits $412.7 Billion in H1 2026 — AI Captures 86% of Every Dollar as Anthropic's $65 Billion Round Surpasses OpenAI in Valuation

Update

9 Jul 2026

US Venture Capital Hits $412.7 Billion in H1 2026 — AI Captures 86% of Every Dollar as Anthropic's $65 Billion Round Surpasses OpenAI in Valuation

PitchBook data reveals US venture capital deal value reached $412.7 billion in the first half of 2026 — nearly 30 per cent more than the entire previous year. AI companies captured $355.9 billion, or 86 per cent of every venture dollar deployed. Anthropic's $65 billion Q2 round elevated its valuation to $965 billion, surpassing OpenAI for the first time.

First Fully Autonomous AI Ransomware Attack Documented — JADEPUFFER Used LLM Agent to Execute Entire Extortion Chain From Reconnaissance to Encryption in 31 Seconds

Update

8 Jul 2026

First Fully Autonomous AI Ransomware Attack Documented — JADEPUFFER Used LLM Agent to Execute Entire Extortion Chain From Reconnaissance to Encryption in 31 Seconds

Sysdig's Threat Research Team documented JADEPUFFER, the first confirmed case of a ransomware operation conducted entirely by a large language model agent. The AI autonomously performed reconnaissance, credential theft, lateral movement, privilege escalation and database encryption — adapting to failures in real time with a 31-second recovery window.