Welcome to the BrevFeed daily tech briefing for Friday, August 14. We've got 13 stories for you today across A.I., security, software, cloud, hardware and startups. Let's get into it.

In AI news, Anthropic has launched Claude Science, an AI workbench designed to streamline scientific research. This platform integrates NVIDIA's BioNeMo Agent Toolkit, allowing researchers to use GPU-accelerated AI for tasks like drug discovery and genomics. Claude Science aims to consolidate various tools and datasets into a single environment, making scientific workflows more efficient.

Concurrently, Anthropic has expanded its Claude Cowork tool to mobile and web platforms. Originally a desktop app, Cowork now allows users to manage tasks across devices, with sessions running in the cloud. This expansion reflects a shift in usage, as 90% of Claude Cowork sessions are for non-coding tasks, indicating its growing role as a general productivity enhancer for administrative and content creation.

These developments highlight Anthropic's strategy to deepen its impact in both specialized fields like life sciences and broader productivity sectors. By offering advanced AI tools for scientists and extending task management capabilities, Anthropic is positioning itself competitively with integrated, AI-driven solutions.

In AI, Google has launched new Gemini Flash models, including Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. These models are designed for efficiency and cost savings, aiming to improve token usage and coding accuracy for developers and businesses.

However, the anticipated Gemini 3.5 Pro model has been delayed. Google cited underperformance in coding capabilities compared to rivals like OpenAI and Meta as the reason. This delay has reportedly led to a four percent drop in Alphabet's share price.

The delay highlights the intense competition in the AI sector, putting pressure on Google to accelerate its innovations. While the new Flash models show Google's focus on optimization, the 3.5 Pro delay underscores challenges in maintaining competitive parity in the rapidly evolving AI market.

Turning to security, the Los Angeles Police Department has ended its contract with Flock Safety, a company that provides automated license plate reader technology. The decision comes after the LAPD raised significant concerns about civil liberties, privacy, and how data collected by the system is shared.

An audit revealed 161 wrongful vehicle stops over two months, where the ALPR system incorrectly flagged vehicles as stolen. This raised questions about the technology's reliability and security. The LAPD also cited disputes over data ownership and the sharing of information with other agencies, including ICE, as key factors in their decision.

The LAPD's Chief Information Officer, Dean Gialamas, highlighted these concerns, emphasizing the need for clearer data ownership and stricter privacy terms in future contracts. This move by one of the nation's largest police forces reflects a growing national debate over surveillance practices and data privacy.

In AI, the robotaxi race is heating up as both Tesla and Waymo announce significant expansions of their autonomous driving services across the U.S.

Tesla has launched its robotaxi service in three new Florida cities: Miami, Orlando, and Tampa. This move follows earlier expansions in Texas, as the company continues to grow its presence in the autonomous ride-hailing market.

Meanwhile, Waymo is also expanding, bringing its driverless operations to San Diego, Las Vegas, Tampa, and Denver. Initially, these services will be available to Alphabet employees, with a public rollout expected later. Waymo currently operates in over 10 U.S. cities and maintains a strong lead in the autonomous taxi sector.

These expansions highlight the intense competition in the autonomous vehicle industry, as both companies vie for market share and technological leadership in the evolving urban mobility landscape.

In AI, AMD has announced a strategic partnership with Anthropic, committing up to five billion dollars to enhance Anthropic's computing infrastructure. This deal includes deploying up to two gigawatts of AMD's Instinct MI450 Series GPUs within Anthropic's systems.

The first gigawatt of these GPUs is expected to be deployed in the first half of 2024. This collaboration also involves a multi-year engineering effort, with AMD integrating Anthropic's AI model, Claude, into its software development processes.

This partnership highlights AMD's efforts to compete with other major players in the AI hardware market and strengthens Anthropic's capabilities as it expands its AI infrastructure.

In AI, the way models are priced based on tokens is drawing criticism for being misleading. Different AI models use unique tokenization methods, meaning the same text can result in varying token counts and, ultimately, different costs, even if the per-token rate seems similar.

This complexity was highlighted when DeepSeek cut the price of its V4-Pro model by 75 percent. While this sounds like a big saving, a lower token rate doesn't always guarantee lower costs, especially for advanced AI workflows that can consume many more tokens per operation.

However, new research from Writer shows that optimizing the orchestration layer, or 'AI harness,' around foundation models can significantly reduce token usage. This approach can cut token spend by up to 40 percent without sacrificing accuracy, offering a more cost-effective way to deploy AI.

In startup news, China's largest memory chipmaker, ChangXin Memory Technologies, or CXMT, is seeing a massive valuation discrepancy between its upcoming IPO and crypto trading platforms. On the decentralized exchange Hyperliquid, crypto traders are valuing CXMT at an astonishing $425 billion.

This figure is significantly higher than the company's official IPO valuation of $86 billion for its Shanghai listing, which is scheduled for next Monday. The $425 billion valuation on Hyperliquid is driven by perpetual futures contracts, with shares trading near $6.35, and peaking recently at $8.60.

Offshore investors are using these crypto derivatives to gain exposure to CXMT because they face restrictions on participating directly in the Shanghai IPO. Analysts suggest this premium reflects both the scarcity of access and strong conviction in CXMT's business, with the crypto market essentially forecasting a potential opening price for the stock.

In development, Kubernetes, known for managing containerized applications, is now being considered for desktop infrastructure. This move aims to unify operations and reduce costs associated with older virtual desktop systems, which have traditionally been managed separately from cloud-native models.

Legacy virtual desktop systems operate on pre-allocated virtual machine pools and custom management, keeping them isolated from modern cloud-native platforms. By bringing desktops into the Kubernetes framework, organizations hope to streamline operations and integrate them into a unified architecture.

However, while Kubernetes simplifies application deployment, it also highlights the complexities of database management. Many developers are skilled in application development but often lack the specialized expertise needed for database operations, creating a new challenge for DevOps teams.

In AI, Google's Gemini assistant may soon offer users the ability to customize its voice. A new feature, found in the latest beta of the Google app, suggests users will be able to adjust characteristics like Energy, Formality, Speed, and Warmth.

These parameters could be set to low, medium, or high increments, giving users more control over how Gemini sounds. This personalization would apply to both Gemini Live and other chat experiences.

This potential update follows Apple's introduction of similar voice customization options for Siri in iOS 17, indicating a growing trend towards more adaptable voice AI solutions. While Google hasn't officially announced the feature, it points to a future of enhanced user experience through greater personalization.

In hardware, Apple is reportedly ramping up production of its first foldable iPhone, tentatively named the 'iPhone Ultra'. The company now aims to produce 10 million units by the end of 2026, an increase from earlier estimates, signaling strong demand expectations for the device.

The iPhone Ultra is anticipated to launch in September 2026 with a price tag between $2,300 and $2,500. This premium pricing could lead to an 18% rise in the average cost of foldable smartphones across the market. However, initial availability might be limited, with only 500,000 to 1 million units expected in the first quarter of its release.

Apple's entry into the foldable market is projected to significantly boost global foldable smartphone shipments by 21% year-over-year in 2026. This move is also expected to shift market share, with competitors like Samsung, Huawei, Motorola, and Honor seeing their shares decrease as Apple enters the segment.

In Android news, Google is updating its backup policy starting July 7th. Now, more data types, including SMS messages, call history, and device settings, will count towards your Google Account storage limit.

This change means your Android backup size could increase by about 40 megabytes on average. However, Google is also introducing new controls, allowing users to manage their backup preferences for SMS, MMS, and RCS messages, as well as device settings.

Existing users will see this transition over the next 45 days. Google is providing information via email and support documentation, and users can also upgrade their storage with Google One plans if needed.

Turning to the markets, shares of Samsung Electronics and SK Hynix recently saw significant drops, falling over seven and nine percent respectively. This decline contributed to a broader downturn in South Korea's benchmark Kospi index, where these two companies hold substantial weight.

The sell-off reflects investor concerns about the sustainability of AI-driven growth, even as some companies report healthy profits. Samsung's shares fell despite its profits surpassing those of companies like Nvidia and Apple, indicating a recalibration of growth expectations within the tech sector.

This trend isn't isolated to South Korea; it mirrors a wider global sell-off in semiconductor stocks, including other major Asian firms like SoftBank and Tokyo Electron. The downturn highlights growing investor anxiety about the financial viability of AI innovations and the impact of high valuations across the industry.

In AI, Amazon is enhancing its generative AI assistant, Amazon Quick, to boost efficiency across finance, sales, and supply chain operations. AWS finance teams are already using Quick to save hundreds of hours each month by automating data preparation and analytics, allowing them to focus on strategic analysis.

Sales teams are also benefiting, with Quick helping to reduce administrative tasks like CRM updates and email drafting. This frees up sales representatives to spend more time selling, potentially increasing productivity and deal closures. Companies like 3M and AWS Global Sales are already leveraging Quick for these benefits.

Additionally, Amazon Quick is making waves in business analytics and supply chain management. Tradeshift, which operates in over 70 countries, has migrated to Quick for faster query responses and expanded analytics capabilities. For supply chains, Quick integrates with NVIDIA NeMo to transform data dashboards into actionable recommendations, streamlining decision-making in complex operational environments.

That's the BrevFeed daily briefing. We'll be back tomorrow with the stories that matter in tech. Thanks for listening.