Enterprises may be cautious about adopting agentic artificial intelligence browsers, due to worries about the technology’s autonomy, Palo Alto Networks CEO Nikesh Arora said. “I think unless there are controls built into agentic browsers, which are oriented around credentials and enterprise security, they’re not going to be allowed in enterprises in 24 months,” Arora said. Arora explained that as a consumer, he likes the idea of an agentic browser — one that can, for example, perform tasks like booking plane tickets, making reservations or calling Ubers. However, he said that browsers would need “credentials” from users, and so users’ “ability and desire to give them autonomy becomes important.” He predicted that many major tech companies will start to develop agentic browsers, noting they are already spending billions to create and operate impressive models. But he said agentic browsers are “at odds with the enterprise,” as companies would be wary of the AI agents’ autonomy. Palo Alto Networks’ planned $25 billion acquisition of cybersecurity company CyberArk will help the company provide a solution that will help enterprises protect their privileged information.
UCL and Huawei’s memory-augmented MDP framework lets LLM agents learn in real-time from experience without fine-tuning; reaching 79.40% in GAIA test
A new learning paradigm developed by University College London (UCL) and Huawei Noah’s Ark Lab enables LLM agents to dynamically adapt to their environment without fine-tuning the underlying language model. The method allows agents to continuously improve their performance by using a structured memory system that updates itself as the agent gathers experience. An implementation of the paradigm, which the researchers call Memento, has achieved top scores on key benchmarks for deep research and complex, multi-step reasoning tasks. For enterprises, this offers a scalable and efficient pathway for developing generalist LLM agents that are capable of continuous, real-time learning without the high cost and downtime associated with traditional training methods. Inspired by human memory, the memory-based learning framework that enables continual adaptation without modifying the LLM. Instead of fine-tuning the base model, agents leverage an external memory to store past experiences. When faced with a new task, the agent draws from similar past situations to guide its decision-making. This process builds on the Markov decision process (MDP), a classic framework in AI for teaching an agent to make optimal decisions. The researchers formalize their new approach as a Memory-augmented MDP (M-MDP), which enhances this framework by allowing the agent to consider not just its current state and potential actions, but also a rich memory of past events. The system has three main components: a planner and a tool-enabled executor that work in an alternating loop to complete tasks, and a growing “case bank” that stores past experiences. The more advanced parametric version uses reinforcement learning with a lightweight neural network to address a common real-world challenge: sparse feedback. For tasks where success or failure signals are infrequent, this method helps the feedback “propagate through various stages,” ensuring the agent learns reliably over time.
Retail stablecoin transfers surpass 2024 totals by August, with sub-$250 payments reaching $5.84B; emerging-market users cite fees and delays as adoption drivers
Stablecoin transactions in the retail segment have reached record levels in 2025, with volumes already surpassing last year’s total by August, a CEX.io report noted. Retail-sized transfers, counting transactions under $250, crossed $5.84 billion in August alone, the highest ever recorded, according to data by Visa and Allium cited in the report. With nearly four months left in the year, 2025 has already become the busiest period yet for stablecoin transfer volume at the consumer level. Survey data from emerging markets, asking over 2,600 consumer in Nigeria, India, Bangladesh, Pakistan and Indonesia, reinforced this picture, CEX.io analysts. A majority of respondents said they turned to stablecoins to avoid high banking fees and slow transfers, the report said. Nearly 70% of them reported using stablecoins more frequently than last year, and more than three-quarters expect usage to keep rising, the report said. The distribution of activity among blockchains have shifted, the report noted. The Tron blockchain, traditionally popular for retail transfers due to its low fees and wide support for Tether’s USDT (USDT), has given up market share. Monthly transaction counts fell by 1.3 million, or 6%, and its growth in volume lagged behind its closest competitors. In its place, Binance Smart Chain (BSC) emerged as the top choice for retail users, capturing nearly 40% of retail stablecoin activity, the report said. The network’s transaction count jumped 75% this year with transfer volume rising 67%. Much of the momentum came after Binance delisted USDT in March for European users and a resurgence of memecoin trading on PancakeSwap on BSC. The Ethereum complex, with the base chain and layer-2 networks combined, made up over 20% of transfer volume and 31% of transaction counts, the report noted. While small transfers largely took place on L2s, the mainnet enjoyed a significant rise in the retail segment. Sub-$250 transfers on the mainnet rose 81% in volume and 184% in count.
Baseten positions inference as AI’s foundation, launching Model APIs and Training to speed production, as $100B market accelerates and applications scale to millions
Baseten, the company powering inference for the world’s fastest-growing AI products, announced a $150 million Series D fundraise at a $2.15 billion valuation bringing total capital raised to over $285 million. This fundraise arrives as AI adoption accelerates across sectors, with applications scaling to millions of users in a matter of months. At the center of this surge is inference, the process of running the AI models that power these products and applications. Inherently tied to all AI application revenue, the over $100 billion inference market represents one of the largest and fastest-growing in history. Baseten’s platform is purpose-built for inference, delivering the performance, cost efficiency, and scalability that modern AI applications demand. By uniting applied AI research, flexible infrastructure, and seamless developer tooling, Baseten enables companies operating at the frontier of AI to bring cutting-edge models into production. As firms become increasingly more demanding around the performance, reliability, and economics of their AI products, Baseten’s Inference Stack has quickly become a cornerstone for hypergrowth AI companies. With the Series D funding, Baseten will scale its team and product to meet accelerating customer demand. The company will double down on investing in model performance research, infrastructure, and developer tooling while also expanding its customer teams. These efforts will help the fastest-growing AI companies deliver reliable, high-performance products at scale. “If cloud was the foundation that enabled the latest generation of great technology companies, inference is the foundation for the next,” said Tuhin Srivastava, Co-founder and CEO of Baseten. “Baseten makes that possible. In the same way Stripe became an index of the internet economy, Baseten will become an index of the AI economy.”
Nvidia’s new GPU is designed for long context inference targets video and coding workloads with million‑plus token windows; shipping end‑2026
Nvidia announced a new GPU called the Rubin CPX, designed for context windows larger than 1 million tokens. Part of the chip giant’s forthcoming Rubin series, the CPX is optimized for processing large sequences of context and is meant to be used as part of a broader “disaggregated inference” infrastructure approach. For users, the result will be better performance on long-context tasks like video generation or software development. Nvidia’s relentless development cycle has resulted in enormous profits for the company, which brought in $41.1 billion in data center sales in its most recent quarter. The Rubin CPX is slated to be available at the end of 2026. Inference consists of two distinct phases: the context phase and the generation phase, each placing fundamentally different demands on infrastructure. The context phase is compute-bound, requiring high-throughput processing to ingest and analyze large volumes of input data to produce the first token output result. In contrast, the generation phase is memory bandwidth-bound, relying on fast memory transfers and high-speed interconnects, such as NVLink, to sustain token-by-token output performance. Disaggregated inference enables these phases to be processed independently, enabling targeted optimization of compute and memory resources. This architectural shift improves throughput, reduces latency, and enhances overall resource utilization.
Gen Z embraces “reverse catfishing”: they intentionally downplay their profiles to prioritize authenticity, filter superficial matches, and align online impressions with real‑life interactions
The latest trend of “reverse catfishing” shows that Gen Z is now choosing to value authenticity over performance in terms of finding a partner. Reverse catfishing refers to deliberately underselling yourself online in the hope of finding something more real. It’s quite literally the opposite of catfishing, which involves presenting a false version of yourself and sometimes even a completely different identity online, usually with the aim to attract romantic interest. This leaves many users disillusioned with digital dating and pushes them toward reverse catfishing in pursuit of less performative connections. With dating apps, first impressions often happen digitally. While this can generate interest, it can also lead to a mismatch between online perception and real-life interaction. This is where reverse catfishing can play a crucial role. The findings showed that text-based interactions produced higher levels of perceived social attraction compared to videoconferencing. However, after meeting face-to-face, while social attraction persisted, romantic attraction declined. This clearly suggests that idealized online self-presentation can inflate social impressions but may create disappointment when offline reality does not match expectations. These findings underscore the importance of authenticity in online dating. Reverse catfishing might help ensure that the positive impressions formed online translate into real-world interactions. Intentionally downplaying appearance by using unfiltered photos or presenting themselves more modestly than in real life can help users filter out people who are only interested in surface-level attraction. This approach challenges the swiping-driven culture that prioritizes looks over personality. In a way, it also helps manage expectations by presenting a more authentic version of oneself and can even serve as a subtle safety measure by deterring unwanted attention. In essence, reverse catfishing can act as a tool to counter many of the frustrations and risks experienced by dating app users. Dissatisfaction on dating apps reflects a systemic issue where superficiality and inflated expectations dominate the online dating landscape. In this context, reverse catfishing emerges as more than a quirky trend; it’s more a deliberate strategy to reclaim control.
Enterprise can build inclusive growth without politics: Align workforce with customers, use outcome‑based metrics over quotas, and brand efforts as leadership development and team performance, not DEI
Despite the political backlash against DEI initiatives, the need for inclusive teams is more pressing than ever. Companies that overlook diversity risk missing out on top talent, innovative ideas, and the ability to connect with an increasingly diverse customer base. Across race, gender, generation, income level, and geography, consumers are seeking products and services that reflect their values and lived experiences. Inclusive brands don’t just gain loyalty; they grow market share. We’ve seen this work across industries with both positive and negative outcomes: Costco’s recommitment to DEI policies has won them increased customer loyalty and brand favorability. Nike’s new shift to align with Kim Kardashian for their women’s primer product rather than an elite woman athlete reveals a new strategy and gambles that the trend in hyper-feminism aspiration is more representative of their current female customer base. Fenty Beauty’s wide shade range didn’t just disrupt the cosmetics industry—it redefined what it means to see yourself represented. Target’s decision to withdraw from venture diversity programs is directly impacting its bottom line and customer loyalty. If we can’t use the language of DEI, we must reframe the strategy. Here’s how businesses can continue building diverse, high-performing teams without triggering political backlash: Culture Strategy or Team Optimization: Focus on the measurable business benefits of inclusive culture: reduced turnover, higher engagement, stronger innovation. Customer-Centric Hiring: Align workforce representation with the demographics of your customer base. Inclusive Processes: Embed inclusive principles into hiring, onboarding, promotion, and vendor selection without labeling them as DEI. Manager Effectiveness: Train leaders in inclusive behaviors under the umbrella of leadership development and strategic agility. Outcome-Based Metrics: Track business outcomes like performance, retention, and collaboration quality instead of identity-based quotas.
Virtual credit cards can reshape operational workflows of travel sector by enabling booking platforms, hotel management systems, payment processors, and card issuers to work together to create a unified data and payment flow while also reducing disputes and fraud risk
Vantage Market Research estimates the global virtual card market will triple by 2030. For an industry like travel—where payments move across borders, pass through multiple systems, and often involve fragmented reconciliation—this technology has the potential to reshape both operational workflows and financial strategies. At their core, virtual credit cards work much like traditional corporate credit cards, with one defining difference: they exist only in digital form. A VCC is typically issued for a single transaction or for use within a limited time window. Each is assigned a unique number and can be configured with precise parameters such as a fixed spending limit, an expiration date, or merchant category restrictions. This design offers two important advantages. First, it sharply limits the potential for fraud. If a number is compromised, it becomes useless after the specified transaction or timeframe. Second, it enables transaction-level control, allowing businesses to tie payments directly to specific invoices or bookings. The value of VCCs is maximized when booking platforms, hotel property management systems, payment processors, and card issuers work together to create a unified data and payment flow. DerbySoft’s approach reflects this reality, partnering with established payment technology providers like Conferma and Voxel Group to integrate VCC processing across multiple distribution channels. These collaborations help hotels and distributors operate with greater speed and accuracy, while also reducing disputes and fraud risk. DerbySoft has developed its Payment Connector to help distributors and hotels process VCCs with some of the lowest rates available in the market. The platform also embeds VCC handling directly into the booking workflow, so payment details arrive pre-configured and easy for hotel staff to access, reducing front desk confusion and streamlining reconciliation. This kind of integration aims to balance the security and efficiency benefits of VCCs with the operational realities of running a hotel.
Amazon launches Zoox robotaxi pilot on Las Vegas Strip; offers free rides in purpose-built electric pods with lidar, radar, cameras, 270-degree coverage and bidirectional movement
Five years after acquiring autonomous vehicle startup Zoox Inc, Amazon.com Inc. entered the robotaxi market with the debut of a service in Las Vegas. On debut, rides using Zoox robotaxis are free and available via the Zoox app, which is on both Android and iOS. The service is currently available only on and around the Las Vegas Strip and possible destinations are limited, such as to specific resort and entertainment properties in the area. The idea behind starting the service for free is that it allows riders to become familiar with Zoox and share feedback before the service scales up. Where Zoox gets interesting, particularly compared with Waymo, is that Zoox is not using retrofitted vehicles but instead offers what they call a “purpose-built autonomous pod”. Each Zoox vehicle is fully electric, measures about 3.6 meters (11.8 feet) in length and has a symmetrical, carriage-style design. The interior of the Zoox vehicles, or pods, includes two bench seats that face each other with no driver’s seat, steering wheel or pedals. The design has allowed Zoox to optimize for autonomy, passenger comfort and safety in ways that conventional vehicle conversions cannot achieve. The technology deployed in each pod includes lidar, radar, high-resolution cameras and infrared sensors covering a full 270-degree field of view at each corner of the vehicle. The overlapping array has been designed to eliminate blind spots and allows the pod to detect obstacles, pedestrians and vehicles in challenging conditions, including low light and adverse weather. The pod design also supports bidirectional movement, meaning the pod can drive forward and backward without having to turn around, negating the need for U-turns or reversing.
U.S. Bank is selling to small businesses the ability to use a U.S.-regulated and insured bank to execute payments directly in other countries in local currencies without converting U.S. dollars beforehand, in response to tariff threats
With tariff threats mounting, U.S. Bank is pushing a payment strategy to enable businesses to operate internationally while storing funds inside the U.S. “Tariffs are the most obvious thing that’s been in the news. But the other big issue is the weakening dollar,” Tarek El-Yafi, head of U.S. Bank Global Transaction Services, told American Banker. “That’s a double whammy.” U.S. Bank has built what it calls a “suite” of foreign currency accounts, which enable businesses to hold funds in up to 23 foreign currencies, such as euros, pounds, yen and Australian dollars, while keeping those balances in the U.S. in FDIC-insured accounts. The bank says this enables companies to manage payments to and from suppliers, minimizing currency conversions and foreign exchange volatility. “You can derisk that cash by bringing it back into the U.S.” El-Yafi said. U.S. Bank is particularly focusing on U.S. based companies with cross-border payment needs that want to avoid managing accounts in non-U.S. jurisdictions. Banks and fintechs have applied numerous methods in recent years to enable smaller businesses to move funds between countries while avoiding third parties such as correspondent banks, which add fees for payment processing and heighten foreign exchange risk given the longer settlement times. To avoid this, firms such as Ripple and the bank-led R3 blockchain consortium use distributed ledgers to sell fast processing and lower FX risk for cross-border payments. U.S. Bank is relying on its own scale and ability to support multiple currencies inside the U.S. to sell small businesses on the ability to use a U.S.-regulated and insured bank to execute payments directly in other countries in local currencies without converting U.S. dollars beforehand. The bank is also offering a reporting dashboard to consolidate visibility across global banking accounts. In addition to economic challenges such as higher prices for international supply chains, the bank is bracing its clients for potential retaliatory moves from other countries that may complicate working with banks based outside of the U.S. “You don’t know what these countries are going to do,” El-Yafi said. “They make it harder to do business in that country. “What will that do for middle-market businesses? Will there be a risk of cash being trapped outside of the U.S.?” El-Yafi said. U.S. Bank is also backing up earlier investments in business payments technology. The bank in July began offering working capital loans to business clients through a partnership with fintech Liberis, countering payment firms such as PayPal and Block that use future payment flows to offer fast decisions on short-term loans to merchants. In another July partnership, U.S. Bank teamed with the blockchain-powered WaveBL platform to streamline trade finance by encrypting digital document transfers between trade partners and their banks, reducing the need for paper. And earlier in 2025.