New custom chips and advanced AI models are enabling the deployment of sophisticated AI agents capable of performing complex, real-world tasks. This trend is reshaping enterprise operations, from financial transactions executed by AI agents to the creation of searchable service directories. The focus is shifting toward building secure, agentic systems on top of foundational hardware.
OpenAI and Broadcom unveiled Jalapeño, a custom inference chip designed to run ChatGPT faster and more efficiently than general-purpose GPUs.
OpenAI and Broadcom have just unveiled Jalapeño, a purpose-built inference chip that promises to redefine AI compute efficiency. This marks a pivotal moment in AI hardware, as OpenAI aims to reduce reliance on Nvidia's GPUs while delivering industry-leading performance per watt. For companies scaling AI workloads, this could translate to lower costs and faster deployment cycles. The chip's focus on inference optimization also signals a shift toward more specialized hardware in the AI supply chain. How will this development influence your organization’s AI infrastructure strategy in the next 12 months?
OpenAI released GPT-5.5 Instant, an updated version with improved conversational abilities, intent recognition, and recommendation features.
OpenAI has launched GPT-5.5 Instant, a major update to its conversational AI that promises stronger intent recognition, context retention, and personalized recommendations. This release underscores the trend toward AI that adapts in real-time to user needs, making interactions feel more natural and intuitive. For businesses integrating AI into customer-facing applications, these improvements could lead to higher engagement and conversion rates. The focus on shopping and local recommendations also hints at a broader shift toward AI-driven decision support. How will your team leverage these enhanced conversational capabilities to improve customer interactions?
Claude Fable 5 and Mythos are approaching public availability after being offline due to export restrictions.
Anthropic’s Claude Fable 5 and Mythos models are finally nearing public release after being offline due to export restrictions. Recent developments, including federal lawsuits and congressional pressure, suggest a phased rollout may already be underway. This development is critical as it highlights the challenges of global AI deployment amid geopolitical constraints. For enterprises relying on advanced AI models, this could mean renewed access to cutting-edge capabilities. How will your organization adapt to potential shifts in model availability and compliance requirements?
Google introduced native computer-use capabilities for Gemini 3.5 Flash, enabling AI agents to interact with desktop interfaces.
Google has taken a significant step toward autonomous AI agents with the introduction of native computer-use capabilities in Gemini 3.5 Flash. This feature allows AI models to navigate desktop interfaces by clicking, scrolling, and interacting with applications through continuous screenshots. For developers, this opens up a new frontier in automation, enabling AI to perform real-world tasks across different software environments. The implications for productivity and workflow automation are profound. How will your team integrate these agentic capabilities into existing systems?
A fund manager uses machine learning models to execute buy/sell orders for their fund.
Machine learning isn’t just for Silicon Valley startups anymore—it’s now powering investment strategies at the fund level. A fund manager is leveraging machine learning models to automate buy/sell orders, demonstrating how AI is transforming asset management. This approach allows for data-driven decision-making, scalability, and potentially improved risk-adjusted returns. As AI adoption in finance accelerates, professionals must ask: Are traditional investment methods still sufficient, or is AI-driven investing the future of portfolio management?
A fund manager is finalizing an offering memorandum and subscription package for a private fund launch.
The road to launching a fund is paved with paperwork—and precision. A fund manager is finalizing an offering memorandum and subscription package, a critical step in ensuring regulatory compliance and investor transparency. This process streamlines the onboarding experience for high-net-worth individuals and institutional investors, setting the stage for a smooth fund launch. For those in the industry, this underscores the importance of meticulous preparation in fund operations. How can fund managers balance compliance demands with operational efficiency?
Stablecore launched a stablecoin and digital asset programme for credit unions in collaboration with Circuit.
Stablecore and Circuit have launched a groundbreaking programme to enable credit unions to offer stablecoins, tokenised deposits, and digital asset products. This initiative marks a pivotal moment for community banks, bridging traditional banking with modern digital assets. By leveraging proven Credit Union Service Organisation (CUSO) infrastructure, the programme ensures regulatory alignment and operational scalability. As credit unions seek to remain competitive, this could accelerate mainstream adoption of digital assets in everyday banking. How can traditional financial institutions balance innovation with regulatory compliance in this rapidly evolving landscape?
Mastercard and PrivatBank completed Ukraine’s first payment executed by an AI agent using agentic commerce framework.
Mastercard and PrivatBank have executed Ukraine’s first payment processed by an AI agent, marking a transformative step in agentic commerce. This breakthrough enables AI-powered agents to autonomously manage shopping, subscriptions, and daily transactions while ensuring full transparency and security. As AI agents become visible and governed participants in payment flows, we’re entering a new era of automated commerce where efficiency meets trust. What opportunities or challenges do you foresee as AI agents increasingly participate in consumer transactions?
Stripe released Directory in public preview, enabling developers and AI agents to discover businesses and services via a searchable network.
Stripe has launched Directory in public preview, a searchable network that allows developers and AI agents to discover businesses, services, and payment endpoints. This positions Stripe as critical infrastructure for agentic commerce, enabling machines to programmatically search, provision, and pay for services. As AI agents evolve from assistants to autonomous actors, Stripe’s move cements its role in the backend of the digital economy. How will your business adapt to an ecosystem where AI agents are the primary users and integrators of digital services?
Modern Treasury partnered with Sardine to enhance transaction monitoring and fraud detection for businesses.
Modern Treasury and Sardine have joined forces to deliver enhanced transaction monitoring and fraud detection capabilities across global payment flows. In an era where financial crime is increasingly sophisticated, this partnership promises businesses earlier detection and faster response times. By integrating Sardine’s real-time risk engine with Modern Treasury’s network, companies can now protect their payment rails with cutting-edge AI-driven insights. How can your organisation stay ahead of evolving fraud tactics while maintaining seamless user experiences?
Thredd implemented Visa Cloud Connect (VCC) in Asia Pacific as part of its cloud transformation strategy.
Thredd has rolled out Visa Cloud Connect (VCC) in the Asia Pacific region, accelerating its cloud-first transformation for issuer processing. This move allows Thredd to deliver scalable, resilient, and globally interoperable payment solutions powered by Visa’s cloud infrastructure. For fintechs and banks in APAC, this means faster time-to-market and reduced operational overhead. As cloud-native architectures become the norm in payments, how will legacy systems adapt to remain competitive?
Intercontinental Exchange (ICE) and OKX formed a joint venture to build a regulated US broker-dealer and futures commission merchant for tokenized equities.
ICE and OKX have announced a 50-50 joint venture to establish a regulated US broker-dealer and futures commission merchant for tokenized NYSE equities and ICE futures. This partnership underscores the accelerating convergence between traditional finance and digital assets, offering institutional-grade access to on-chain securities. With ICE’s regulatory expertise and OKX’s global user base, this venture could redefine how institutions interact with tokenized markets. How might your firm position itself to capitalise on the growing demand for regulated tokenized assets?
Meta invested $900 million in Indian fintech Cred, acquiring a 20% stake and appointing Cred’s founder to lead WhatsApp in India.
Meta has made a bold $900 million investment in Cred, India’s leading fintech platform, securing a 20% stake and appointing Cred’s founder, Kunal Shah, to lead WhatsApp in the region. This deal signals Meta’s intensified focus on India’s fast-growing digital economy, particularly in fintech and payments. By integrating Cred’s capabilities with WhatsApp’s massive user base, Meta aims to supercharge financial inclusion and commerce. How will this partnership reshape India’s digital payments landscape and consumer behaviour?
Anthropic is retiring its existing Claude in Slack connector in favor of an always-on, agentic Slack coworker called Claude Tag.
Anthropic is reimagining AI collaboration with the retirement of its Claude in Slack connector and the introduction of Claude Tag—a new always-on, agentic AI coworker. This shift moves beyond simple chat interactions to a proactive team member that learns from channel context and can be delegated tasks with ambient updates or scheduled work. For enterprises, this represents a leap toward AI-driven workflows that operate seamlessly within existing tools. The ability to delegate multi-hour or multi-day tasks to an AI agent could redefine productivity in remote and hybrid work environments. How might your team leverage an always-on AI coworker to streamline daily operations?
Qualcomm is acquiring Modular for about $3.9B to strengthen its AI software layer and compete with Nvidia's CUDA ecosystem.
Qualcomm's $3.9B acquisition of Modular marks a bold move to challenge Nvidia's CUDA ecosystem in the AI software space. By enabling AI workloads to run across different chips without rewriting software, Modular's technology could democratize AI development across heterogeneous hardware. This acquisition underscores the growing importance of software layers in AI, beyond just hardware acceleration. For tech leaders, it signals a shift toward more flexible and interoperable AI infrastructure. How can your organization prepare for a future where AI workloads seamlessly span diverse hardware platforms?
Gartner predicts AI coding costs may surpass average developer salaries by 2028 due to token-based pricing and increased usage.
Gartner's latest analysis raises a red flag: AI coding costs could outpace developer salaries by 2028, driven by token-based pricing and unchecked usage. As AI tools become integral to workflows, IT leaders must adopt FinOps-style governance to track usage, assign ownership, and ensure measurable business value. The days of unmonitored AI experimentation are numbered—costs compound quickly when left unchecked. Are you confident your organization has the right controls in place to manage AI spending before it spirals out of control?
Data lakehouses are becoming the foundation for enterprise AI due to their combination of broad data access, governance, and security.
Data lakehouses are emerging as the backbone of enterprise AI, merging the scalability of data lakes with the structure and governance of warehouses. As agentic AI systems demand semantic context, audit trails, and strict access controls, lakehouses provide the ideal foundation. This shift highlights the growing importance of the data layer in AI deployments, where security and compliance are non-negotiable. For data leaders, this means rethinking traditional data architectures to support autonomous AI systems. How will your organization adapt its data infrastructure to meet the demands of next-gen AI agents?
Bunny.net is eliminating DNS query fees and offering free DNS hosting for up to 500 domains per account.
Bunny.net is disrupting the DNS market by making it free with its new offering—no query limits, no per-request billing, and features like smart records and health monitoring included. This move is a strategic play to capture more edge platform users by lowering the barrier to adoption. For DevOps teams, free DNS simplifies the integration of CDN acceleration, edge security, and modern record management. How might free DNS change your approach to cloud and edge infrastructure planning?
A new website tracks major apps and services that do not offer passkey login, including Instagram, Netflix, and Spotify.
Security researcher Scott Helme has launched a website to publicly shame apps that still lack passkey support, including household names like Instagram, Netflix, and Spotify. As passkeys shift from a security upgrade to a trust signal, organizations risk reputational damage by ignoring modern authentication methods. This trend underscores the growing consumer demand for frictionless, secure logins. How will your organization address the passkey gap before it becomes a liability?
Cisco introduces AI Troubleshooting for Industrial Networks to monitor factory-floor network signals and diagnose issues faster.
Cisco is bringing AI to the factory floor with its new AI Troubleshooting for Industrial Networks, an always-on agent that monitors network signals, clusters alerts, and diagnoses root causes in real time. By identifying issues like bad cables or VLAN mismatches faster, it reduces downtime and the need for scarce network experts. For OT teams, this is a game-changer in proactive maintenance and operational efficiency. How can your organization leverage AI to minimize costly downtime in critical infrastructure?
Jamf highlights Declarative Device Management (DDM) for real-time Mac fleet visibility and faster compliance tracking.
Jamf is elevating Mac fleet management with Declarative Device Management (DDM), which provides real-time visibility and proactive status reporting instead of traditional polling. This shift enables faster compliance tracking, quicker update visibility, and reduced manual follow-up for lean IT teams. For organizations with large Apple deployments, DDM is a step toward more responsive and efficient device management. How can your team benefit from proactive, real-time fleet insights?
Superhuman, formerly Grammarly, is acquiring GPTZero to add AI detection, plagiarism checks, and writing-origin tools to its platform.
Superhuman, the company formerly known as Grammarly, is acquiring GPTZero to deepen its AI-powered productivity suite with detection, plagiarism checks, and writing-origin tools. This move signals a broader trend toward AI tools that not only generate content but also ensure its integrity and authenticity. For teams relying on AI-generated outputs, this could be a critical safeguard. How will your organization balance the productivity gains of AI with the need for content verification?
Google Meet hardware now supports room codes in early preview, allowing users to start meetings from rooms using short codes.
Google Meet hardware is introducing room codes in early preview, simplifying meeting room access with short codes instead of manual invites or calendar events. This feature reduces friction for users in shared spaces and aligns with the growing demand for intuitive, accessible meeting solutions. For IT teams managing hybrid workspaces, this could streamline room booking and usage tracking. How might your organization leverage room codes to improve the meeting experience for employees and guests?
A sports club founder was jailed for committing £700,000 in Gift Aid fraud.
Fraud remains a pressing challenge for charities, as evidenced by a sports club founder sentenced to prison for £700,000 in Gift Aid fraud. This case highlights the importance of robust internal controls and transparency in financial reporting for organizations handling public funds. Compliance teams must prioritize due diligence and real-time monitoring to prevent misuse of charitable donations. What steps is your organization taking to safeguard against financial misconduct and maintain donor trust?
Former Google engineer Justin Poehnelt was fired for creating an open-source Google Workspace CLI tool that allows AI agents and humans to control Workspace apps via command line.
Google’s decision to fire an engineer for building an open-source tool that made Workspace apps more accessible via CLI has sparked widespread debate. The Google Workspace CLI, which went viral on Hacker News and gained thousands of GitHub stars, exemplifies the frustration many developers feel toward clunky, admin-heavy enterprise tools. This incident underscores a growing disconnect between corporate incentives and developer innovation. How can companies better align their policies with the tools their engineers and users actually need?
Noam Shazeer, co-lead of Google’s Gemini model, left for OpenAI less than two years after returning from Character.AI.
Noam Shazeer, the co-lead behind Google’s Gemini model, has returned to OpenAI—just two years after Google paid $2.7B to bring him back from Character.AI. Shazeer’s move highlights the intense talent war shaping the AI industry, where top researchers frequently cycle between competitors. For companies, this underscores the challenge of retaining AI talent amid rapid advancements and open competition. What strategies can organizations employ to foster loyalty and innovation without stifling mobility?
GPT-5 Pro helped solve a 3-year-old immunology mystery about how immune cells fight cancer.
A team led by immunologist Derya Unutmaz used GPT-5 Pro to crack a 3-year-old mystery about how T cells specialize to fight cancer and viruses. The AI model analyzed complex biological data to reveal new pathways, opening doors for targeted cancer therapies and autoimmune disease treatments. This case exemplifies AI’s growing role in accelerating scientific discovery, bridging gaps where human analysis falters. How can researchers and institutions better integrate AI into their workflows to unlock similar breakthroughs?
Fable 5 may return as part of Claude Code’s subscriptions with weekly usage limits, hinted by strings in v2.1.190.
Rumors suggest Fable 5, an AI-generated game, could make a comeback as part of Claude Code’s subscription plans, with weekly usage limits. The hints emerged from strings in the latest update (v2.1.190), sparking excitement among developers and gamers alike. If confirmed, this would blend AI creativity with monetization, reflecting a broader trend of integrating AI tools into subscription ecosystems. How might subscription-based AI services transform the gaming and creative industries?
Nabla Bio’s JAM-2 became the first AI model to design drug-quality antibodies from a computer, matching traditional lab success rates.
Nabla Bio’s JAM-2 has achieved a first: an AI model that designs drug-quality antibodies directly from a computer, achieving binding success rates comparable to traditional lab methods. This breakthrough could drastically reduce the time and cost of drug discovery, particularly for hard-to-target proteins like those in cancer research. For biotech and pharma companies, this signals a paradigm shift toward AI-driven R&D. How soon will we see AI-designed antibodies moving from labs to clinical trials?
Arc Institute released Proto, an open framework integrating multiple AI biology tools for combined protein, RNA, and gene regulator design.
Arc Institute has launched Proto, an open framework that connects multiple AI tools for designing proteins, RNA, and gene regulators in tandem. By enabling holistic, multi-modal design, Proto empowers researchers to tackle complex biological challenges more efficiently. This open-source approach democratizes access to cutting-edge AI tools, accelerating innovation across the life sciences. How might such integrative frameworks redefine collaboration in scientific research?
NVIDIA released its BioNeMo Agent Toolkit, a framework for building AI agents that automate repetitive research tasks in life sciences and drug discovery.
NVIDIA has unveiled its BioNeMo Agent Toolkit, a framework that lets researchers build AI agents capable of autonomously reading papers, generating hypotheses, writing code, and iterating on results—all without human intervention. Aimed at life sciences and drug discovery, this toolkit could significantly accelerate early-stage research by automating tedious tasks. For industries reliant on scientific innovation, this represents a leap toward agentic AI in R&D. How will AI agents transform the role of human researchers in the coming decade?
Netflix replaced its Compute Managed Batch system with Kueue, a Kubernetes-native job queuing system, to simplify batch compute operations.
Netflix has just streamlined its batch compute infrastructure by migrating from its legacy Compute Managed Batch system to Kueue, a Kubernetes-native job queuing and scheduling platform. This shift isn’t just about infrastructure optimization—it’s a testament to how modern systems prioritize scalability and fairness through preemption-based resource sharing. By maintaining API compatibility and restructuring tenant hierarchies into Kueue Cohorts, Netflix demonstrates how legacy systems can evolve without disruption. For companies managing large-scale batch workloads, this move underscores the growing importance of Kubernetes-native solutions in driving operational efficiency. How might your team rethink legacy batch systems in light of these advancements?
DoorDash engineer shares insights on building long-running AI agents grounded in structured context and persistent artifacts.
Ever wonder how to make long-running AI agents truly effective? A DoorDash engineer’s deep dive reveals that grounding agents in structured context—like persistent artifacts, repo research, and verification reports—turns chaotic workflows into state-machine-driven processes. By leveraging Agentic Orchestrator to manage handoffs, these agents maintain coherence even when tackling complex, multi-step tasks. In an era where AI agents are expected to handle increasingly sophisticated workflows, this approach highlights the critical role of structured data in enabling reliable, scalable automation. What aspects of your workflows could benefit from this kind of structured, agentic approach?
Zepto built a Dual Sequence ReRanker for real-time personalization combining user history and current session behavior.
Zepto’s new Dual Sequence ReRanker is redefining real-time personalization by dynamically combining a user’s long-term history with their in-session behavior. Using separate transformer encoders for history and session data, along with target-aware pooling and real-time signals, this system ensures that recommendations stay relevant even as user intent shifts. For e-commerce and content platforms, this represents a leap toward hyper-personalized experiences that adapt instantly—without sacrificing performance. How will your team leverage real-time behavioral signals to enhance user engagement?
Zalando implemented a high-performance client-side load balancer in-process for their Product Read API to handle over a million requests per second.
Handling over a million requests per second isn’t just a benchmark—it’s a reality for Zalando’s Product Read API, thanks to their new client-side load balancer. Built in-process, this solution replaces edge ingress Skipper for batch requests while maintaining consistent hashing and Kubernetes watch-based discovery. With features like N-ring fade-in for smooth scaling and AZ-aware routing, it sets a new standard for high-throughput, low-latency systems. For teams grappling with internal fan-out traffic, this approach offers a blueprint for balancing performance and reliability. What bottlenecks in your infrastructure could benefit from similar client-side optimizations?
Technically excellent data teams often fail to drive business impact due to lack of strategic influence and decision-making focus.
Strong technical execution isn’t enough—data teams must also drive decisions. Many technically brilliant teams remain irrelevant because they focus solely on delivering data without shaping business outcomes. The shift requires clear perspectives (opinionated analysis) and proactive influence on decision-making. For data leaders, this means moving beyond ‘reporting’ to ‘reshaping’—where data isn’t just a byproduct but a catalyst for action. How can your team transition from data providers to strategic partners?
Clustering unstructured text with LLM embeddings and HDBSCAN enables automatic topic discovery in unlabeled documents.
Traditional text clustering methods are getting a major upgrade with LLM embeddings and HDBSCAN. This approach combines sentence-transformers for semantic richness with UMAP for dimensionality reduction and HDBSCAN for density-based clustering—automatically determining the number of topics while gracefully handling noise. For teams drowning in unstructured data, this modern pipeline offers a scalable way to extract insights without manual labeling. How could your organization leverage automated topic discovery to uncover hidden patterns in your data?
PJ One Piece reduced analytics lead time from two weeks to ten minutes using a generative AI agent.
What if your analytics lead time dropped from weeks to minutes? PJ One Piece achieved this feat with a generative AI agent, cutting analysis from two weeks to just ten minutes. By leveraging staged metadata, SQL guardrails, and supervisor agents, they’ve enabled hundreds of analyses while freeing data scientists to focus on higher-value discovery. This isn’t just about speed—it’s a paradigm shift in how teams operate, moving from reactive reporting to proactive, real-time insights. How could similar automation transform your team’s output and agility?
Future calendar events should store user’s intended local time and timezone, not just UTC, to avoid issues from timezone rule changes.
Storing calendar events in UTC might seem like the right call, but timezone rules can turn that into a ticking time bomb. A new recommendation suggests storing the user’s intended local time and timezone instead, ensuring appointments remain accurate even as political boundaries shift. This isn’t just a niche issue—it affects global teams, scheduling apps, and any system handling user-facing time data. For developers, this is a reminder that edge cases often become critical failures. How is your team handling timezone-sensitive data, and are you prepared for the next rule change?
Comments