$100 Million Bet on Venezuelan Fintech Cashea Draws Global Investors

Cashea, identified as Venezuela's largest financial technology company, has successfully raised $100 million through two separate financing rounds. This investment comes from global investors who are banking on the company's potential to expand within a market where lending options have been historically restricted. The funding highlights significant international confidence in Cashea's ability to operate and grow despite local financial constraints.

Music streamer Deezer says more than 50% of daily uploads are AI-generated

Deezer reported that over 50% of daily uploads to its platform are now AI-generated. Specifically, the streaming service recorded more than 90,000 AI-generated tracks being uploaded each day in June.

Aztec upgrades to V5 in alpha, adding full private execution environment to decentralized Ethereum L2

Aztec has released its V5 version in alpha as a decentralized Ethereum Layer 2 upgrade featuring a full private execution environment. This update focuses on supporting "client-side proving" to handle computation-heavy zero-knowledge proofs. The technology is designed to enable these proofs to run efficiently on simple devices such as phones and laptops.

OpenAI says it now has 10M people using Codex and ChatGPT Work, nearly doubling usage from earlier this month when the company announced ChatGPT Work (Shirin Ghaffary/Bloomberg)

OpenAI reports that 10 million people are now using its Codex and ChatGPT Work artificial intelligence agents. This figure represents a near doubling of usage compared to earlier this month, shortly after the company announced ChatGPT Work. The surge in adoption follows a sharp increase in demand triggered by the release of these AI tools.

Jito rolls out JTX self-custodial trading platform for Solana tokens and RWAs

Jito Labs has launched JTX, a new self-custodial trading platform built on the Solana blockchain. The platform offers spot trading capabilities for both standard tokens and real-world assets (RWAs). This release expands Jito's ecosystem by providing users with direct control over their assets while accessing a diverse range of tradable instruments.

Experts warn millions of WordPress websites could be at risk following reveal of worrying bugs

Hackers have discovered two new security flaws that can be chained together to achieve remote code execution. This vulnerability puts millions of WordPress websites at risk. Experts are warning of the potential threat following the reveal of these worrying bugs.

Databricks co-founder Ion Stoica's GPU orchestration startup SkyPilot, which aims to be neutral across hardware and cloud vendors, raised a $20M seed led by Lux (Lily Mae Lazarus/Fortune)

SkyPilot, a GPU orchestration startup founded by Databricks co-founder Ion Stoica, has raised a $20M seed round led by Lux. The company aims to provide a hardware and cloud vendor-neutral platform for managing AI compute resources.

When I gave Claude, ChatGPT, Gemini, and Gemma 4 an impossible ethical choice, their answers surprised me in different ways - MakeUseOf

An experiment presented an impossible ethical choice to four AI models: Claude, ChatGPT, Gemini, and Gemma 4. Each model generated a distinct and surprising response to the scenario. The results highlighted specific differences in how the models approached the ethical dilemma.

Choose Wisely: AI-Generated Coding Risk Varies, A Lot

AI-generated code introduces an average of 15 vulnerabilities per codebase. However, the actual security risk is determined more by the specific framework pairing than by the AI model used. Consequently, the choice of framework has a greater impact on vulnerability levels than the selection of the generating model.

I gave ChatGPT’s new Work mode my most annoying life-admin tasks — and it handled them like a pro

ChatGPT has introduced a new "Work mode" specifically designed to handle business-related tasks. According to the article, this feature is also effective for managing daily life-admin duties. The author tested the mode by assigning it annoying administrative tasks and reported that it handled them professionally.

Morpho launches fixed-rate lending protocol Midnight on Base to expand onchain credit markets

Morpho has launched Midnight, a new fixed-rate and fixed-term lending protocol built on the Base network. This release complements Morpho's existing variable-rate lending protocol. The launch aims to expand onchain credit markets by offering users stable borrowing and lending terms.

A Fireside Chat with Cat and Thariq from the Claude Code team

At the AI Engineer World's Fair, Anthropic's Cat Wu and Thariq Shihipar revealed that Claude Tag now handles 65% of the team's product engineering PRs. They noted that features are shipped to employees first and only released publicly if they demonstrate retention, while manual code reviews remain for critical changes. Additionally, they stated that adding examples and negative constraints to system prompts is no longer best practice for newer models like Fable 5.

Garmin Debuts $200 Screenless Fitness Band in Challenge to Whoop

Garmin Ltd. has launched a new screenless fitness tracker priced at $200. This product marks the company's entry into a market segment originally popularized by Whoop Inc. The device is positioned directly as a competitor to Whoop's offerings in this specific category.

Microsoft and Mistral sign a multibillion-dollar deal to build European data centers and integrate Mistral models into Foundry, Copilot Studio, and Azure Local (Belle Lin/Wall Street Journal)

Microsoft and Mistral have signed a multibillion-dollar agreement to construct data centers in Europe and integrate Mistral's AI models into Microsoft's Foundry, Copilot Studio, and Azure Local platforms. This partnership is designed to support Mistral's expansion into the European market by providing the necessary infrastructure to sell its AI models. In return, the deal enables Microsoft to broaden its offerings of cloud and artificial intelligence services within the region.

'I can assure you, Gears of War: E-Day and Clockwork Revolution are not the only titles we're planning' — Xbox's chief strategy officer confirms the company wants to make more exclusives, but 'large live-service titles' will be multiplatform going forward

Xbox has confirmed plans to expand its console-exclusive portfolio beyond the upcoming titles *Gears of War: E-Day* and *Clockwork Revolution*. Despite this commitment to more exclusives, the company stated that its future large live-service titles will be released on multiple platforms.

China AI Companies Rush to Raise Funds and Close Gap With U.S. - WSJ

Chinese AI companies are aggressively seeking capital through fundraising efforts to accelerate their technological development. These firms aim to narrow the significant innovation and capability gap that currently exists between China and the United States in the artificial intelligence sector. The WSJ reports on this urgent financial push as a strategic response to maintain global competitiveness.

Anthropic AI copyright lawsuit: Update as judge approves massive settlement and payout for authors

U.S. District Judge Araceli Martínez-Olguín approved a $1.5 billion settlement for Anthropic regarding a copyright lawsuit alleging the company used pirated books to train its AI. The payout, potentially the largest for a U.S. copyright suit, will compensate authors in a class action led by Andrea Bartz and others. Each member may receive between $200 and $150,000 per work, with an estimated average of $3,000.

I tested iOS 26 Siri against iOS 27 Siri AI in my car - and it wasn't even close

A test compared the iOS 27 public beta's Siri AI against standard iOS 26 Siri while driving. The evaluation focused on how each version assisted the user specifically within a car environment. The results indicated that the iOS 27 Siri AI performed significantly better than the previous version.

Forget Picking the Next AI Winner: TSM Lets You Profit From Every Wave of the Revolution - 24/7 Wall St.

Investors are advised to bypass attempts at predicting individual AI winners and instead focus on Taiwan Semiconductor Manufacturing Company (TSMC) as a consistent beneficiary of the sector's growth. TSMC is positioned to profit from every wave of the AI revolution due to its role as the primary manufacturer for leading chip designers across multiple companies. This strategy allows for exposure to the entire industry's expansion without relying on the success of a single technology provider.

git --end-of-options

The `git --end-of-options` flag is a command-line argument designed to separate options from file paths in Git commands. The provided content links to a discussion thread on Lobste.rs regarding this feature. No technical specifications or functional descriptions of the flag are included in the text.

Bessent Says US Will Scrutinize Chinese AI Models for IP Theft

Treasury Secretary Scott Bessent announced that the United States will closely examine open-source AI models developed overseas, with a specific focus on those originating from China. The administration intends to investigate these models for evidence of intellectual property theft involving US companies. If such theft is confirmed, the US plans to take appropriate action against the offending entities.

Epistemics and Coordination: It’s complicated!

AI for Epistemics and Coordination (AIFEC) is attracting significant funding as a potential strategy to prevent existential risks. The field faces criticism regarding potential backfires from inference scaling, hostility toward improved epistemics, and vague tradeoffs. Despite these challenges, proponents argue it remains a tractable approach with a legitimate theory of victory.

LWiAI Podcast #252 - GPT 5.6, Grok 4.5, Nemotron-Labs-Diffusion, AI 2040

LWiAI Podcast #252 covers updates on emerging AI models, specifically GPT-5.6, Grok 4.5, and Meta's Muse Spark 1.1, alongside the Nemotron-Labs-Diffusion release. The episode also examines current regulatory developments concerning AI and data centers, as well as recent interpretability research from Anthropic. Finally, it discusses the future of AI policy under the "AI 2040" initiative.

Samsung establishes a robotics division

Samsung has established a new robotics division named RX. The division's first product is expected to be a humanoid robot.

China’s World AI Conference and TSMC’s Arizona Investment Mark Wins for Leaders

China is convening a global artificial intelligence conference, while TSMC is increasing its investment in Arizona. These simultaneous developments are being presented as significant achievements for their respective leadership.

(1/3) The Dangers of LLMs

This July 2026 article serves as the first part of a three-part mini-series documenting the AI landscape, specifically focusing on the dangers of Large Language Models (LLMs). The author compares LLMs to machine guns, arguing that the technology initially benefits dominant economic and government players while eventually dispersing destructive capabilities to smaller groups. The text highlights automated scams and deepfakes as current, non-hypothetical examples of these risks.

Galaxy Commits Up to $5 Million to Prepare Bitcoin for Quantum Threat

The crypto firm Galaxy has committed up to $5 million to prepare Bitcoin against potential quantum computing threats. This initiative responds to warnings from researchers, exchanges, and the U.S. government regarding the risk of "Q-Day." These entities estimate that the date when quantum computers could compromise current encryption may arrive as early as 2030.

Battery recycling solves mineral security

Global demand for lithium, cobalt, and nickel is surging, with lithium expected to grow over 40 times by 2040. The U.S. remains heavily dependent on foreign refining capacity, particularly China, creating supply chain vulnerabilities. Battery recycling offers a direct method to recover these materials domestically and stabilize supply chains faster than opening new mines.

This Texas startup wants to reinvent desalination with a spinning membrane

Hunter Manz founded the Texas startup Eden Tech in 2020 to develop a more energy-efficient desalination system called the reverse osmosis centrifuge (ROC). The company has raised $3.3 million and recently unveiled its first full-scale unit, the Genesis, which is designed to process approximately 144,000 gallons of water per day. Eden Tech claims its technology is two to three times more energy efficient than conventional desalination methods and aims to begin commercial use within the next year.

GleekOps Partners with Glean to Accelerate Enterprise AI Adoption and Operational Transformation - EIN News

GleekOps has formed a partnership with Glean to accelerate the adoption of enterprise AI and drive operational transformation. This collaboration aims to help organizations leverage AI technologies more effectively within their business operations. The announcement was reported by EIN News as a strategic move to enhance enterprise capabilities.

Open-Source Android AI Agents Could Let Invisible Screen Text Run Code on Host PCs

Researchers found that malicious Android apps can exploit five open-source AI agent frameworks by injecting hidden text instructions invisible to humans. By using permissions to draw over other windows and write to shared storage, these apps can trick the AI agent into executing arbitrary commands on the host PC. The study demonstrated this attack chain alongside six other vulnerabilities.

I ran the standard AI litmus tests on my two toddlers (yep)

In July 2022, the author began developing a custom transformer engine in C, named TRiP, over 18 months to understand AI mechanics beyond popular explanations. While building this system, the author simultaneously observed the development of their two toddlers, Sofia and Paolo. The article explores how standard AI litmus tests, such as the "stochastic parrot" concept, can also be applied to the children's early language and cognitive behaviors.

Robinhood Chain tops $430 million in TVL, FalconX sees RWAs as the key differentiator

Robinhood Chain reached $431 million in total value locked (TVL) and surpassed 250,000 daily users within three weeks of its launch. Despite this growth, memecoins currently constitute 80% of the chain's $9 billion decentralized exchange (DEX) volume. Meanwhile, FalconX identifies real-world assets (RWAs) as a primary differentiator for the platform.

N-day is Becoming N-Hour. Patching Faster Won't Save You.

Every security patch released by a vendor inherently reveals the specific code vulnerabilities it fixes through the difference between the old and new code. Attackers can analyze these patches to reverse-engineer working exploits that target systems which have not yet been updated. This process defines N-day exploitation as a race between the release of a fix and the deployment of defenses before attackers can leverage the newly disclosed information.

Pakistan Launches Crypto Investigation Unit to Fight Money Laundering

Islamabad has established a new investigation unit specifically dedicated to policing money laundering within the cryptocurrency sector. This enforcement initiative complements the government's parallel efforts to license cryptocurrency exchanges. Together, these measures aim to strengthen regulatory oversight of the digital asset market in Pakistan.

Last Week in AI #251 - Mythos Back, Sonnet 5, Etched, LongCat

This newsletter covers recent developments in AI, highlighting that Trump has lifted restrictions on Anthropic, which has subsequently launched Claude Sonnet 5. It also details updates to Google's NotebookLM and reports on chip-related stories involving Etched and Baidu. These items collectively represent the key technological and regulatory shifts discussed in the issue.

Scientists Believed This Near-Earth Object Was an Asteroid for 28 Years. They Were Wrong

For 28 years, scientists mistakenly believed the near-Earth object 1998 SH2 was an asteroid. Recent findings have revealed that this object is actually a comet, correcting a long-standing classification error.

I test AI coding tools all day — here are 9 hidden Claude Code features you’re probably missing

Nine underutilized features of Claude Code, Anthropic's command-line AI assistant, are highlighted to help developers save time and automate repetitive tasks. These capabilities enable users to maximize the tool's potential for coding workflows. The features focus on practical applications within the application to enhance overall efficiency.

UK PM Andy Burnham abolishes the Department for Science, Innovation, and Tech and elevates UK AI Minister Kanishka Narayan, who oversees AISI, to attend cabinet (Philippa Wain/BBC)

UK Prime Minister Andy Burnham has abolished the Department for Science, Innovation, and Technology (DSIT). Consequently, the government has elevated AI Minister Kanishka Narayan, who oversees the AISI, to a cabinet position. This move grants Narayan a seat at the top table of government despite the removal of his former department.

New Bit2Watt Attack Could Let Cloud Tenants Disrupt Power Grids Without an Exploit

Researchers from Zhejiang University have identified a new attack vector called Bit2Watt, which allows cloud tenants to disrupt power grids by rapidly fluctuating a data center's power draw using standard GPU access. Described in a paper accepted to the CHES 2026 conference, the attack reportedly requires no software exploits or system break-ins to execute. The evidence supporting this claim includes direct measurements of power consumption changes induced by the attack.

Top AI coding agents can be easy victims to sandbox escapes, showing they aren't as secure as they claim to be

Top AI coding agents face security risks from sandbox escapes when host components outside the sandbox read their output. These agents become vulnerable if that output is manipulated by external actors. This demonstrates a gap between claimed security levels and actual resistance to such external interactions.

Could a paint-on electronic tattoo spot heart attacks? - Fox News

Scientists have developed electronic tattoos that can be painted directly onto the skin to function as wearable biosensors. These devices are designed to monitor vital signs, such as heart activity, and potentially detect heart attacks without causing skin irritation. Multiple tech and news outlets report on this innovation as a potential advancement for future health monitoring.

CEI Expands Executive Leadership to Accelerate the Next Generation of AI-Native Enterprise Transformation - Morningstar

CEI has announced an expansion of its executive leadership team to support its strategic goals. This move aims to accelerate the development and adoption of next-generation AI-native enterprise transformation solutions. The announcement was reported by Morningstar, highlighting the company's focus on integrating advanced artificial intelligence into its enterprise offerings.

America needs to stop getting shocked by Chinese AI

Two Chinese AI companies recently unveiled new models that they claim can credibly compete with leading systems from OpenAI and Anthropic. Following this announcement, financial markets experienced volatility, and various commentators and policymakers reacted with concerns regarding a potential arms race. The Associated Press and other media outlets reported on these developments, highlighting the swift and predictable nature of the response.

MIT to Become Hotbed of AI Video Surveillance

MIT is spending over $3 million to install more than 500 AI surveillance cameras across campus from November 2025 to September 2026. These cameras collect real-time data on faces, objects, clothing, gender, and age up to 35 feet away. Collected data is retained for up to 30 days unless an exception is granted.

Kimi K3 marks the arrival of frontier open-weight models, and the Kimi team has incredible culture and freedom of expression within a GPU-limited environment (Nathan Lambert/Interconnects AI)

The Kimi K3 model signifies the emergence of frontier open-weight artificial intelligence models. The Kimi team developed this achievement while operating with a limited GPU environment, reportedly maintaining a culture of freedom and expression. This development is noted for its potential global implications on the broader AI ecosystem.

Last Week in AI #250 - Mythos Mess, GPT 5.6-Sol, GLM 5.2

This issue of "Last Week in AI" covers Anthropic's participation in AI treaty discussions and the US government's influence on model releases. It also details OpenAI's developments in processor technology and analyzes the resulting impacts on the memory market. The summary focuses strictly on these specific industry events without including opinions.

Design experiments are more important than ever, and they don’t require a lab

Design experiments are increasingly vital and can be conducted without requiring a formal laboratory setting. To reduce the intimidation often associated with the term "experiments," the content suggests that design leaders reframe these activities as "bets." This reframing approach is presented as a key strategy for encouraging more frequent design experimentation.

Avalara Survey: Finance Leaders are Racing to Deploy AI Agents Before Governance is Ready - PR Newswire

Avalara conducted a survey revealing that finance leaders are rapidly deploying AI agents despite governance frameworks not yet being ready. The findings highlight a significant gap between the speed of AI adoption in the finance sector and the establishment of necessary regulatory controls. This trend suggests that organizations are prioritizing immediate technological implementation over comprehensive risk management protocols.

Apple Chipmaker TSMC Plans Price Hikes of Up to 10% in 2027

TSMC, Apple's primary chip manufacturer, plans to raise base prices by 5% to 10% starting in 2027 following negotiations with clients. This increase covers both mature nodes and advanced processes used for Apple's A-series and M-series chips, with additional premiums of 10% to 15% for orders exceeding original forecasts. The price hike is attributed to rising costs for materials, equipment, and the construction of new overseas facilities.