Tech news digest for Saturday, August 8, 2026

Google’s top hacker hunter explains why hacking groups get codenames

Google has introduced a new naming convention for hacking groups, moving away from the APT system previously used by Mandiant. The updated format features a memorable first name and a second word indicating the group's country of origin, such as Castle for China and Neptune for North Korea. This change, according to Shane Huntley, CTO of Google Threat Intelligence Group, aims to improve clarity for cybersecurity professionals, as Google currently tracks over 5,000 activity clusters worldwide. The consistent naming is crucial for organizations to understand threats and respond effectively.

Google's AI claimed Flock cameras hold $650 in gold and 23 pounds of copper — It was duped by meme

On August 5, Google's AI erroneously reported that Flock license plate cameras contained up to 23 pounds of copper and several grams of gold, valued at approximately $650. This misinformation stemmed from a joke circulating among privacy activists, suggesting the cameras held significant scrap value. Despite the Flock cameras weighing only about 3 pounds, the AI repeated the meme as fact, citing unreliable sources including an anonymous Substack post and an AI-generated Instagram post. By August 6, the AI had corrected itself, stating that the claims were based on internet memes and not factual evidence.

• Gold was trading at approximately $4,350 per ounce on August 7, 2026.

Next big push in ETF industry? Why these risk assets are gaining traction as interest rate uncertainty persists

Collateralized loan obligations (CLOs) are emerging as a significant focus in the ETF industry, driven by investor demand amid ongoing interest rate uncertainty. VettaFi's Todd Rosenbluth highlighted that CLOs, which pool floating-rate secured loans, offer stability and attractive yields. The Federal Reserve's recent decision to maintain interest rates has further fueled interest in fixed income ETFs. Companies like Reckoner Capital Management are responding by launching new CLO ETFs, indicating a robust innovation trend in this sector.

• AAA-rated CLO tranches have near-zero default rates, but lower-tier tranches face risks.

Here’s how we played the massive rebound in AI stocks this week

AI stocks experienced a significant rebound this week, driven by a combination of corporate earnings and a weaker-than-expected July jobs report. The Dow Jones Industrial Average and S&P 500 both reached record highs, with gains of 3% and 3.6%, respectively. The Nasdaq Composite surged 5.2%, now just 1.5% below its all-time closing high. The unexpected drop of 23,000 in nonfarm payrolls eased fears of a Federal Reserve interest rate hike in September, leading to lower Treasury yields and increased investor confidence in AI infrastructure stocks, particularly following the forced liquidation of the Situational Awareness hedge fund last week.

• Nonfarm payrolls unexpectedly fell by 23,000 in July, reducing rate hike expectations.
• The iShares Semiconductor ETF (SOXX) rose by 7.6% for the week, outperforming the Dow.

Here’s why Apple is skipping its M6 Pro and M6 Max chips to accelerate M7 launch

Apple is skipping the M6 Pro, M6 Max, and M6 Ultra chips to accelerate the development of its M7 chip, which is set to launch in spring 2027. The only chip from the M6 family to be released will be the base M6, while redesigned MacBook Pros will debut this fall with M5 Pro and M5 Max chips. This strategic shift aims to enhance Apple's capabilities in on-device AI tasks, with the M7 Ultra expected to significantly improve AI performance, potentially rivaling dedicated accelerators like Nvidia's Blackwell.

• The M7 chip is expected to launch in spring 2027, ahead of the usual schedule.
• The M7 Ultra chip may support up to 1.5TB of RAM, matching the 2019 Mac Pro.

Apple @ Work: What IDC’s latest shipments mean for the MacBook Neo and Air in the enterprise

In a challenging economic climate for the PC industry, Apple stands out with a 10% increase in MacBook shipments, while overall shipments across the sector have declined nearly 5%. This growth is attributed to Apple's effective supply chain management and the popularity of its M series chips. Despite rising prices and component shortages driven by demand from AI data centers, Apple is the only major manufacturer experiencing shipment growth, contrasting sharply with competitors like HP, Dell, and Lenovo, which have all reported negative unit growth. As enterprises navigate these market conditions, many are shifting towards longer refresh cycles for hardware purchases.

• IDC predicts further price hikes into 2027 as inventory concerns rise.

Anthropic sets Claude Code to Auto Mode by default to protect developers from bad approvals

Anthropic's Claude Code will transition to Auto Mode by default on August 14, 2026, allowing the AI coding tool to operate autonomously without requiring manual approvals at each step. This mode includes a classifier that only seeks confirmation for actions deemed dangerous or irreversible. In tests involving 1,053 paid testers, Auto Mode demonstrated superior safety, catching 89% of dangerous commands compared to just 13.6% by human reviewers. Additionally, teams utilizing Auto Mode generated approximately 25% more pull requests, indicating increased productivity.

• Starting August 14, Auto Mode will be enabled by default for Pro, Max, and Team plans.
• Claude Code is currently the most widely used AI coding tool.

Fields Medalist who published a paper on AI-driven human extinction now works for OpenAI

Jacob Tsimerman, recently awarded the Fields Medal, is joining OpenAI to work on AI safety, emphasizing the need for increased focus on the technology's risks. He argues that AI is a transformative force and that mathematicians can significantly contribute to understanding its implications, as current AI systems operate largely on empirical data without solid guarantees. Tsimerman previously published a paper discussing 'omnicide events,' scenarios where AI could lead to human extinction. He believes that while panic is not warranted, a serious assessment of AI risks is essential.

• Former DeepMind CEO Demis Hassabis commented on AI's math capabilities, noting they aren't yet groundbreaking.
• OpenAI's latest model, Astra, has struggled with solving Millennium Prize Problems.

xAI's Imagine Image 2.0 lands just behind OpenAI's GPT-Image-2 in Arena benchmarks

xAI has unveiled Imagine Image 2.0, a new model that incorporates advanced editing tools and preconfigured templates, ranking second in the Arena benchmarks just behind OpenAI's GPT-Image-2. As of August 7, 2026, Imagine Image 2.0 achieved an Elo rating of 1,439 in the Image Edit Arena and 1,320 in the Text-to-Image Arena. The model is designed to follow instructions with high accuracy, maintain clean typography, and ensure consistency across multiple generations. API access is expected to be available soon, expanding its usability across platforms.

• The model features tools like Magic Wand for selective editing and Smart Resize for aspect ratio adjustments.
• xAI plans to release API access for Imagine Image 2.0 in the near future.