Hirundo AI Ltd., a startup that’s helping AI models “forget” bad data that causes them to hallucinate and generate bad responses, has raised $8 million in seed funding to popularize the idea of “machine unlearning.” Hirundo’s approach to AI hallucinations is about making fully trained AI models forget the bad things they learn, so they can’t use this mistaken knowledge to generate their responses later on, down the line. It does this by studying the behavior of AI models in order to locate the directions users can go in order to manipulate them. It identifies any bad traits, then investigates the root cause of those bad outputs, before steering the model away from them. It pinpoints where hallucinations originate from in the billions of parameters that make up their knowledge base. This retroactive approach to fixing undesirable behaviors and inaccuracies in AI models means it’s possible to improve their accuracy and reliability without needing to retrain them. That’s a big deal, because retraining models can take many weeks and cost thousands or even millions of dollars. “With Hirundo, models can be remediated instantly at their core, working toward fairer and more accurate outputs,” Chief Executive Ben Luria added. Besides helping models to forget bad, biased or skewed data, the startup says it can also make them “unlearn” confidential information, preventing AI models from revealing secrets that shouldn’t be shared. What’s more, it can do this for both open-source models such as Llama and Mistral, and soon it will also be able to do the same for gated models such as OpenAI’s GPT and Anthropic PBC’s Claude. The startup says it has successfully managed to remove up to 70% of biases from DeepSeek Ltd.’s open-source R1 model. It has also tested its software on Meta Platforms Inc.’s Llama, reducing hallucinations by 55% and successful prompt injection attacks by 85%.
PayPal’s reward offers on stablecoin balances could force super apps to offer interest-earning prepaid cards and wallets in a closed-loop, scalable ecosystem
PayPal and Circle, two U.S. firms that are trying to build what are called financial super apps, are making stablecoins a major piece of how users link to other functions. Both companies made recent deals to expand usage of their own stablecoins while easing access to other financial services. PayPal and Circle are both positioning themselves to capitalize on the increasing popularity of stablecoins as a means to power their super apps. PayPal recently expanded its partnership with cryptocurrency exchange Coinbase to power free conversions between PayPal’s PYUSD stablecoin and traditional money. Coinbase will offer PYUSD to PayPal’s merchant network, making it easier for the stablecoin to be used at the point of sale. It also can drive PayPal’s super app strategy, a long-standing initiative to enable a mobile app to shop at millions of merchants, make payments, access a debit card, use installment credit, pay bills, invest, make P2P transfers, receive paychecks and dozens of other services. The addition of the Xoom remittance and Venmo transfer app, PayPal’s PYUSD and support for stablecoin and cryptocurrency investments at Venmo and PayPal add to that super-app potential.
OCC clarifies banks may buy and sell assets crypto-assets; can also outsource custody to third parties
The OCC has clarified that institutions under its oversight can now buy and sell crypto assets on behalf of their customers. In addition, the OCC stated that national banks may outsource crypto-asset services to third parties, including custody and trade execution, provided those third parties maintain sound risk management practices. The latest OCC letter follows a similar directive issued in March, which rescinded the 2021 policy requiring banks to seek prior supervisory approval before engaging in crypto-related services. “The services national banks may provide in relation to the cryptocurrency they are custodying may include services such as facilitating the customer’s cryptocurrency and fiat currency exchange transactions, transaction settlement, trade execution, recordkeeping, valuation, tax services, reporting, or other appropriate services,” the March letter stated. It further clarified: “A bank acting as custodian may engage a sub-custodian for cryptocurrency it holds on behalf of customers and should develop processes to ensure that the sub-custodian’s operations have proper internal controls to protect the customer’s cryptocurrency.”
Meanwhile, the Federal Reserve recently dropped its supervisory guidelines that previously required American banks to notify it in advance of any crypto-asset activities. Banks are also no longer required to obtain formal approval from the Fed before engaging in stablecoin-related operations. The decisions by both US regulators reflect the broader shift toward more crypto-friendly policies under the Trump administration.
Morgan Stanley forecasts the market for humanoids will grow to become materially larger than the global auto industry and reach approximately $4.7 trillion by the year 2050
Humanoid will arrive sooner than expected, says Morgan StanleyHumanoid will arrive sooner than expected, says Morgan Stanley. The investment banking firm projects that “the team set forth their proprietary humanoid TAM model that projects the global market for humanoid robots will grow to become materially larger than the global auto industry.” This accelerated timeline suggests a potentially disruptive force with broad economic and sector-wide implications. With the arrival of humanoids, the MS research estimates this potential impact by estimating that the market for humanoids could reach approximately $4.7 trillion by the year 2050. This provides the long-term investment opportunities and the transformative nature of humanoid technology for the investors. Within this AI-driven landscape, the development of humanoids, considered a subset of “Embodied AI,” is identified as a key area of focus. However, 2025 will be the year of Agentic AI, where companies can use Agentic AI tools to improve their businesses. Morgan Stanley says, “we believe the magnitude of benefits from AI adoption is vastly underestimated.” “AI spend is set to rise dramatically, but we see a $1.1 trillion revenue opportunity as early as 2028, with contribution margins of 34% in 2025, rising to 67% by 2028,” Morgan Stanley said. On the nuclear power front, Morgan Stanley believes that “nuclear renaissance will be worth $1.5 trillion through 2050 in the form of capital investment in new global nuclear capacity, which is based on the assumption that there will be 383.5GW of new nuclear capacity to be added globally, which is roughly equal to the current global nuclear capacity of 390GW.” Both AI and humanoids are seen as pivotal forces shaping investment strategies and offering significant opportunities for alpha generation within the evolving technological landscape.
Mastercard and MoonPay team to promote stablecoin payments in an API-driven implementation letting businesses, neobanks, and other payment participants manage payouts and disbursements more efficiently
Mastercard has launched a stablecoin-focused partnership with cryptocurrency payments FinTech MoonPay. The collaboration will allow consumers and businesses to send and receive stablecoin payments across global markets. Companies and FinTechs will be able to employ Mastercard-branded cards linked to users’ stablecoin balances, allowing cardholders to spend their stablecoins, which will simultaneously be converted to fiat currency, at more than 150 million locations where Mastercard is accepted around the world. “By providing solutions that unlock stablecoin utility and ubiquity, we are redefining how money moves globally and driving a shift in payments as we know it,” Scott Abrahams, executive vice president, Global Partnerships at Mastercard, said. T he partnership will leverage the API-driven stablecoin infrastructure from Iron, acquired by MoonPay in March, to facilitate stablecoin transactions, turning “crypto wallets into new digital bank accounts for seamless global transactions.” This will let businesses, neobanks, and other payment participants manage payouts and disbursements more efficiently, improving cross-border money transfers, and help businesses offer stablecoin-based payouts to gig workers, contractors and creators.
Gravitee Topco’s open-source API management platform offers an array of tools for developers that span API design, access, management, deployment and security with support for both asynchronous and synchronous APIs
Digital traffic pipeline management startup Gravitee Topco has closed on a $60 million Series C funding round, bringing its total amount raised to date to more than $125 million. The company is the creator of an open-source API management platform that provides developers with the tools they need to easily manage both legacy and newer data streaming protocols. It also provides a wealth of API security tools with its platform. Gravitee’s core offering is split into two products, with the Gravitee API Management tool designed for API publishers, and the Gravitee Access Management offering aimed at the developers who need to use those APIs. Through the two platforms, it provides tools that span API design, access, management, deployment and security. Gravitee can therefore be thought of as a kind of control plane for APIs, which often come with a confusing array of protocols and tools that can quickly overwhelm developers, despite their intention of making life simpler. Companies can deploy Gravitee’s core, open-source offering in the cloud or on-premises, or they can access the premium platform through the startup’s software-as-a-service offering. Its core features include a tool for designing and deploying APIs, mock testing and a dashboard that provides an overview of team’s API deployments. What makes Gravitee different is that it supports both asynchronous and synchronous APIs, meaning APIs that deliver data at a later point in time, and those that deliver data immediately, in real time.
Google’s updates to media apps on Android Auto to allow apps to show different sections in the browsing UI and offer more flexibility in layout to build richer and more complete experiences
Google introduced two new changes to media apps on Android Auto. The first change is to the browsing interface in media apps. The new “SectionedItemTemplate” will allow apps to show different sections in the browsing UI, with Google’s example showing “Recent search” above a list of albums. The other change is the to “MediaPlaybackTemplate,” which is used as the “Now Playing” screen. It appears that Google is going to grant developers more flexibility in layout here, with the demo shown putting the media controls in the bottom right corner instead of the center, and in a different order than usual – although that might become the standard at some point. The UI isn’t drastically different or any harder to understand, but it’s a different layout than we usually see on Android Auto, which is actually a bit refreshing. Google is also allowing developers to build “richer and more complete experiences” for media apps using the “Car App Library.” This could make it easier to navigate some apps, as most media apps on Android Auto are shells of their smartphone counterpart in terms of functionality. This category is just in beta for now, though.
FICO partners with AWS to accelerate the adoption of AI-driven, automated decision workflows in its platform and simplify procurement for customers through its availability on AWS Marketplace
FICO has signed a new strategic collaboration agreement with Amazon Web Services (AWS) to bring more organizations worldwide the power of AI-driven, automated decision workflows with FICO Platform, which runs on AWS, and FICO will broaden its participation in AWS partner programs to accelerate client adoption of FICO Platform. FICO solutions will also be available through AWS Marketplace Private Offers. The first of these solutions, FICO Decision Modeler, is now available in AWS Marketplace. FICO Decision Modeler is the advanced decision management system that powers greater flexibility, ease of use and decision accuracy, and is at the core of FICO Platform. Many financial institutions are already using FICO Platform on AWS to drive successful business initiatives, such as Westpac NZ, one of New Zealand’s largest retail banks. “With FICO Platform, which supports transformation at scale on AWS, we’ve been able to shift from focusing on individual decisions we make on a customer’s account, to thinking strategically about how we manage customers across all of their accounts and throughout their lifetime with the bank,” said Regan Goble, Risk Analytics senior manager at Westpac NZ. “FICO’s decision management solutions provide customers with powerful tools for making data-driven financial decisions,” said Scott Mullins, managing director, Worldwide Financial Services at AWS. “Making these solutions available in AWS Marketplace simplifies procurement for our customers while providing the security, scalability, and performance benefits of AWS.
Startup New Gen is building AI storefronts using “AI subdomain” to receive AI-driven traffic; it reindexes a brand’s entire product catalog and uses AI models to create dynamic product pages and recommendation
Startup New Generation, or simply New Gen, is building AI storefronts – versions of brands’ websites that dynamically interact with AI chatbots and agents. Its technology helps brands automate product tagging, personalize responses and adapt to real-time market trends. To build the AI storefront, New Gen creates an “AI subdomain” — a dedicated microsite such as “ai.brand.com” designed to receive AI-driven traffic from tools like ChatGPT or website-browsing agents like Operator. Their system reindexes a brand’s entire product catalog, combines structured and unstructured data such as product descriptions and social media posts, and uses AI models to create dynamic product pages and recommendations. This data pipeline is available through an API for other companies to access. For image generation, New Gen uses Gemini 2.5. For prose and copy, the team relies on Anthropic’s Claude 4. For code and front-end generation, OpenAI’s models are preferred. Brands can control the experience through a merchant dashboard, setting preferences for tone of voice, which products to highlight, and seasonal merchandising strategies. Over time, the AI storefronts would integrate with marketplaces such as Shopify or WooCommerce. New Gen has partnered with Visa as part of the payments giant’s Intelligent Commerce initiative to enable AI agents to make purchases on behalf of consumers.
Postman looks to streamline API and agentic AI development with Agent Mode that automates API design, testing, documentation, and monitoring through natural language inputs
Postman Inc. is rolling out a suite of AI-driven features aimed at transforming API development. The company’s latest introduction, Agent Mode, automates API design, testing, documentation, and monitoring through natural language inputs. This feature acts as a fully capable execution agent, streamlining development workflows and reducing manual effort. Beyond Agent Mode, Postman is enhancing real-time API observability, enterprise-ready integrations, and support for the Model Context Protocol (MCP), which standardizes the way AI agents interact with third-party tools. Developers will soon have the ability to create their own AI agents and deploy them in their workspaces, improving efficiency in both daily engineering tasks and broader operations. One standout addition is Postman Insights, which provides real-time tracking for API usage, failure patterns, and proactive debugging. The Repo Mode feature further simplifies testing, allowing developers to reproduce API failures for easier troubleshooting. Meanwhile, integration with the Model Context Protocol enables APIs to function as callable agent tools, generate MCP servers, and connect with Postman’s newly launched MCP server network. The company is also introducing workflow integrations designed to accelerate API delivery and shorten development cycles. The integration with GitHub enables real-time collection synchronization and branch-based governance, while Jira supports context-aware issue tracking. Postman is strengthening collaboration among developer teams by linking its platform with Slack and Microsoft Teams.