New research from the Bank for International Settlements (BIS) reveals striking statistics about stablecoin market influence. While stablecoin issuers have been noted as major holders of short term Treasuries, surpassing the holdings of countries such as China, the BIS highlights that during 2024, they were the third largest purchasers of Treasuries bills*. That figure is based on the net increase in stablecoin reserves. Regarding the impact on Treasury rates, the BIS study notes that a naive analysis of a $3.5 billion change in stablecoin holdings of Treasury bills implies a 25 basis point (0.25%) change in short term Treasury yields. However, it says this significantly overstates the impact, because many factors will simultaneously influence both stablecoin demand and Treasury rates. To try to isolate the impact of stablecoins, it explored major crypto events unrelated to movements in interest rates. Based on its research, the BIS concludes that the impact of sales of $3.5 billion of Treasuries by stablecoin issuers causes an increase in Treasury bill yields of six to eight basis points (0.06% – 0.08%). This is a much bigger effect than purchases, because sales are frequently more urgent given they may involve a mini crisis. A similarly sized purchase of Treasuries would result in a decline in Treasury rates of three basis points (-0.03%). The paper also noted that stablecoins are still comparatively small and the research is based on current volume levels. As stablecoins grow, the relative impact will increase. This will also create financial stability risks because of their effect on Treasury rates if there’s a run on a stablecoin. Additionally, with larger volumes of stablecoins, it will reduce the ability of the Federal Reserve to influence interest rates.
Multimode encoding could improve quantum error correction, also leakage errors, which remove the qubit from the encoding space, can now be detected and corrected
Nord Quantique has successfully developed bosonic qubit technology with multimode encoding, outlining a path to a significant reduction in the number of qubits required for quantum error correction. This provides the system protection against many common types of errors, including bit flips, phase flips, and control errors. Another key advantage over single-mode encoding is that leakage errors, which remove the qubit from the encoding space, can now be detected and corrected. The Tesseract code allows for increased error detection, and it is expected that this will translate into additional quantum error correction benefits as more modes are added. These results are therefore a key stepping stone in the development of this hardware-efficient approach. The core concept of the multimode approach centres on simultaneously using multiple quantum modes to encode individual qubits. Each mode represents a different resonance frequency inside an aluminium cavity and offers additional redundancy, which protects quantum information. The number of photons populating each mode can also be increased for even more protection, further escalating QEC capabilities. This breakthrough enables additional quantum error correction capacity and extra means for detecting errors, while maintaining a fixed number of qubits. It also delivers more benefits, which compound as they scale, opening new avenues for fault-tolerant quantum computing. Through this scientific advance, Nord Quantique now has a clear path to delivering fault tolerance at utility scale. The team will continue to improve its results by leveraging systems with additional modes to push the boundaries of quantum error correction.
InvestorFlow AI private markets capital solution integrates seamlessly into firm-wide capital formation and deployment workflows, automatically extracting proprietary often unstructured data found in in firm-wide interactions and meeting notes
InvestorFlow, the AI-powered front-office solution for private markets, announced new AI capabilities that transform unstructured interactions into actionable insights for capital formation and deployment. InvestorFlow AI delivers game-changing outcomes: from the early access program with top private market firms, clients recorded a 15X increase in actionable insights, including critical KPIs and in-quarter deal opportunities, and a 10X improvement in data accuracy — turning data into advantage at AI-scale. InvestorFlow AI integrates seamlessly into firm-wide capital formation and deployment workflows, automatically extracting proprietary often unstructured data found in firm-wide interactions and meeting notes, enriching that information with industry data providers including Preqin and Pitchbook, and generating intelligent, dynamic briefs for every fund, investor or deal opportunity. These insights are embedded directly into existing pipelines and workflows, accelerating decision-making and productivity where users already work. InvestorFlow AI cuts through the complexity, rapidly surfacing the most strategic and actionable insights from vast amounts of unstructured data. Whether in private equity, venture capital, infrastructure, real estate or credit, capital allocators increase optionality and originate the best opportunities by synthesizing insights from their proprietary data. InvestorFlow AI helps by identifying investor preferences and investment targets, generating precision-targeted investor lists for investor relations leaders, streamlining meeting preparation and follow-up, and accelerating co-investment campaigns, all at an accelerated pace.
Tether USDT and, Tron dominate fast-growing stablecoin payments arena, survey shows; Tron was the preferred settlement network, hosting around 60 percent of volume
Tether’s USDT token and the Tron blockchain network dominate the rapidly growing stablecoin payment industry, according analytics firm Artemis with help from investment firms Dragonfly and Castle Island Ventures. Their report looked at data from 31 stablecoin payment companies, and found USDT, the largest stablecoin, accounted for 90 percent of payment transaction volume, followed by Circle’s USDC, the second-largest. Tron was the preferred settlement network, hosting around 60 percent of volume, followed by Ethereum, Binance Smart Chain and Polygon. It’s perhaps surprising that the share of Circle’s USDC isn’t larger, given the firm’s involvement in payments and recent plans to introduce a dedicated cross-border payments network. In addition, Circle, which this week filed for an initial public offering on the New York Stock Exchange, has been taking market share from Tether in terms of issuance, so the expectation might have been a similar or pro-rata level when it comes to payments volume, said Dragonfly general partner Rob Hadick. “For the 31 providers we got data from at least, it’s clear that’s not the case for the payments use case,” Hadick said. “In fact, a higher portion of the volume, relative to the issuance, is happening with Tether, and it’s happening primarily on Tron and then Ethereum. This was quite surprising to us.”
Walmart’s agentic AI strategy to follow ‘surgical’ approach where agents will have expertise at retail-specific tasks and their work outputs will be stitched together to solve complex workflows
Walmart U.S. Chief Technology Officer Hari Vasudev unveiled the retailer’s agentic AI strategy and implementation plans, preparing for an era where robot shoppers will buy products and services from robot sellers, accessing websites optimized for them with the goal of delivering fast, hyper-personalized experiences to human shoppers. First, Walmart identifies core agentic AI capabilities that would work best for the retailer, are cohesive and can scale globally, per the post. Next, it uses a “surgical” approach to agentic AI. That means its agents will be experts at specific tasks, unlike the more generic solutions from other providers. Finally, Walmart agents’ work outputs will be stitched together to solve complex workflows. As an example, Walmart taps its retail-expert large language model to build agents within its generative AI shopping assistant, which appears as a smiley face chatbot. These agents can do specific tasks such as deep personalization, item comparison and shopping journey completion, among others. The model is trained on the retailer’s data and can be combined with other models to contextually address the customer’s needs Walmart’s existing generative AI-powered tools are on their way to becoming fully autonomous agents. Walmart is also exploring using AI agents across the company, from doing in-store tasks to merchandising planning at the home office and beyond. Shoppers are already using Walmart’s shopping assistant to find products, and the next step is to let an agent do the research, make decisions and place the order. This autonomous task would be ideal for repeat purchases of everyday necessities. Walmart is aware of the risk of hallucinations, or AI models making things up. So, it is adding a layer of governance, checks and balances, as well as evaluating which parts of agentic AI need human oversight and approval
Experiential anchors- booksellers, fitness centers and food & beverage brands are serving to drive customer visits to malls, accounting for 8-16% of visitor share
Macy’s and JCPenney still play a key role in drawing customers to malls, but empty anchor implants like entertainment brands, fitness centers, and restaurants are increasingly building new traffic across longer ranges of hours, declares Placer.ai’s latest Mall Report. One key participant in this trending phenomenon is a brand that most retail experts thought had run its course: Barnes & Noble. In recent years, the land’s leading bookseller has reinvented itself in smaller stores (15,000 sq. ft. versus 25,000 sq. ft.) redesigned to be “hangouts” for local customers with better lighting, more open layouts, and opportunities for social interaction. At the Coronado Center in Albuquerque, N.M., the Barnes & Noble accounted for 7.9% in 2024, according to Placer.ai, outperforming both Macy’s and JCPenney. Key food-and-beverage brands, too, are now outpacing department stores with their traffic counts. At Northridge Fashion Center in Northridge, Calif., Porto’s Bakery & Café was the No. 1 customer draw with a 15.6% share of overall center visitor, a full 3.6 percentage point lead over No. 2 Dick’s Sporting Goods. And while Placer.ai had Target as the top tenant at Glendale Galleria in Glendale, Calif., Placer.ai with a 14.4% share of visitors, it was followed by by In-N-Out Burger’s 8.6% share in second place. “Increasingly, shopping centers are turning to fitness centers as experiential anchors,” according to the Placer.ai report. “And since many people work out early in the morning, these gyms are having a significant impact on the distribution of mall visits across dayparts.” At Northshore Mall in Peabody, Mass., where a Life Time gym opened in 2021, visits between 7:00 a.m. and 12:00 p.m. rose from 13% to 15% over the past five years. Similarly, at Jackson Crossing in Jackson, Mich., where Planet Fitness arrived in 2022, the morning visit share increased from 14% to 16%.
Microsoft proposes GenAI Intent-based routing (IBR)- Customer Intent Agent discovers and manages intents, while IBR uses those intents to route conversations, connecting customer needs to the right support resources with speed and precision
Intent-based routing (IBR) is a generative AI-powered capability that routes customer queries based on real-time intent recognition and dynamic user group assignment. It is enabled by the Customer Intent Agent, which autonomously discovers and manages customer intents by analyzing past interactions and builds an evolving intent library. Customer Intent Agent discovers and manages intents, while IBR uses those intents to route conversations, connecting customer needs to the right support resources with speed and precision. Once an intent and its group are identified, IBR routes the conversation to the appropriate user group based on the mapped intent group. Next, IBR assigns it to the best-suited support representative within the group, based on their capacity, presence, and other routing attributes. This enables faster, more accurate resolutions. By turning intent from a passive insight into an active, intelligent routing decision, IBR becomes the operational backbone of an intent-driven contact center. Subsequently, this results in better assisted and self-service experiences. Implementing intent-based routing in your contact center can offer numerous benefits: Enhanced precision and personalization; Dynamic intent discovery; Streamlined routing configuration; Smarter workforce management and load handling; and Scalable and adaptable.
Apple’s LLM FOR Siri with 150 billion parameters equals the quality of ChatGPT’s recent releases but shows higher levels of hallucination
A new report claims that internally, Apple has already been testing Large Language Models for Siri that are vastly more powerful than the shipping Apple Intelligence, but executives disagree about when to release it. Apple is said to be testing models with 3 billion, 7 billion, 33 billion, and 150 billion parameters. For comparison, Apple in 2024 said that Apple Intelligence’s foundation language models were of the order of 3 billion parameters. That version of Apple Intelligence is intentionally small in order for it to be possible to run on-device instead of requiring all prompts and requests to be sent to the cloud. The larger versions are cloud-based, and in the case of the 150 billion parameter model, now also said to approach the quality of ChatGPT’s most recent releases. However, there reportedly remain concerns over AI hallucinations. Apple is said to have held off releasing this Apple Intelligence model in part because of this, implying that the level of hallucinations is too high. There is said to be another reason for not yet shipping this cloud-based and much improved Siri Chatbot, though. It is claimed that there are philosophical differences between Apple’s senior executives over the release.
Google’s new app lets users find, download, and run openly available AI models that generate images, answer questions, write and edit code, and more on their phones without needing an internet connection
Google quietly released an app that lets users run a range of openly available AI models from the AI dev platform Hugging Face on their phones. Called Google AI Edge Gallery, the app is available for Android and will soon come to iOS. It allows users to find, download, and run compatible models that generate images, answer questions, write and edit code, and more. The models run offline, without needing an internet connection, tapping into supported phones’ processors. Google AI Edge Gallery, which Google is calling an “experimental Alpha release,” can be downloaded from GitHub. The home screen shows shortcuts to AI tasks and capabilities like “Ask Image” and “AI Chat.” Tapping on a capability pulls up a list of models suited for the task, such as Google’s Gemma 3n. Google AI Edge Gallery also provides a “Prompt Lab” users can use to kick off “single-turn” tasks powered by models, like summarizing and rewriting text. The Prompt Lab comes with several task templates and configurable settings to fine-tune the models’ behaviors. Your mileage may vary in terms of performance, Google warns. Modern devices with more powerful hardware will predictably run models faster, but the model size also matters. Larger models will take more time to complete a task — say, answering a question about an image — than smaller models. Google’s inviting members of the developer community to give feedback on the Google AI Edge Gallery experience. The app is under an Apache 2.0 license, meaning it can be used in most contexts — commercial or otherwise — without restriction.
Fintech Chime is said to plan IPO launch as soon as Monday weighing a valuation of about $11 billion
Chime Financial Inc. is planning to launch its initial public offering as soon as Monday, as US listings bounce back from a disappointing April. The no-fee banking services company is weighing a valuation of about $11 billion. Chime was valued at $25 billion after raising $750 million in a funding round in 2021, according to an announcement at the time. The firm which filed for a listing with the US Securities and Exchange Commission in May, is set to disclose how much it will seek to raise in the IPO. No final decisions have been made and details of the offering including valuation and timing could change. Chime disclosed net income of $12.9 million on revenue of $518.7 million for the first three months of 2025, according to its latest filing. That compares with net income of $15.9 million on revenue of $392 million a year earlier. Chime’s largest investors include affiliates of DST Global, Crosslink Capital, Len Blavatnik’s Access Industries, General Atlantic, Menlo Ventures, Sino French (Innovation) Fund and Iconiq Strategic Partners, the filing shows. The offering is being led by Morgan Stanley, Goldman Sachs Group Inc. and JPMorgan Chase & Co. The company plans for its shares to trade on the Nasdaq Global Select Market under the symbol CHYM.