Tech Express | July 11, 2026 OpenAI Unleashes GPT-5.6 and ChatGPT Work Agent — The AI Agent Era Begins

---

📰 OpenAI Launches Most Powerful GPT-5.6 Model Family Alongside ChatGPT Work Autonomous Agent

On July 9, OpenAI opened its most powerful system yet to the public — GPT-5.6, offered in three tiers: Sol (flagship), Terra (mid-range), and Luna (fast and affordable). Following a two-week U.S. government safety review, the models are now available across ChatGPT, Codex, and the API. Sol scored 53.6 on the ALE benchmark and 80.0 on the AA Coding Index, with an Ultra mode achieving 91.9% on Terminal-Bench, marking a new frontier in generative AI capabilities.

The same day, OpenAI introduced ChatGPT Work, a cloud-based AI agent designed to autonomously handle complex tasks across email, Slack, and calendars for hours at a time without human intervention. The agent can create documents, slides, spreadsheets, and web apps, support scheduled tasks, and operate desktop browser tools — fundamentally transforming ChatGPT from a Q&A tool into an autonomous workflow executor. This dual launch signals OpenAI's all-out push into the enterprise market.

— VentureBeat

---

📰 Meta's Custom "Iris" AI Chip Enters Production This September, Built by TSMC

According to an internal memo reviewed by Reuters, Meta's in-house AI chip code-named "Iris" cleared testing in just six weeks and is scheduled to enter mass production in September 2026. Co-designed with Broadcom and manufactured by TSMC, Iris is purpose-built to handle AI training and inference workloads more efficiently than off-the-shelf GPUs, reducing Meta's dependence on Nvidia. The company plans to deploy 7 gigawatts of computing capacity this year and double that to 14 GW by 2027.

The move comes amid an unprecedented AI component shortage that has sent chip prices soaring. By bringing chip design in-house, Meta joins the ranks of Google, Amazon, and Microsoft in the race for AI silicon independence. Iris marks a pivotal step in Meta's transformation from a software platform company into a serious chipmaker, while posing a new competitive threat to Nvidia's dominance in the AI accelerator market.

— The Motley Fool

---

📰 Meta Pulls Instagram AI Image Feature After Massive Privacy Backlash From Hollywood

Meta's newly launched Muse Image tool, which allowed users to generate AI images using publicly posted Instagram photos, was abruptly discontinued on July 10 after just days of fierce backlash. Hollywood unions, privacy advocates, and users protested that the feature automatically opted public photos into AI training datasets without consent, raising serious privacy and copyright concerns. Meta admitted it "missed the mark" and made the feature "no longer available."

The swift rollback underscores the growing tension between rapid AI commercialization and user privacy rights. Meta had been aggressively integrating AI across its social media platforms, but the Muse Image debacle reveals the limits of "move fast" culture when applied to AI content generation. The incident may accelerate regulatory efforts globally to impose stricter guardrails on AI use of user-generated content, particularly in the creative industries.

— Los Angeles Times

---

📰 China Temporarily Bans Helium Exports, Stirring Global Chip Supply Chain Fears

On July 10, China's Ministry of Commerce announced an immediate temporary export ban on helium, citing the risk of supply disruptions caused by renewed military conflict in the Middle East. Helium is an irreplaceable gas in semiconductor manufacturing, essential for cooling and creating vacuum environments in EUV lithography systems. While China's share of the global helium market remains limited, the move has nonetheless rattled the chip industry at a time of surging AI-driven demand.

Analysts see dual motivations behind the ban: legitimate domestic supply security concerns and a strategic play to gain leverage in the global semiconductor supply chain. With AI chip demand straining fabrication capacity worldwide, any supply-side disruption carries outsized psychological impact. However, most assessments conclude that China's limited helium market share means the ban is unlikely to cause catastrophic damage to global semiconductor production in the near term.

— Reuters

---

📰 Nubia and ByteDance Unveil World's First AI Agent Smartphone at WAIC 2026

ZTE subsidiary Nubia has officially announced it will showcase the "world's first AI agent phone," jointly developed with ByteDance, at the 2026 World Artificial Intelligence Conference opening July 17 in Shanghai. The device is positioned as a mass-production flagship, featuring a Snapdragon 8 Elite chip, a 6.78-inch LTPO display, and a 6,000mAh battery. Its core differentiator is on-device AI capable of autonomously operating apps, enabling true agent-level human-computer interaction.

ByteDance's Doubao AI assistant will serve as the phone's core AI engine. Notably, Chinese AI startup StepFun is also racing to debut its own AI agent smartphone at the same conference, setting up a "world's first" showdown at WAIC. The competition signals that China's smartphone market is pivoting from hardware spec battles to AI-capability-driven differentiation — a paradigm shift with global implications.

— DigiTimes

---

📰 China's CXMT Launches $4.3 Billion STAR Market IPO to Challenge Global Memory Giants

ChangXin Memory Technologies, China's leading DRAM maker, has kicked off the issuance process for its STAR Market IPO with subscriptions set for July 16. Targeting upwards of $4.3 billion, the listing will be China's largest IPO of 2026. According to Bloomberg, CXMT's revenue grew by a staggering seven-fold in the first half of the year — the fastest in the industry — and founder Zhu Yiming's fortune is expected to soar to $12 billion post-IPO.

CXMT's public debut comes amid a global memory chip shortage and represents a milestone in China's semiconductor self-sufficiency drive. As the only Chinese company with large-scale DRAM production capability, CXMT's rise is reshaping a global memory market long dominated by Samsung, SK Hynix, and Micron. The IPO's success will serve as a key barometer for China's semiconductor independence ambitions and its ability to compete in cutting-edge chip manufacturing.

— TechNode

---

📰 SK Hynix Surges 13% in Nasdaq Debut, $26.5 Billion IPO Ranks as Second-Largest in US History

South Korean memory chip giant SK Hynix made a blockbuster Nasdaq debut on July 10, with its ADS shares closing at $168 — up roughly 13% from the $149 offer price. The $26.5 billion IPO ranks as the second-largest in U.S. stock market history, trailing only SpaceX's June listing, and was 7x oversubscribed, reflecting immense investor appetite for AI infrastructure plays.

SK Hynix is the global leader in HBM high-bandwidth memory, a critical component for AI accelerators like Nvidia's GPUs. With HBM supply remaining severely constrained amid the AI boom, SK Hynix's strategic position has never been stronger. The Nasdaq listing not only raised massive capital for capacity expansion but also gave U.S. investors direct access to a core pillar of the global AI supply chain. Analysts increasingly view memory chips as strategic assets of the AI era, rather than mere cyclical commodities.

— Money Morning

---

📰 Musk Praises Anthropic as "Leader in AI," Vows Not to Cut Off Its Compute Access

In a striking about-face, Elon Musk publicly praised rival Anthropic as "the leader in AI" on X this week, while explicitly promising that SpaceX would not maliciously cut off the AI company's computing resources. The comments stand in sharp contrast to his earlier criticism of Anthropic as "woke" and "hypocritical," and are widely seen as driven by business interests tied to SpaceX's massive compute contract with the company.

Per SpaceX's S-1 filing, Anthropic has agreed to pay $1.25 billion per month through May 2029 for access to SpaceX's computing infrastructure, with either party able to terminate with 90 days' notice. That puts the total potential value of the deal at up to $75 billion over five years. Meanwhile, Musk's own xAI continues to rapidly advance its Grok models, creating a delicate dynamic where Anthropic is simultaneously a top customer and fierce competitor — a relationship that is reshaping the AI infrastructure landscape.

— TechCrunch

---

📰 North American Startup Funding Shatters Records at $392 Billion in H1 2026, Fueled by AI Frenzy

Venture investment in U.S. and Canadian startups reached a staggering $392 billion in the first half of 2026, according to Crunchbase data, dwarfing all previous records. The overwhelming driver of this historic funding surge is artificial intelligence — from foundation model developers to AI infrastructure providers, virtually every AI subsector attracted astronomical levels of capital.

The AI investment fever has spilled into adjacent sectors as well. Canadian cooling technology startup Wafr Technologies recently raised $100 million to develop water-efficient chilling systems for AI data centers, with plans to raise an additional $200 million for U.S. and European expansion. However, alongside the unprecedented funding, bubble concerns are emerging — uncertainty around Fed rate policy and growing scrutiny of AI investment returns could introduce headwinds in the second half of 2026.

— Crunchbase News