Thursday, March 26, 2026

Gemini 3.1 Flash Live: Making audio AI more natural and reliable

Google has introduced Gemini 3.1 Flash Live, its highest-quality audio model designed for natural and reliable real-time dialogue. This model enhances speed and tonal understanding, allowing developers and enterprises to create more intuitive voice-first applications. With its availability across Google products and support for over 200 countries, it aims to transform user interactions in various environments.

Meta gets ready to launch two new Ray-Ban AI glasses

Meta, in collaboration with EssilorLuxottica, is preparing to launch its next-generation Ray-Ban AI glasses, the Scriber and Blazer, following recent FCC filings. These filings indicate that the glasses are production-ready, and the new models may feature significant hardware upgrades, including improved connectivity for enhanced AI functionalities.

Investing Without Training Wheels: Are Unsupervised Teen Brokerage Accounts Really a Smart Idea?

Fidelity and Charles Schwab have launched unsupervised brokerage accounts for teenagers aged 13 and older, allowing them to trade stocks and ETFs independently without needing parental approval for each transaction. This initiative aims to foster financial literacy and capitalize on the power of compound growth, as illustrated by the potential for a $1,000 investment at age 13 to grow to $142,043 by age 65 at a 10% annual return. However, the lack of required parental oversight raises concerns about the risks of impulsive trading and the development of speculative habits among inexperienced young investors.

ByteDance’s new AI video generation model, Dreamina Seedance 2.0, comes to CapCut

ByteDance has announced the rollout of its new AI video generation model, Dreamina Seedance 2.0, in its editing platform CapCut. This model enables creators to draft and edit video and audio content using simple prompts or reference materials, and is initially available to users in select countries, including Brazil and Indonesia. The company is also implementing safety measures to prevent copyright infringement as it expands the model's availability.

Israel says it has killed Iran naval chief overseeing Strait of Hormuz blockade

Israel's Defense Forces confirmed the killing of Iranian naval commander Alireza Tangsiri in a precision airstrike in Bandar Abbas, attributing to him the orchestration of efforts to blockade the vital Strait of Hormuz. The U.S. Central Command supported this claim, asserting that his death contributes to regional security, while Israeli Prime Minister Benjamin Netanyahu reiterated the country's commitment to targeting Iranian military leadership.

Wikipedia bans AI-generated articles

Wikipedia has officially banned editors from using AI to write or rewrite articles, a decision made to uphold its core content policies. While the English version of the site will no longer permit AI-generated content, editors can still utilize AI for basic copy editing and translations, provided they verify the accuracy of the original text. This policy change follows months of community discussions and aims to address the challenges posed by AI-written articles.

Major outgoing CEOs are citing AI as a factor in their decisions to step down

Coca-Cola CEO James Quincey and former Walmart CEO Doug McMillon have both indicated that the rise of artificial intelligence played a significant role in their decisions to resign. Quincey, who has led Coca-Cola since 2017, stated that it was time for someone else to lead the company into the next phase of growth, while McMillon, who stepped down after nearly a decade, expressed the need for a leader who could accelerate AI-driven transformations at Walmart.

Defense technology startup Shield AI valued at $12.7 billion in latest funding round

Defense technology startup Shield AI has raised $2 billion in a Series G funding round, achieving a valuation of $12.7 billion. Led by Advent International and JPMorganChase, the funding will support the expansion of its artificial intelligence-powered software, particularly in response to increased demand for autonomous technologies in military operations. The company also plans to acquire Aechelon Technology to enhance its simulation capabilities for military applications.

Conntour raises $7M from General Catalyst, YC to build an AI search engine for security video systems

Conntour, a startup focused on AI-enhanced video surveillance, has raised $7 million in a seed round from investors including General Catalyst and Y Combinator. The company aims to develop a search engine for security video systems that allows users to query camera feeds using natural language, making it easier for security personnel to find specific footage. With notable clients like Singapore’s Central Narcotics Bureau, Conntour is selective about its customer base, emphasizing ethical considerations in its operations.

Cohere launches an open-source voice model specifically for transcription

Cohere has introduced Transcribe, its first open-source automatic speech recognition model designed for transcription tasks like note-taking and speech analysis. With a lightweight architecture of 2 billion parameters, it supports 14 languages and claims to have a superior word error rate of 5.42, outperforming several established competitors. The company plans to integrate Transcribe into its enterprise platform, North, and make it freely available through its API.

Webtoon is adding AI localization tools to its comics platform

Webtoon is set to enhance its Canvas platform by introducing AI localization tools aimed at helping comic creators reach a broader audience and increase their earnings. The new features will include an AI-powered translation tool that allows creators to localize their work into multiple languages, starting with a beta release for English-speaking users this spring. Additionally, Webtoon will update its dashboard to offer improved analytics and expand its ad revenue share program to all creators.

A Google AI breakthrough is pressuring memory chip stocks from Samsung to Micron

Google's recent introduction of TurboQuant, a new compression method, is causing a significant decline in memory chip stocks, with shares of major players like SK Hynix and Samsung falling sharply. The technology promises to reduce memory requirements for large language models by six times, raising concerns about future demand for memory chips, even as analysts caution that this development may ultimately lead to increased demand for more advanced AI capabilities.

Daily Spotlight: Stocks Can Survive a Pullback

The stock market has recently experienced a 5% pullback, with the S&P 500 dropping below 6,500 amid rising oil prices and geopolitical tensions related to the war in Iran. Such pullbacks, which occur about three times a year, do not necessarily indicate a market downturn; historically, stocks have rebounded within a month after such declines. Current market fundamentals, including economic growth and falling interest rates, suggest a resilient outlook despite recent volatility.

OpenAI halts "Adult Mode" as advisors, investors, and employees raise red flags

OpenAI has decided to pause the development of its erotic chatbot, known as 'Adult Mode', due to significant concerns raised by employees, investors, and its well-being advisory board. The board unanimously opposed the project, citing potential societal harms, while technical issues with age verification, which misidentified minors as adults in 12% of cases, further complicated the initiative. With a user base of 100 million underage users weekly, OpenAI is now prioritizing long-term research on the implications of sexual AI content and exploring alternatives focused on productivity.

Vercel Releases JSON-Render: A Generative UI Framework for AI-Driven Interface Composition

Vercel has unveiled json-render, an open-source framework that enables AI models to generate structured user interfaces from natural language inputs. Released under the Apache 2.0 license, json-render allows developers to define a catalog of components and actions, which an AI then uses to create a JSON specification. This innovative approach, described by Vercel CEO Guillermo Rauch as a 'very disruptive technology', integrates AI directly into the rendering layer, facilitating a new method of UI composition.

GitHub will use Copilot interaction data to train AI models starting April 2026

Beginning April 24, 2026, GitHub will implement a new data policy for its Copilot service, allowing the collection of user interaction data from Free, Pro, and Pro+ plans to train AI models, unless users choose to opt out. This data will include prompts, outputs, code snippets, and user feedback, with the aim of enhancing model performance through real-world usage insights. Notably, Copilot Business and Enterprise customers will remain unaffected by this change.

ARC-AGI-3 offers $2M to any AI that matches untrained humans, yet every frontier model scores below 1%

The ARC Prize Foundation has introduced the ARC-AGI-3 benchmark, which challenges AI systems in interactive game environments that untrained humans can navigate easily. Despite the $2 million prize for any AI that can match human performance, all tested frontier models, including Gemini 3.1 Pro Preview and GPT 5.4, scored below 1 percent, highlighting the significant gap in adaptability between AI and human intelligence.

Building shared coding guidelines for AI (and people too)

As software engineering evolves, coding agents are taking on more responsibilities, leading to a need for clear coding guidelines that align with existing standards. Unlike human developers, these agents require explicit instructions to navigate the complexities of a shared codebase, making the quality of documentation paramount for maintaining consistency and reducing errors in the code they generate.

WhatsApp joins the appalling trend for AI-written replies to messages

WhatsApp has introduced a controversial new feature that generates AI-powered suggested replies to messages, echoing similar functionalities in Apple's iMessage. This update, part of a broader rollout, aims to streamline communication but has drawn criticism for promoting less genuine interactions. Alongside this, users can now manage two accounts on iPhone and utilize AI tools to enhance photos directly in chats.

Samsung now rolling out AirDrop support to Galaxy S26 devices in the US

Samsung has begun rolling out AirDrop support for its Galaxy S26 series devices in the US, expanding from initial launches in Korea. The update, which is part of the Quick Share feature, allows users to share files seamlessly with other devices. Reports indicate that users across various states, including California and Texas, are already seeing the update go live, enabling AirDrop functionality by default.

Apple details 4 new ways it’s bringing product component manufacturing to the US

Apple is significantly increasing its U.S. manufacturing efforts by adding four new partners to its American Manufacturing Program (AMP). The company plans to invest $400 million through 2030 with Bosch, Cirrus Logic, TDK, and Qnity Electronics, aiming to produce essential components domestically and create jobs, while enhancing its supply chain resilience.

Moonshot Is Said to Consider Hong Kong IPO as AI Stocks Flourish

Moonshot AI, the Beijing-based developer of the chatbot Kimi, is reportedly exploring an initial public offering in Hong Kong as interest in AI stocks surges. The company has initiated discussions with major financial institutions, including China International Capital Corp. and Goldman Sachs, although the timing of the IPO remains uncertain.

Green IT: How to Reduce the Impact of AI on the Environment

AI's rapid growth is creating substantial environmental challenges, as each query demands significant energy and contributes to hardware churn, with GPU chips lasting only 2-3 years. Ludi Akue, speaking at QCon London, argues that current regulatory frameworks like the EU AI Act are inadequate in enforcing sustainability measures. She advocates for integrating sustainability into the design of AI systems through model compression, quantization, and a cultural shift towards transparency in AI's environmental costs.

Apple’s rumoured 200MP camera said to have larger sensor

Apple is reportedly developing a 200-megapixel camera for its future iPhones, potentially debuting in 2027 or 2028. The new camera aims to address pixel density issues by utilizing a larger sensor size of 1/1.12 inches, which could enhance low-light photography. This move follows a trend where Apple has previously prioritized sensor quality over sheer megapixel counts, ensuring that the camera remains competitive in various lighting conditions.