The Trump administration has proposed a rule to significantly expand drone operations, which could alter America’s shopping habits, boosting retailers like Walmart and Amazon as they expand into delivering consumer packages by autonomous aircraft. The proposal aims to safely integrate drones — technically called unmanned aircraft systems — into the national airspace. Under current rules, operators must seek individual waivers for flights beyond the drone operator’s direct line of visual sight. The Federal Aviation Administration’s (FAA) Bryan Bedford said comments accompanying the rule announcement that the “Beyond Visual Line of Sight” proposal is “key to realizing drones’ societal and economic benefits.” He cited package delivery first, followed by agriculture, aerial surveying, public safety, recreation and flight testing. An FAA fact sheet said that under the proposal, drone operations would occur at or below 400 feet above ground level, from pre-designated and access-controlled locations. Operators would need FAA approval for the areas where they intend to fly, and proposals for a single operator to fly multiple drones would be evaluated on a case-by-case basis.
US-based crypto and fintech companies are expanding their platforms to build ‘super apps’ to offer a broad range of services that include tokenized RWAs, stocks, derivatives, ETF tokens and prediction markets, indicating unification of crypto UX
US-based crypto and fintech companies, including Coinbase, Robinhood, and X, are expanding their platforms to offer a broader range of financial and communication services, as the “super app” model popularized in Asia gains traction in the West. The trend has accelerated as U.S. regulators push forward with crypto-friendly legislation and policies that provide long-awaited clarity to the industry. Coinbase’s head of consumer and business products, Max Branzburg, announced that the largest U.S. centralized exchange (CEX) would soon offer its domestic users access to tokenized real-world assets (RWAs), stocks, derivatives, and prediction markets. Coinbase’s CEO, Brian Armstrong, has told media as far back as 2023 that the CEX is moving toward “super app” status, emulating the model popular in Asia. SEC Chair Paul Atkins praised the idea of crypto-powered “super-apps” and emphasized that this new policy represents more than a regulatory shift — it is a generational opportunity. Robinhood announced that it will offer more than 200 tokenized U.S. stocks and exchange-traded fund (ETF) tokens to users in Europe, with plans to expand that number to 2,000 by the end of 2025. Coinbase and Robinhood are embracing a trend to emulate Asia’s super app model, exemplified by WeChat, which combines payments, messaging, shopping, and other services in one platform. Tesla CEO Elon Musk has expressed a similar goal, aiming to turn X into a global marketplace that combines comprehensive communications with the ability to conduct the entire financial world. The crypto-powered super app trend is seen as confirmation that traditional finance is moving on-chain and that mainstream crypto user experience is moving toward unification, instead of fragmentation across networks, apps, and protocols. However, some industry experts argue that creating an “everything app” with many services would discourage competition and be overwhelming from a UX perspective. The regulatory framework hasn’t caught up yet, and whoever can balance compliance and composability will win the battle.
Google DeepMind’s foundation world model can simulate real world scenarios by remembering what it has generated and reasoning over long time horizons and can be used to train general-purpose AI agents essential to reaching “artificial general intelligence”
Google DeepMind has revealed Genie 3, its latest foundation world model that can be used to train general-purpose AI agents, a capability that the AI lab says makes for a crucial stepping stone on the path to “artificial general intelligence,” or human-like intelligence. Still in research preview and not publicly available, Genie 3 builds on both its predecessor Genie 2 and DeepMind’s latest video generation model Veo 3. Perhaps most importantly, Genie 3’s simulations stay physically consistent over time because the model can remember what it previously generated — a capability that DeepMind says its researchers didn’t explicitly program into the model. While Genie 3 has implications for educational experiences, gaming or prototyping creative concepts, its real unlock will manifest in training agents for general-purpose tasks, which he said is essential to reaching AGI. “We think world models are key on the path to AGI, specifically for embodied agents, where simulating real world scenarios is particularly challenging,” Jack Parker-Holder, a research scientist on DeepMind’s open-endedness team, said. Genie 3 is supposedly designed to solve that bottleneck. Like Veo, it doesn’t rely on a hard-coded physics engine; instead, DeepMind says, the model teaches itself how the world works by remembering what it has generated and reasoning over long time horizons. That memory, the company says, lends to consistency in Genie 3’s simulated worlds, which in turn allows it to develop a grasp of physics, similar to how humans understand that a glass teetering on the edge of a table is about to fall, or that they should duck to avoid a falling object. Notably, DeepMind says the model also has the potential to push AI agents to their limits — forcing them to learn from their own experience, similar to how humans learn in the real world.
Microsoft’s paper discusses closed-domain hallucination detection method, designed to provide traceability for processes with any number of generative steps, through an evidence trail that identifies error stage and verifies each claim separately
Microsoft paper presents VeriTrail, the first closed-domain hallucination detection method designed to provide traceability for processes with any number of generative steps. VeriTrail outperforms baseline methods commonly used for hallucination detection. In this blog post, we provide an overview of VeriTrail’s design and performance. A key idea leveraged by VeriTrail is that a wide range of generative processes can be represented as a directed acyclic graph (DAG). Each node in the DAG represents a piece of text (i.e., source material, an intermediate output, or the final output) and each edge from node A to node B indicates that A was used as an input to produce B. Each node is assigned a unique ID, as well as a stage reflecting its position in the generative process. VeriTrail takes as input a DAG representing a completed generative process and aims to determine whether the final output is fully supported by the source text. It begins by extracting claims (i.e., self-contained, verifiable statements) from the final output using Claimify. VeriTrail verifies claims in the reverse order of the generative process: it starts from the final output and moves toward the source text. Each claim is verified separately. In addition to assigning a final “Fully Supported,” “Not Fully Supported,” or “Inconclusive” verdict to each claim, VeriTrail returns (a) all Verdict Generation results and (b) an evidence trail composed of all Evidence Selection results: the selected sentences, their corresponding node IDs, and the generated summaries. Collectively, these outputs provide traceability: Provenance: For “Fully Supported” and “Inconclusive” claims, the evidence trail traces a path from the source material to the final output, helping users understand how the output may have been derived. Error Localization: For “Not Fully Supported” claims, VeriTrail uses the Verdict Generation results to identify the stage(s) of the process where the unsupported content was likely introduced. Error stage identification helps users address hallucinations and understand where in the process they are most likely to occur. The evidence trail also helps users verify the verdict: instead of reading through all nodes – which may be infeasible for processes that generate large amounts of text – users can simply review the evidence sentences and summaries.
‘White-space’ opportunity for buying or originating fintech loans is estimated at US$280 billion over the next five years; banks are partnering with private-credit funds to originate and distribute loans off balance sheet expanding fee-based revenue while offloading credit risk
Fintech founders are facing a significant challenge as giant private-credit funds are lining up with term sheets so large they would have broken cap tables a year ago. A joint study from BCG and QED Investors puts the “white-space” opportunity at US$280 billion over the next five years: capital earmarked for buying or originating fintech loans. Private credit has grown nearly ten-fold since 2010 to roughly US$1.5 trillion of assets under management (AUM) in 2024, and consultants expect it to hit US$3.5 trillion by 2028, a compound annual growth rate north of 19%. Big banks are partnering with private-credit titans to originate and distribute loans off balance sheet, such as Citi x Apollo and Citi x Carlyle. The model is simple: banks keep the origination and servicing fees, funds take the credit risk, and regulators get comfort that risky assets live outside the deposit-backed system. Private-credit exuberance has pushed unitranche pricing down, but funds can lever those assets 1.5-2x and still net solid returns, making it an attractive asset class. However, many CEOs prefer a 14 percent cost of capital that preserves ownership over a 35% down round in an unforgiving venture market. The venture capital power-law reset threatens that maths, as private-credit recycling threatens that maths. Funds are aggressively scouring growth markets where banking pull-back is most acute, such as India, Southeast Asia, and Latin America, where dollar funding married to local-currency wallets is a tantalizing carry trade, provided FX hedges hold. However, there are several risks to consider, including credit deterioration, funding squeeze, regulatory shock, and FX blow-ups in emerging markets. To mitigate these risks, investors should monitor the spread compression Pace, Reg-Tech Build-Out, Private-Fund Reporting, Basel Endgame Final Rule, and Structured-Credit Revival.
AWS new neurosymbolic AI feature to mix symbolic or structured thinking with the neural network nature of generative AI to validate truth or correctness in an AI system against a set of policy or ground truth data, lending greater confidence in deploying AI agents
AWS is banking on the fact that by bringing its Automated Reasoning Checks feature on Bedrock to general availability, it will give more enterprises and regulated industries the confidence to use and deploy more AI applications and agents. It is also hoping that introducing methods like automated reasoning, which utilizes math-based validation to determine ground truth, will ease enterprises into the world of neurosymbolic AI, a step the company believes will be the next major advancement — and its biggest differentiation. Byron Cook, vice president at AWS’s Automated Reasoning Group, told the preview rollout proved systems like this work in an enterprise setting, and it helps organizations understand the value of AI that can mix symbolic or structured thinking with the neural network nature of generative AI. Automated Reasoning Checks validates truth or correctness in an AI system by proving that a model did not hallucinate a solution or response. This means it could offer regulators and regulated enterprises worried that the non-deterministic nature of generative AI could return incorrect responses more confidence. Cook brought up the idea that Automated Reasoning Checks help prove many of the concepts of neurosymbolic AI. Automated reasoning works by applying mathematical proofs to models in response to a query. It employs a method called the satisfiability modulo theories, where symbols have predefined meanings, and it solves problems that involve both logic (if, then, and, or) and mathematics. Automated reasoning takes that method and applies it to responses by a model and checks it against a set of policy or ground truth data without the need to test the answer multiple times. Cook said that agentic use cases could benefit from automated reasoning checks, and granting more access to the feature through Bedrock can demonstrate its usefulness.
U.S. Bank’s Business Resources Central (BRC) is centered on ease of access and comprehensive educational offerings, “open not just to U.S. Bank clients but is actually open to everyone”
US Bank Chief Product Officer for Business Banking Shruti Patel provided an assessment of the sentiment among SMBs. Results from a U.S. Bank survey show just where the worries lie, she said. “Almost 60% to 80% of our respondents are feeling very overwhelmed with the macro environment stressors right now,” she said, adding that her unit’s focus is on enterprises with sales under the $25 million threshold. “The uncertainty around tariffs, the impact on cost of operations, whether they’re going to absorb the cost, pass it down, the rate environment, access to capital” are top of mind. Last month, the bank launched U.S. Bank Business Resources Central (BRC). The free, online hub builds on the bank’s financial literacy and community access initiatives, said Patel, who added that the business access advisor program was set up in 2021. The online offering, launched in collaboration with Next Street, a firm focused on training programs for small businesses, functions as a sort of Coursera for enterprises. The overarching goal is to provide foundational business skills that remain relevant regardless of economic shifts. The collaboration uses Next Street’s experience to co-create content that addresses the broader needs of small businesses beyond banking and payments, Patel said. The functionality of the BRC is centered on ease of access and comprehensive educational offerings. The hub is “open not just to U.S. Bank clients but is actually open to everyone,” Patel said. The content (contained in 10 courses) covers essential topics such as how to secure financing, protect against cybercrime and improve digital literacy. There’s also emphasis on artificial intelligence, given its growing adoption by small businesses. “We made AI a very big portion of the business resource hub because we hear a lot from our customers … that they are deploying AI more and more in everyday operations,” Patel said. “Almost 80% of our small business survey respondents deploy AI. They’re trying to minimize the cost of their operations.” The hub offers dedicated toolkits to help businesses learn about and use AI tools for efficiency and cost reduction. The content is also customizable, tailored to the specific industry and stage of development of each small business. Beyond general business acumen, the BRC also helps SMBs navigate specific opportunities, such as becoming vendors for larger institutions, with the help of the business access advisors that act as procurement specialists and are located across more than a dozen markets around the United States. The focus on active use and return visits will help fine-tune the platform’s value to SMBs, Patel said. “The success of the platform is all about keeping up with the challenging times and continuing to iterate on the curated content to make sure that engagement metrics stay high,” she said. Early results show that nearly 20% of users in the initial weeks are already defined as U.S. Bank customers, and many are reaching out for one-on-one coaching. U.S. Bank is promoting the BRC through various marketing channels, including publications and social media, and through direct outreach by its business access advisors in communities across its footprint markets. “If we can help small businesses, whether they became a vendor, whether they get access to capital, whether they were able to understand how they can [undertake] sustaining and growing their business, that would be a big success story for us,” Patel said.
UWM swings back to profitability in Q2; originated $39.7 billion in mortgages during the quarter, up 23% from Q1 2025 and 18% higher than Q2 2024; launched Loan Estimate Optimizer (LEO) and Mia, an AI-powered assistant for loan officers
United Wholesale Mortgage (UWM), swung back to profitability in the second quarter of 2025, driven by higher loan volumes and wider margins. UWM originated $39.7 billion in mortgages during the quarter, up 22.5% from Q1 2025 and 18.2% higher than Q2 2024. Of that total, $27.3 billion was tied to purchase loans, putting the lender on pace to reach $100 billion in purchase volume for the year. “We delivered our best quarter since 2021, and it wasn’t because of any market tailwinds,” chairman and CEO Mat Ishbia said. “While others have pulled back, UWM has doubled down.” UWM’s refinance volume doubled compared to Q2 2024 and rose 17% from the first quarter to $12.4 billion. UWM said it captured 11% of the industry’s total refi volume during the period. The day before its earnings release, it announced a 90 basis-point incentive for rate-and-term refis that’s available through mid-September. The wholesale lender’s gain-on-sale margin improved to 113 basis points, up from 94 bps in Q1 2025 and 106 bps in Q2 2024. Regarding the wholesale channel market share, it reached nearly 30% from April through June — its highest share since 2009 — although UWM’s goal is to exceed 50%. According to Ishbia, 97% of consumers who work with a mortgage broker give a five-star review, but only 10% remember who their broker was when it’s time to refinance. To help brokers stay top of mind, UWM launched technologies like the Loan Estimate Optimizer (LEO) and Mia, an artificial intelligence-powered assistant for loan officers. Overall, UWM posted net income of $314 million for the quarter, reversing a $247 million loss in Q1 2025 and considerably higher than its $76.2 million profit in Q2 2024. But documents filed with the Securities and Exchange Commission (SEC) show a non-GAAP net profit of $249 million for the quarter. In its servicing portfolio, the company’s mortgage servicing rights totaled $211 billion in unpaid principal balance as of June 30, slightly down from $214 billion at the end of March. “Most people think you need to have the client in your servicing book to refinance them. First, we don’t refinance any borrowers — our brokers do. But second, we only own about 2% of the industry servicing’s market,” Ishbia said. “So, for us, to do 11-plus percent [of the industry’s refinances] really disproves the age-old theory that you must have to own the service to do the refi.” The lender has decided to bring its servicing in-house, a process that’s expected to be completed in Q1 2026. Ishbia said this move will have a positive financial impact and increase borrower loyalty to brokers. “We recently partnered with a company called Bilt, which will help create an amazing front-end experience for consumers,” he added. UWM ended the quarter with $2.2 billion in available liquidity, including $490 million in cash and borrowing capacity. Looking ahead, the lender expects to originate between $33 billion and $40 billion in the third quarter, with a gain-on-sale margin projected between 100 and 125 bps. Rami Hasani, UWM’s chief financial officer since April, said the company believes it is positioned to “handle twice our current production volume with minimal impact to fixed costs.” He also noted that UWM continues to assess and evaluate the opportunistic refinancing of $800 million in unsecured notes maturing in November 2025. UWM shares were trading around $4.60 as of Thursday morning , up roughly 8% from the previous day’s close.
JPMorgan’s report says DeFi and tokenization growth still disappoint- Total Value Locked (TVL) in DeFi remains below 2021 highs, with most activity still driven by crypto-native and retail users
The growth of decentralized finance (DeFi) and asset tokenization continues to underwhelm, JPMorgan’s Nikolaos Panigirtzoglou said in a research report Wednesday, citing the stagnant recovery since the 2022 crypto winter. Total Value Locked (TVL) in DeFi remains below 2021 highs, with most activity still driven by crypto-native and retail users, the report noted. Institutional adoption has lagged despite the development of compliance-ready infrastructure, such as permissioned lending pools and KYC-enabled vaults. Major barriers remain, as institutions face regulatory fragmentation, legal uncertainty around on-chain assets, and concerns about smart contract security. As a result, most institutional crypto activity remains concentrated in bitcoin. Tokenization has also struggled to deliver; while the sector has seen some traction, with $25 billion in tokenized assets, $8 billion in tokenized bonds, and growing adoption in money market funds, most initiatives remain small, illiquid, or experimental. Prominent efforts like BlackRock’s BUIDL and Broadridge’s Distributed Ledger Repo (DLR) platform offer efficiency gains but lack scale. In private markets, tokenization is heavily concentrated among a few players and lacks meaningful secondary market activity. Many traditional investors remain skeptical, especially given blockchain’s transparency, a drawback for institutions that favor opaque trading venues like dark pools, and the continued rise of off-exchange equity trading illustrates this preference. Despite regulatory initiatives like the SEC’s “Project Crypto,” Panigirtzoglou doubts whether rule changes alone can overcome the deeper issue: traditional finance doesn’t yet see a clear need for blockchain. Fintech has already improved speed and efficiency within the current system, reducing the urgency to adopt tokenized alternatives, the report added.
PNC’s Covid time partnership with Pirates to support SMEs continues to pick up momentum, having now aided a total of 36 small businesses in the Pittsburgh area
Pirates and PNC Bank teamed up to establish the “Going to Bat for Small Business” program to help local entrepreneurs navigate the challenges of the COVID-19 pandemic. The initiative was well-received and has continued to pick up momentum, having now aided a total of 36 small businesses in the Pittsburgh area. Each year, six small businesses are selected for the program. Representatives from the class of 2025 were invited to PNC Park on Monday for a meet-and-greet with officials from the Pirates and PNC Bank. They also took part in a pregame ceremony and had the opportunity to watch the game between the Pirates and San Francisco Giants from the ballpark’s World Series Suites. Those chosen for the “Going to Bat for Small Business” program receive a promotional period where the Pirates provide in-game, social media, online and in-park assets. Those include live drop-ins by Pirates broadcasters, radio and television commercials produced by the Pirates, social media sponsorship, and signage inside PNC Park. Each business also receives mentorship from Pirates and PNC professionals as well as a financial grant. Lou Cestello, head of Regional Presidents and regional president of Pittsburgh and southwestern Pennsylvania for PNC said, “This program is making a meaningful impact for the small businesses we serve, particularly through the guidance provided by PNC and the Pirates in areas such as financing and marketing. Our goal is to help them transform their ideas into successful realities.”