Tech Express | July 22, 2026 Google Ships Three Gemini Flash Models, Pro Still Missing

---

📰 Google DeepMind Releases Three New Gemini Models: 3.6 Flash Leads, Pro Version Still Absent

Google DeepMind released three new Gemini AI models on July 21 — Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber — but the highly anticipated 3.5 Pro remains missing, raising fresh questions about the company's AI roadmap.

Gemini 3.6 Flash is positioned as Google's "workhorse model," delivering improved capabilities in coding, knowledge work, and multimodal performance while reducing token usage by up to 17%, making it cheaper than its predecessor. 3.5 Flash-Lite is the most cost-effective model in its class, designed for lightweight agentic workflows.

3.5 Flash Cyber is a specialized model fine-tuned for detecting and fixing cybersecurity vulnerabilities, available exclusively to governments and trusted partners under a limited access pilot. According to Bloomberg, Gemini 3.5 Pro was delayed after falling short of internal benchmarks, and Google has already begun pretraining Gemini 4.

— TechCrunch

---

📰 Samsung Galaxy Unpacked Today: Flex Titanium Display, Z Fold 8, and Z Flip 8 Debut

Samsung is holding its Galaxy Unpacked event in London on July 22, unveiling the next-generation Galaxy Z Fold 8, Z Flip 8, and an all-new Z Fold Wide variant. Pre-orders open immediately following the event at 2pm local time.

The headline feature is Samsung's new Flex Titanium display technology, which uses a titanium-alloy substrate to make the Z Fold 8's main display thinner and more durable while dramatically reducing crease visibility. The previous Z Fold 7 already had a nearly imperceptible crease, and Samsung claims the eighth generation pushes this even further.

Alongside the foldables, Samsung is also expected to announce new Galaxy Watch smartwatches and Galaxy Buds earbuds. The choice of London over traditional venues like New York or San Francisco underscores Samsung's growing focus on the European market.

— Mashable

---

📰 Tesla Robotaxi Blitz: Orlando and Tampa Go Live Just 18 Days After Miami

Tesla flipped the switch on its Robotaxi service in Orlando and Tampa on July 21, extending its Florida footprint to three metro areas in under three weeks. The blistering pace — just 18 days after the Miami launch on July 3 — far exceeds Tesla's earlier expansion tempo in Texas.

The Florida Robotaxi fleet consists of Model Y vehicles equipped with rear camera washers, hardware exclusive to Tesla's autonomous fleet. Combined with existing operations in Austin, Dallas, and Houston (Texas) plus the San Francisco Bay Area (with safety monitors), Tesla Robotaxi now covers seven U.S. cities.

Tesla's Texas rollout was far more gradual: Austin launched in June 2025 with supervised driving, transitioned to unsupervised in December 2025, and only expanded to Dallas and Houston in April 2026. The compressed Florida timeline signals rapidly growing confidence in the autonomous driving system.

— BASENOR

---

📰 Pratt & Whitney Canada Begins Ground Testing of Hybrid-Electric Propulsion System

Pratt & Whitney Canada announced at the Farnborough International Airshow on July 21 that ground testing of the flight-standard hybrid-electric propulsion system and propeller has begun at its Longueuil, Quebec facility. Upon completion, the system will be installed on a De Havilland Canada Dash 8-100 experimental aircraft, with first flight targeted for 2027.

The hybrid-electric system combines an advanced Pratt & Whitney Canada thermal engine with a 1-megawatt electric motor and controller developed by Collins Aerospace, plus a battery system from Swiss firm H55. The electric motor provides additional power during demanding phases like takeoff and climb, optimizing efficiency throughout the mission. The project targets up to 30% improved fuel efficiency for a typical 250-nautical-mile regional turboprop route.

The project brings together leading aerospace companies including De Havilland Aircraft of Canada, GKN Aerospace, and the National Research Council of Canada, with funding support from the Canadian federal and Quebec provincial governments.

— RTX

---

📰 TSMC to Raise Chip Prices by Up to 10% in 2027, Semiconductor Stocks Rally

TSMC, the world's largest contract chipmaker, plans to increase prices for both advanced and mature chip production by up to 10% beginning in 2027, according to Nikkei Asia. Price negotiations with clients began in June and were finalized in July.

For advanced process nodes at 7nm and below — which accounted for about 77% of TSMC's recent quarterly revenue — base price increases will range from 5% to 10% depending on the customer and product. Additional high-performance computing orders exceeding forecasts could face a 10-15% surcharge, potentially pushing total increases above 10% for some clients. Mature nodes including 12nm, 16nm, and 28nm will also see increases of up to 10%.

TSMC Chairman C.C. Wei emphasized "strategic, not opportunistic" pricing during the July earnings call. The news came as semiconductor stocks rallied broadly on July 21, with AMD, Intel, and Marvell surging 6-9% and the iShares Semiconductor ETF gaining 5.5%.

— MacDailyNews

---

📰 Kai-Fu Lee's 01.AI Launches "C-Suite AI," Aims to Become China's Palantir

01.AI (零一万物), founded by AI pioneer Kai-Fu Lee, unveiled its Wance AI platform and three "C-Suite Decision AI" products — Boss AI, Top Sales AI, and Investment Officer AI — on July 7. The company has completed a strategic pivot from large model pretraining to enterprise AI, targeting a 2027 IPO as "China's first profitable AI 2.0 company."

Lee is a heavy user of Boss AI himself, using it to track meeting commitments, identify operational risk signals, and even ask "what did I say that was inappropriate?" In a company-wide letter five weeks ago, he disclosed that 01.AI achieved 500 million yuan in orders and 250 million in audited revenue in 2025, with orders surpassing 1.5 billion yuan in the first five months of 2026.

Lee identifies three reasons enterprise AI transformations fail: AI doesn't understand the business, companies treat AI as software procurement, and decision-making authority sits with the wrong person — it must be the CEO. Benchmarking against Palantir, Lee declared: "Stop calling us one of the 'Six Tiger Cubs.' Call us the Money Leopard."

— 36Kr

---

📰 China's STAR Market Turns Seven: 600+ Listed Companies, Market Cap Exceeds 13 Trillion Yuan

China's Nasdaq-style STAR Market (科创板) celebrates its seventh anniversary on July 22, 2026. Over seven years, the board has grown to host more than 600 listed companies with a combined market capitalization exceeding 13 trillion yuan, becoming a critical capital market platform for hard-tech enterprises.

By the end of 2025, cumulative R&D spending by STAR Market companies approached 900 billion yuan, maintaining a nearly 13% R&D intensity ratio for seven consecutive years. Cumulative invention patents have surpassed 150,000. The board has become the core funding platform for China's semiconductor, biopharmaceutical, and high-end equipment sectors.

Launched on July 22, 2019, the STAR Market was designed as a "testing ground" for China's registration-based IPO reform. It has nurtured a generation of globally competitive tech companies, playing a key role as a capital catalyst in chip design, innovative drug development, and artificial intelligence.

— China News

---

📰 Meta Replaces Human Moderators with AI: 13% Fewer Errors, But Wrongful Account Bans Spark Outrage

Meta is aggressively replacing human content moderators with AI systems to police Facebook and Instagram, according to the New York Times. The company claims its latest AI moderation tools make 13% fewer errors than human reviewers and detect 10% more policy violations.

However, the transition has triggered widespread user complaints. Numerous legitimate business accounts and personal profiles have been wrongfully banned by AI systems, including a couple with nearly one million followers running an English-teaching business. Making matters worse, the appeals process is also handled by AI, leaving users unable to reach a human. Some accounts were only restored after journalists intervened.

Meta spokesperson Daniel Roberts responded that the accounts examined were banned by older moderation systems, not the company's latest AI tools — an admission that implicitly acknowledges a performance gap between new and old systems. Still, users remain deeply skeptical of an "AI-judged-by-AI" appeals mechanism.

— Axadle Times