DeepSeek’s latest version of its hit open source model DeepSeek, R1-0528 is already being adapted and remixed by other AI labs and developers, thanks in large part to its permissive Apache 2.0 license. German firm TNG Technology Consulting GmbH released one such adaptation: DeepSeek-TNG R1T2 Chimera, the latest model in its Chimera large language model (LLM) family. R1T2 delivers a notable boost in efficiency and speed, scoring at upwards of 90% of R1-0528’s intelligence benchmark scores, while generating answers with less than 40% of R1-0528’s output token count. That means it produces shorter responses, translating directly into faster inference and lower compute costs. This gain is made possible by TNG’s Assembly-of-Experts (AoE) method — a technique for building LLMs by selectively merging the weight tensors (internal parameters) from multiple pre-trained models. R1T2 is constructed without further fine-tuning or retraining. It inherits the reasoning strength of R1-0528, the structured thought patterns of R1, and the concise, instruction-oriented behavior of V3-0324 — delivering a more efficient, yet capable model for enterprise and research use.
Dust helps enterprises build AI agents capable of taking real actions across business systems and secures sensitive information by separating data access rights from agent usage rights
AI platform Dust helps enterprises build AI agents capable of completing entire business workflows, has reached $6 million in annual revenue — a six-fold increase from $1 million just one year ago. The company’s rapid growth signals a shift in enterprise AI adoption from simple chatbots toward sophisticated systems that can take concrete actions across business applications. The startup has been selected as part of Anthropic’s “Powered by Claude” ecosystem, highlighting a new category of AI companies building specialized enterprise tools on top of frontier language models rather than developing their own AI systems from scratch. Instead of simply answering questions, Dust’s AI agents can automatically create GitHub issues, schedule calendar meetings, update customer records, and even push code reviews based on internal coding standards–all while maintaining enterprise-grade security protocols. The shift toward AI agents that can take real actions across business systems introduces new security complexities that didn’t exist with simple chatbot implementations. Dust addresses this through a “native permissioning layer” that separates data access rights from agent usage rights. The company implements enterprise-grade infrastructure with Anthropic’s Zero Data Retention policies, ensuring that sensitive business information processed by AI agents isn’t stored by the model provider.
Agent2.Ain’s AI agent can instantly turn complex research tasks into usable outputs in multi-formats like structured spreadsheets and presentation slides through a transparent, step-by-step breakdowns of how it searched, evaluated sources, and reached conclusions
Agent2.Ain has launched Super Agent, a powerful new AI tool designed to help users tackle complex research tasks and instantly turn them into usable outputs—like structured spreadsheets, presentation slides, and more. What sets Super Agent apart is its open process. Every step in its reasoning is visible—users can review, edit, or guide the workflow in real time. Behind the scenes, multiple AI models collaborate on each task. The system compares their outputs, refines them, and delivers a final version that reflects stronger reasoning from multiple angles. Super Agent fits into existing workflows with support for formats like Excel, PowerPoint, Docs, Markdown, and more. And when deeper context is needed, users can securely log in to enterprise tools within a virtual machine, allowing the agent to factor in private business data alongside public research. The Agent2.AI Super Agent is designed to take a prompt and deliver usable results across multiple formats and tools. Some examples of what it can do include: Deep Research: Transparent, step-by-step breakdowns of how the agent searched, evaluated sources, and reached its conclusions;
AI Sheets: Structured spreadsheets that organize research findings, metrics, and summaries. Exportable with one click; AI Slides: Presentation decks built from research or reports, complete with titles, visuals, and speaker notes; Other Outputs: From timelines and tables to emails and internal docs, the agent adapts its output based on what the user needs.
FortePay’s integration with Sequence payments infra platform to offer end-to-end, multi-chain payment solution with integrated local payments and currency options and dynamic compliance tools to power web3 monetization
Forte and its wholly-owned subsidiary LemmaX has announced a strategic alliance with Sequence to integrate FortePay as a central component of the Sequence platform infrastructure. This collaboration equips developers with an end-to-end, multi-chain payment solution that enables compliant web3 payment processing and seamless value transfer, advancing the shared vision of creating safe, user-friendly applications with versatile monetization pathways in the web3 ecosystem. Through this partnership, FortePay will serve as Sequence’s foundational payments infrastructure layer, providing global card payment network coverage with integrated local payments and currency options. The solution incorporates dynamic compliance tools featuring graduated KYC flows and asset-based ID verification, resulting in faster, more efficient purchase flows. Sequence users will be able to seamlessly buy and convert between crypto and fiat currencies, purchase NFTs using fiat or cryptocurrency, and transact with custom tokens through the platform’s comprehensive capabilities. “Integrating FortePay into Sequence’s core infrastructure unlocks a new level of flexibility and compliance tools for developers building web3 applications,” said Taylan Pince, CTO of Sequence. “This partnership enables seamless, secure, and scalable payment flows, whether in fiat or crypto, without compromising on user experience. It brings us one step closer to a future where web3 monetization is as intuitive and robust as any traditional platform.”
SIFMA is concerned about SEC’s exemptive orders for trading tokenized securities on digital asset platforms and wants a rulemaking process which allows for public notice and comment, oversight, and broad industry engagement
The CEO of the Securities Industry and Financial Markets Association (SIFMA) has written to the SEC expressing concerns relating to tokenized securities. SIFMA members have read press reports implying that digital asset firms might receive SEC no action letters or exemptive orders for trading tokenized traditional securities on digital asset platforms. In other words, the digital asset firms won’t have to comply with the same securities rules that SIFMA members adhere to. Given this could have far reaching consequences for capital markets, SIFMA would prefer to see the process follow the SEC rulemaking process “which allows for public notice and comment, oversight, and broad industry engagement.” SIFMA raised similar issues in a longer letter last month responding to SEC requests for information. An example given by SIFMA was Coinbase looking to trade tokenized securities in the United States. The potential ripple effects become clear when examining Coinbase’s proposals, which represent very substantial changes to market structure, although they expect that parts may be implemented via rulemaking. One potentially controversial example is the best execution rule requiring that a broker shouldn’t execute an order when there’s a better price displayed on another venue. Coinbase argues that this only accounts for price rather than the size of the order or the ability to meet demand at that price. It claims that bitcoin and ether have “better market quality measures” than the vast majority of stocks. “To this end, the Commission should give markets an opportunity to operate without the cumbersome artifice of legacy order routing requirements until and unless markets do not develop in a way that demonstrates they cannot or chose not to provide high execution quality,” Coinbase argued.
Startup Qedma’s software specializes in quantum error suppression and error mitigation by analyzing noise patterns to suppress some classes of errors while the algorithm is running and mitigate others in post-processing
Startup Qedma specializes in error-mitigation software. Its main piece of software, QESEM, or quantum error suppression and error mitigation, analyzes noise patterns to suppress some classes of errors while the algorithm is running and mitigate others in post-processing. IBM is both working on delivering its own “fault-tolerant” quantum computer by 2029 and collaborating with partners like Qedma. That’s because IBM thinks driving quantum further requires a community effort. “If we all work together, I do think it’s possible that we will get scientific accepted definitions of quantum advantage in the near future, and I hope that we can then turn them into more applied use cases that will grow the industry,” VP of Quantum, Jay Gambetta said. In all likelihood, it will first apply to an academic problem, not a practical one. In this context, it may take more than one attempt to build consensus that it’s not just another artificial or overly constrained scenario. Since last September, Qedma has been available through IBM’s Qiskit Functions Catalog, which makes quantum more accessible to end users. Qedma’s plans are hardware-agnostic. The startup has already conducted a demo on the Aria computer from IonQ, a publicly listed U.S. company focused on trapped ion quantum computing. In addition, Qedma has an evaluation agreement with an unnamed partner Sinay described as “the largest company in the market.” Recently, it also presented its collaboration with Japan’s RIKEN on how to combine quantum with supercomputers.
Pinwheel’s smartwatch for kids aged 7 to 14 prevents access to social media and internet features an AI chatbot that enables them to ask questions about everyday curiosities, social interactions, and homework-related questions
Pinwheel, a kid-friendly tech company, is introducing a new solution for parents who want to stay connected with their children without giving them a phone. The Pinwheel Watch is a recently launched smartwatch designed specifically for kids aged 7 to 14, offering a child-safe alternative that prevents access to social media and the internet. It features parental management tools, GPS tracking, a camera, voice-to-text messaging, fun mini-games, and — here’s a surprise — an AI chatbot. The smartwatch itself features a sleek black design and a screen that is slightly larger than that of an Apple Watch. In addition to a more standard set of parental controls, the feature some parents might be wary of is the watch’s AI assistant, “PinwheelGPT.” PinwheelGPT is designed as a safer alternative to typical AI chatbots, enabling kids to ask questions about various topics, including everyday curiosities, social interactions, and homework-related questions. In addition to the AI feature, kids and tweens can make calls and send texts on the watch by using voice commands or a keyboard. There’s also a camera for video calls and selfies, along with a voice recorder app. The parent-monitoring features are available through the “Caregiver” app. This allows parents to create a “Safelist” of contacts that their children are permitted to talk to, as well as reject certain phone numbers from being added to the list.
One Big Beautiful Bill Act offers advisors an avenue to be really proactive with their clients by sharing their thoughts and opinion on the bill’s massive impact on the rules for federal income taxes and estate planning
After President Donald Trump’s Republican allies raced to meet their July 4 deadline to pass the One Big Beautiful Bill Act, the legislation is on its way to be signed into law. Financial advisors and their clients can now take the rest of the year to plan for 2026 and beyond. The legislation extends and expands many provisions of the Tax Cuts and Jobs Act and will have a massive impact on the rules for federal income taxes and estate planning, alongside other Trump administration priorities such as defense and border security appropriations, work requirements for Medicaid beneficiaries and an increase to the debt ceiling. Trillions of dollars in additional federal debt as a result of the newly passed legislation pose further questions for investors. Over the next decade, the bill will expand deficits by $3.2 trillion, after savings of $1.4 trillion on the overall cost of $4.6 trillion, according to the Penn Wharton Budget Model. Beyond the political upshot and inevitable arguments around the economic impact of the legislation, advisors and their clients will likely want to prepare for an array of new tax rules coming into effect as early as this year. No matter their political bent or opinion on the law, it is “exciting that they can take advantage of something like that,” said Mike Byrnes, founder of advisor growth firm Byrnes Consulting. Since clients will no doubt be asking advisors’ thoughts, it makes a great topic for, say, a client or prospect event, he noted. “It just gives advisors another thing to be really proactive with their clients about,” Byrnes said. “Whether the client leans left or leans right, I think it’s a great opportunity to strengthen their relationship and just be in front of them.”
Paytiko’s AI-powered tool enables merchants to reduce failed transactions and predict customer churn by analyzing a multitude of variables including historical payment success rates, time of day, processor reliability, and regional performance
Paytiko’s Growth Hub emerges as a transformative, AI-powered solution that not only simplifies payment management but also drives strategic growth through data-driven insights and automation. At its core, the Paytiko Growth Hub leverages artificial intelligence to elevate how merchants engage with their payment ecosystems. Far beyond static reporting dashboards, the Growth Hub introduces predictive analytics, intelligent automation, and real-time optimization to empower merchants with actionable intelligence. At its core, the Paytiko Growth Hub leverages AI to elevate how merchants engage with their payment ecosystems. Far beyond static reporting dashboards, the Growth Hub introduces predictive analytics, intelligent automation, and real-time optimization to empower merchants with actionable intelligence. By analyzing a multitude of transaction variables—including historical payment success rates, time of day, processor reliability, and regional performance—the Growth Hub enables merchants to reduce failed transactions and boost payment success rates proactively. These predictive capabilities extend to cash flow forecasting, providing merchants with clear visibility into their expected revenues and enabling more accurate financial planning. The tool’s versatility is particularly valuable for subscription-based businesses. Through AI-driven churn prediction, merchants can identify customers at risk of cancellation and implement timely, targeted retention strategies that directly improve customer lifetime value. Simultaneously, real-time fee optimization guides merchants toward the most cost-effective processors for each transaction, significantly reducing operational expenses.
Meta is partnering data labelling firm Alignerr to build customizable chatbots that reach out to users unprompted, send proactive message and follow up on any past conversations
Meta is working with Data labelling firm Alignerr to train customizable chatbots to reach out to users unprompted and follow up on any past conversations. That means the bots, which users can create in Meta’s AI Studio platform, also remember information about users. AI persona named “The Maestro of Movie Magic” might send as a proactive message on Messenger, WhatsApp, or Instagram. The AI chatbots will only send follow-ups within 14 days after a user initiates a conversation and if the user has sent at least five messages to the bot within that time frame. Meta says the chatbots won’t keep messaging if there’s no response to the first follow-up. Users can keep their bots private or share them through stories, direct links, and even display them on a Facebook or Instagram profile. To address safety, Meta gave a series of disclaimers. One of them warns that an AI’s response “may be inaccurate or inappropriate and should not be used to make important decisions.” Another says that the AIs aren’t licensed professionals or experts trained to help people. One of them warns that an AI’s response “may be inaccurate or inappropriate and should not be used to make important decisions.” Another says that the AIs aren’t licensed professionals or experts trained to help people.