The credit reporting system is no longer fit for purpose. Built for a different era, it fails to account for the dynamics of modern lending, particularly short-term, high-frequency credit products like BNPL. Instead of helping lenders assess risk and enabling consumers to access fair credit, the current system misrepresents creditworthiness. The solution is not to exclude BNPL data from credit reporting. Nor is it to shoehorn new behaviours into old models. We need structural change. A credit reporting system designed for the 2020s should include: 1) Modern scoring logic: Models should focus on actual repayment behaviour and differentiate between product types. High frequency, short term credit such as BNPL should be assessed in its own context. 2) Real-time data infrastructure: Data should be delivered instantly, through APIs and event driven systems, enabling lenders to see the full picture in the moment it matters. Real time capabilities also allow for responsible product design, like adaptive credit limits and instant affordability checks. 3) Fit for purpose standards: Data formats must accommodate transaction-level detail, flexible repayment schedules, and user context. Standards should evolve with innovation, not lag behind it. A shared taxonomy for short-term credit products is also essential to ensure consistent interpretation.
Banks are pushing cryptos into mainstream following regulatory maturation, the real-world quest for stablecoin utility and the institutionalization of digital assets
With new SEC leadership and defined compliance expectations, banks and companies are moving from experimentation to operational integration — using crypto and stablecoins for cross-border payments, corporate treasuries and programmable money at scale. Major global banks, like ING, are in fact beginning to partner on stablecoin projects, motivated both by the fear of being left behind and the opportunity to define new standards. These are robust, multi-bank consortia aiming for real-world use cases — cross-border payments, corporate treasuries and eventually programmable money at scale. The Lynq network — developed by Arca Labs, Tassat Group, and tZERO — promises real-time, yield-bearing settlements, a marked upgrade from legacy systems that still settle on T+2 timelines. This settlement innovation, which also includes participation from U.S. Bank, is critical for both risk management and unlocking new forms of financial products. Putting an exclamation point on today’s landscape, a Cantor Fitzgerald affiliate business is teaming up with SoftBank and Tether to create a multi-billion-dollar corporate treasury vehicle with the goal of accumulating bitcoin. Upexi, a consumer products firm, is raising $100 million to accumulate Solana, echoing the “corporate treasury as crypto hedge” playbook pioneered by MicroStrategy. This signals not just speculative belief, but operational integration: companies see blockchains not only as investment vehicles but as potential infrastructure for their own business models. Circle’s launch of a stablecoin orchestration layer aims to make stablecoins “invisible” in the best sense: moving money across borders, across blockchains and into the hands of consumers without them needing to understand the underlying tech. Major financial institutions are taking notice, not just with experimental projects but with real investment and product launches. The partnership between CompoSecure and MoneyGram exemplifies this. By enabling cash-to-crypto conversions at thousands of global MoneyGram locations, stablecoins are made accessible to the unbanked and underbanked, potentially reshaping remittance and financial inclusion. With Visa, Mastercard and JPMorgan testing tokenized forms of cash, treasuries and even real estate, we’re beginning to see the outlines of a future where everything of value can be transacted in programmable, composable digital units. Taken together, these three trends — regulatory maturation, the real-world quest for stablecoin utility and the institutionalization of digital assets — mark a turning point. The Wild West days of crypto are fading, replaced by a convergence with mainstream finance. Success could mean a financial system that is faster, fairer and more inclusive, leveraging the strengths of both centralized and decentralized models.
Private credit markets with a higher degree of customization are more resilient to restrictive monetary policy compared to public credit markets and bank lending
Fed Gov. Adriana Kugler said the growth of private lending in the wake of the global financial crisis created a market that was relatively immune to the central bank’s restrictive policy stance. “One implication of this strong growth during this past policy tightening is that monetary policy transmission to private credit markets appeared more muted relative to financing through public credit markets or bank commercial and industrial lending,” Kugler said. Kugler attributed the proliferation of private credit to structural advantages such lenders have over banks, including their ability to offer “higher customization” to borrowers and investors alike. The resilience of nonbank lenders was not the Fed’s only takeaway from its post-pandemic tightening cycle. Kugler said the central bank also learned about the impact of excess savings on monetary policy transmission. She noted that the combination of government stimulus and curtailed spending as a result of COVID-19 social distancing “led the personal savings rate to soar.” The saving glut, she said, effectively created a buffer between consumers and higher borrowing costs. “If households are flush with excess cash, they are less likely to respond to elevated interest rates by curtailing demand,” Kugler said. “Instead, they may have funds to avoid financing or may feel they are able to afford higher monthly payments.” Those excess savings have largely evaporated, Kugler said, allowing monetary policy to impact the economy in a more typical fashion. But, she added, the effects have been more pronounced for less creditworthy borrowers, pointing to credit card and auto loan delinquencies, which have risen above pre-pandemic levels. Kugler said the disparate impacts between prime and subprime borrowers could play out in the inverse, once the Fed resumes lowering interest rates. “For these [lower credit] households, easing monetary policy may have larger effects,” she said. Kugler said financial conditions — namely the willingness of banks to provide credit — have front-run some of the Fed’s monetary decisions. She noted that conditions began easing last year even before the Fed began cutting interest rates in September, which corresponded with an increased demand for loans by households and businesses. But, overall, she said banks have only reduced the interest rates they charge modestly from their post-pandemic peaks and stopped doing so early this year in accordance with the Fed’s pause on policy adjustments. “Banks stopped tightening lending standards after nine consecutive quarters, but they left standards unchanged in January,” she said.
Amazon’s new benchmark to evaluate AI coding agents’ ability to navigate and understand complex codebases and GitHub issues
Amazon has introduced SWE-PolyBench, the first industry benchmark to evaluate AI coding agents’ ability to navigate and understand complex codebases. The benchmark, which measures system performance in GitHub issues, has spurred the development of capable coding agents and has become the de-facto standard for coding agent benchmarking. SWE-PolyBench contains over 2,000 curated issues in four languages and a stratified subset of 500 issues for rapid experimentation. The benchmark aims to advance AI performance in real-world scenarios. Key features of SWE-PolyBench at a glance: Multi-Language Support: Java (165 tasks), JavaScript (1017 tasks), TypeScript (729 tasks), and Python (199 tasks). Extensive Dataset: 2110 instances from 21 repositories ranging from web frameworks to code editors and ML tools, on the same scale as SWE-Bench full with more repository. Task Variety: Includes bug fixes, feature requests, and code refactoring. Faster Experimentation: SWE-PolyBench500 is a stratified subset for efficient experimentation. Leaderboard: A leaderboard with a rich set of metrics for transparent benchmarking.
New system for training AI agents focuses on multi-turn, interactive settings where agents must adapt, remember, and reason in the face of uncertainty instead of static tasks like math solving or code generation
A collaborative team from Northwestern University, Microsoft, Stanford, and the University of Washington — including a former DeepSeek researcher named Zihan Wang, currently completing a computer science PhD at Northwestern — has introduced RAGEN, a new system for training and evaluating AI agents that they hope makes them more reliable and less brittle for real-world, enterprise-grade usage. Unlike static tasks like math solving or code generation, RAGEN focuses on multi-turn, interactive settings where agents must adapt, remember, and reason in the face of uncertainty. Built on a custom RL framework called StarPO (State-Thinking-Actions-Reward Policy Optimization), the system explores how LLMs can learn through experience rather than memorization. StarPO-S incorporates three key interventions: Uncertainty-based rollout filtering; KL penalty removal; and Asymmetric PPO clipping. StarPO operates in two interleaved phases: a rollout stage where the LLM generates complete interaction sequences guided by reasoning, and an update stage where the model is optimized using normalized cumulative rewards. This structure supports a more stable and interpretable learning loop compared to standard policy optimization approaches. The team identified three dimensions that significantly impact training: Task diversity, Interaction granularity, and Rollout freshness. Together, these factors make the training process more stable and effective.
OpenAI is planning a truly ‘open reasoning’ AI system with a ‘handoff’ feature that would enable it to make calls to the OpenAI API to access other, larger models for a substantial computational lift
OpenAI is gearing up to release an AI system that’s truly “open,” meaning it’ll be available for download at no cost and not gated behind an API. Beyond its benchmark performance, OpenAI may have a key feature up its sleeve — one that could make its open “reasoning” model highly competitive. Company leaders have been discussing plans to enable the open model to connect to OpenAI’s cloud-hosted models to better answer complex queries. OpenAI CEO Sam Altman described the capability as a “handoff.” If the feature — as sources describe it — makes it into the open model, it will be able to make calls to the OpenAI API to access the company’s other, larger models for a substantial computational lift. It’s unclear if the open model will have the ability to access some of the many tools OpenAI’s models can use, like web search and image generation. The idea for the handoff feature was suggested by a developer during one of OpenAI’s recent developer forums, according to a source. The suggestion appears to have gained traction within the company. OpenAI has been hosting a series of community feedback events with developers to help shape its upcoming open model release. A local model that can tap into more powerful cloud systems brings to mind Apple Intelligence, Apple’s suite of AI capabilities that uses a combination of on-device models and models running in “private” data centers. OpenAI stands to benefit in obvious ways. Beyond generating incremental revenue, a handoff could rope more members of the open source community into the company’s premium ecosystem.
New quantum states that are magnet-freee could support building topological quantum computers that are stable and less prone to the errors
A new study published in Nature reports the discovery of over a dozen previously unseen quantum states in twisted molybdenum ditelluride, expanding the “quantum zoo” of exotic matter. Among them are states that could be used to create what is known, theoretically at the moment, as a topological quantum computer. Topological quantum computers will have unique quantum properties that should make them less prone to the errors that hinder quantum computers, which are currently built with superconducting materials. But superconducting materials are disrupted by magnets, which have until now been used in attempts to create the topological states needed for this (still unrealized) next generation of quantum computers. Lead author from Howard Family Professor of Nanoscience at Columbia, Xiaoyang Zhu’s zoo solves that problem: The states he and his team discovered can all be created without an external magnet, thanks to the special properties of a material called twisted molybdenum ditelluride. These states, including magnet-free fractional quantum Hall effects, could support non-Abelian anyons—key building blocks for more stable, topological quantum computers. The discoveries were made using a pump-probe spectroscopy technique that detects subtle shifts in quantum states with high sensitivity, revealing fractional charges and dynamic quantum behavior.
Bilt Rewards platform enables students to earn rewards on their student housing payments and redeem their rewards toward student loan repayments
Bilt Rewards now enables students to earn rewards on their student housing payments and redeem their rewards toward student loan payments. The first of these new capabilities results from the expansion of the Bilt Rewards network of homes to include student housing properties, beginning with those of its launch partner American Campus Communities (ACC). This partnership with ACC, which is a student housing company and a Blackstone portfolio company, will begin in late May at two properties at Baylor University and then expand in the coming months to the broader ACC portfolio that serves nearly 140,000 students. The collaboration extends to student housing properties the Bilt Rewards payments and commerce network that transforms housing and neighborhood spending into rewards and benefits. The other new feature announced — the ability for Bilt members to redeem their Bilt Points toward eligible student loan payments — is the first of its kind for Bilt Rewards. Starting Wednesday, Bilt members can redeem Bilt Points on student loans with five servicers: Nelnet, MOHELA, Sallie Mae, Aidvantage and Navient.
Synchrony partners Belle Tire to expand its car care credit card use and flexible financing to more than 1 million gas stations, auto parts retailers, and service locations nationwide
Belle Tire has partnered with financial services company Synchrony to make car care more affordable. The Belle Tire private label credit card, part of the Synchrony Car Care network, will provide customers with access to flexible financing options for a wide range of automotive products and services. Cardholders can take advantage of six months of promotional financing on purchases of $199 or more and twelve months of promotional financing on purchases of $1,000 or more. Additionally, the Synchrony Car Care credit card can be used not only at Belle Tire locations but also at more than 1 million gas stations, auto parts retailers, and service locations nationwide. Synchrony collaborated with 1stMILE, Belle Tire’s integrated payment provider, to help to ensure a seamless rollout of financing options across all Belle Tire locations, enabling the company to reach more customers from a single platform. Belle Tire customers can apply for Synchrony financing through multiple channels, including in-store pin pads, QR codes, text-to-apply options, and on BelleTire.com, using Synchrony’s pre-qualification and digital application solutions. With the ability to pay for everything from routine maintenance to unexpected repairs, as well as for parking and car rentals, the Synchrony Car Care Network aims to simplify automotive financial management while providing “flexibility and convenience” for cardholders.
Startup Alpaca’s platform lets financial services firms and brokerages offer trading services especially U.S. stock market to their consumer user bases via an API
The U.S. stock market remains extremely attractive to investors around the world, simply because of its sheer size and liquidity. But it’s still quite difficult for investors in other parts of the globe to trade stocks on U.S. exchanges. Startup Alpaca has quietly capitalized on that opportunity by offering an API to financial services firms that lets them sell trading services to their consumer user bases. Alpaca claims it serves more than 5 million brokerage accounts, and has more than 200 financial clients in 40 countries. To build on that traction, Alpaca has raised $52 million in a Series C funding round to expand into more foreign markets, including the Middle East, Europe, and Asia. The startup just opened a new office in New York, and it plans to use the fresh cash to obtain more regulatory licenses in different regions, similar to those it already has in the U.S., Japan, and the Bahamas, its co-founder and CEO Yoshi Yokokawa told. Alpaca will also use the proceeds to develop new products, add non-U.S. products like European and Asian equities, and support 24/5 trading of U.S. stocks. “When new banks want to improve their products for their customers, they prefer to work with modern partners because their customers want modern solutions. So that is how we are currently winning the market share over them,” Yokokawa added.
