2026-06-16 21:30
Databricks is working to merge the databases used for business operations with those for analysis. At the Data + AI Summit, they introduced the Lake Transactional/Analytical Processing (LTAP) architecture, aiming to streamline data management for AI agents. This approach addresses long-standing challenges in data infrastructure, allowing agents to access both transactional and historical data seamlessly. CEO Ali Ghodsi emphasized that LTAP is a significant advancement that could reshape...
Source: The New Stack
Frederic Lardinois
2026-06-16 18:30
🚀 SpaceX has announced the acquisition of Anysphere, Inc., the maker of AI coding agent Cursor, for $60 billion. This comes shortly after SpaceX's IPO and is expected to finalize in Q3 2026. 🤝 The partnership began in April, allowing SpaceX to enhance Cursor's model training efforts. This acquisition is seen as a move to gain valuable engineering and AI talent. 🔍 Experts suggest that the focus may be on consolidating elite AI talent as the coding assistant market becomes increasingly...
Source: The New Stack
Meredith Shubel
2026-06-16 15:15
Amazon Web Services has introduced the AWS FinOps Agent, a new AI tool designed to help manage cloud spending. This agent simplifies the process of analyzing cost spikes by correlating changes with AWS CloudTrail data. It can automatically open Jira tickets or notify teams on Slack, enhancing collaboration on cloud costs. Additionally, it offers natural-language responses to cost inquiries and generates scheduled reports in various formats. The agent operates mostly with read-only...
Source: The New Stack
Darryl K. Taft
2026-06-16 14:00
The era of siloed data is ending as AI agents evolve. 🚀 With 700 million weekly ChatGPT users, agentic AI is becoming more capable in executing tasks and workflows. According to PwC, 79% of executives report AI adoption, with many seeing productivity gains. 📈 However, challenges persist. An MIT study reveals up to 95% failure rates in enterprise AI projects. High costs and lack of quality data are key issues. 📊 For developers, a new eBook offers guidance on connecting AI agents with trusted...
Source: The New Stack
TNS Staff
2026-06-16 12:07
🚨 Anthropic has paused a billing change for its Claude Agent SDK that was set to take effect today. This decision comes after a challenging week for the company. On June 9, Anthropic launched Fable 5 and Mythos 5, only to withdraw them due to a US government export control directive. The billing change would have split usage allowances, impacting costs for developers. Now, subscribers can continue using their existing plans without increased fees. #Anthropic #ClaudeSDK #TechNews #Developers #AI
Source: The New Stack
Paul Sawers
2026-06-15 18:24
The Anthropic-Mythos-Fable story has generated significant discussion since last Friday. Here’s a brief overview of the key events. In early 2023, Anthropic clashed with the Department of Defense over AI model usage limits. This led to concerns about its models being labeled as a supply-chain risk. By April, Anthropic launched Mythos, an AI model that identified cybersecurity vulnerabilities. They initiated Project Glasswing to help tech companies strengthen their software. As of June,...
Source: The New Stack
Alex Wilhelm
2026-06-15 14:54
Cohere, a Canadian foundation model company, is shifting its focus to developers with the launch of North Mini Code, its first coding model. Previously, Cohere catered to banks and healthcare, emphasizing that AI should run on their infrastructure for data security. Now, they're addressing developers' growing concerns about model access and ownership. North Mini Code is designed for complex coding tasks and can run on a single Nvidia H100 GPU, making it practical for self-hosting. Developers...
Source: The New Stack
Paul Sawers
2026-06-15 13:50
The prompt-to-app process has improved significantly, making it easy to create apps with just a description. Tools like Replit, Lovable, and Base44 are gaining attention for this. However, a key issue arises: these apps run on the builder's cloud, not yours. This matters when transitioning from prototype to production. Without control over the environment, visibility, security, and compliance can suffer. The ability to monitor, test, and meet governance requirements becomes challenging. The...
Source: The New Stack
Oluwadamilola Oshungboye
2026-06-15 13:42
The AI industry is evolving, and new benchmarks are needed to measure the effectiveness of AI models in real-world applications. Recently, the Agent’s Last Exam (ALE) was introduced, evaluating AI performance across 55 occupations and 1,500 tasks. This tool focuses on "economically valuable work," grounded in actual business needs. Dawn Song, leading the research, emphasizes that measuring AI readiness for jobs requires practical, relevant benchmarks rather than abstract scores. #AI...
Source: The New Stack
Adrian Bridgwater
2026-06-15 12:00
Engineers communicate through various channels, but often overlook the importance of code as a communication tool. Code serves as a message for future engineers, teammates, and even your future self. It should be clear and understandable, allowing others to grasp the context without additional questions. Think of pull requests as stories, where each commit is a chapter. This approach ensures that the reasoning flows logically, making it easier to follow the development process....
Source: The New Stack
Hernan Garchtrom
2026-06-14 17:00
🚀 Xiaomi has open-sourced MiMo Code, claiming it outperforms Claude Code in long-horizon coding tasks. 🔍 Current agents often stall around 30 steps due to early hypothesis locking and error compounding. Xiaomi aims to address these common issues, focusing on reliability beyond 200 steps. 📊 A new benchmark from UC Berkeley highlights the shortcomings of coding agents by grading actual shipped work, revealing that many struggle significantly in real projects. #Xiaomi #MiMoCode #Coding #AI...
Source: The New Stack
Janakiram MSV
2026-06-14 16:00
Logs often go unread until issues arise, leaving teams with incomplete evidence for investigations. 📊 Recent shifts highlight the need for effective logging. Regulatory changes and heightened security awareness mean logs must be clear, queryable, and tied to events. 🔍 As AI-powered attacks grow, detailed logs are essential for understanding attack patterns and improving defenses. The focus is now on whether logs can provide valuable insights when it matters. #CyberSecurity #AI #Logging...
Source: The New Stack
Mohit Bansal
2026-06-14 15:00
AI is transforming how software teams operate, enabling faster code deployment. 🚀 However, with 70% of incidents linked to system updates, traditional incident response methods are insufficient. A shift towards a robust AI ecosystem is essential. The Model Context Protocol (MCP) is a key standard facilitating data exchange among tools. Yet, having MCP alone isn't enough; AI agents must access relevant data and adapt to incident response processes. 📊 By integrating various data points, AI can...
Source: The New Stack
João Freitas
2026-06-13 21:09
🚨 Anthropic has suspended its flagship models, Fable 5 and Mythos 5, following a U.S. government alert about a specific jailbreak vulnerability. This decision affects all users, as the export control order applies universally. Anthropic claims the vulnerabilities are minor and similar to those in other models. Amidst ongoing discussions, reports suggest that Amazon's CEO alerted officials about the jailbreak, leading to increased scrutiny. More updates are expected. #AI #Cybersecurity...
Source: The New Stack
Frederic Lardinois
2026-06-13 18:00
AI retrieval architectures are evolving beyond flat vector databases. A recent GigaOm CxO Decision Brief highlights the need for combining semantic search, personalization, and machine learning in production systems. 📊🤖 While vector search has transformed AI infrastructure, real-world queries require evaluating multiple signals. This complexity raises questions about whether flat vector stores can meet future demands. 🧩 The report emphasizes the advantages of tensor-native architectures,...
Source: The New Stack
Bonnie Chase
2026-06-13 17:00
JetBrains has introduced a new initiative aimed at bridging the skills gap in programming education. The JetBrains Course Creators Program allows educators on platforms like Udemy and Coursera to incorporate practical coding exercises into JetBrains IDEs. This approach emphasizes hands-on experience, enabling students to write, run, and debug code in real-world tools. JetBrains believes as AI-generated code increases, developers must strengthen their foundational skills in debugging and...
Source: The New Stack
Darryl K. Taft
2026-06-13 16:00
A shift in AI coding discourse highlights a new focus on "design loops that prompt your agents." This marks the transition from prompt-driven and spec-driven development to a third era where the loop itself is the unit of work. Developers now create systems that evaluate and refine tasks automatically. Key questions arise about verification and the role of engineers in this evolving landscape. #AI #Coding #SoftwareDevelopment #Kubernetes #TechTrends 🖥️🔄📊
Source: The New Stack
Arjun Iyer
2026-06-13 15:00
This week, Anthropic launched Fable 5, the first model in its Mythos-class tier, which claims to surpass Opus 4.8 in capability. 🤖 While there’s excitement among many experts, concerns arose about its pricing and safety features. Fable 5 costs significantly more and initially routed some tasks to the less capable Opus 4.8. 💰 After backlash, Anthropic revised the model's response policy. Testing between Fable 5 and Opus 4.8 revealed interesting insights on reasoning and coding tasks. #AI...
Source: The New Stack
Jessica Wachtel
2026-06-13 10:31
📊 Claude Fable 5 recently debuted, costing $9 for a coding test, while GPT-5.5 came in at $1.50. This highlights a new skill: model triage. Effective AI users are now distinguishing themselves by strategically selecting models based on cost and capability. 🛠️ Users are leveraging Fable for planning and review, while utilizing more affordable models for execution. Stay informed on AI trends to enhance your productivity! #AI #ModelTriage #ClaudeFable #GPT55 #TechTrends
Source: The New Stack
Matthew Burns
2026-06-13 02:12
🚨 Anthropic has disabled Fable 5 and Mythos 5 for all users just three days after their launch. This decision follows a U.S. government export control directive aimed at preventing access to these models for foreign nationals. The company claims compliance was only possible by shutting down the models entirely. Other models remain unaffected. Anthropic has expressed disagreement with the government's concerns regarding a potential jailbreak that could expose vulnerabilities, stating that...
Source: The New Stack
Matthew Burns
2026-06-12 19:50
In the world of enterprise software, many vendors are promoting AI agents supported by a central control plane. Companies like SAP, ServiceNow, and Salesforce have similar offerings. OutSystems recently highlighted its approach at the ONE conference, positioning itself as a neutral orchestrator among these systems. CEO Woodson Martin emphasized their strategy focuses on integrating without owning the underlying data. This neutrality allows OutSystems to coordinate across various platforms,...
Source: The New Stack
Frederic Lardinois
2026-06-12 16:30
Stack Overflow has been a key resource for developers for over 15 years. However, recent trends show a decline in user engagement, partly due to the rise of large language models (LLMs) like ChatGPT. 📉 As LLMs provide quick and polite answers, some users feel Stack Overflow's moderation culture has become a barrier. Discussions are ongoing about how coding agents can share knowledge similarly to developers in the past. Notably, Stack Overflow has now acknowledged this shift and is exploring...
Source: The New Stack
Paul Sawers
2026-06-11 20:38
Chainguard's recent findings highlight concerns about open-source package safety. Their new source code scanner identifies "greyware," which can be transparent yet harmful. With over 52,000 potentially dangerous packages blocked, they emphasize caution for non-technical users relying on open-source solutions. Understanding the risks is essential for safe development. 🔍💻 #CyberSecurity #OpenSource #TechSafety #Greyware #Chainguard
Source: The New Stack
Darryl K. Taft
2026-06-11 18:34
Source: The New Stack
Paul Sawers
2026-06-11 17:35
A shift is occurring in data services as AI agents are becoming more autonomous. Traditional manual data stewardship struggles to keep pace with these changes, raising concerns about data governance. 🌐 lakeFS has introduced its lakeFS for Agentic AI service to address these challenges. This solution aims to ensure governed access to data for AI workloads operating at scale. 📉 Einat Orr, CEO of lakeFS, highlights the risks of unchecked agent activity, which can lead to significant errors, such...
Source: The New Stack
Adrian Bridgwater
2026-06-11 17:00
Debugging in AI presents new challenges as traditional methods rely on deterministic software behavior. With AI, inputs can yield varying outputs, complicating the debugging process. To address this, a new approach called **prompt tracing** is proposed, capturing the entire lifecycle of AI requests. This method aims to enhance observability and reliability in AI systems. Learn more about this evolving paradigm! 🤖🔍💻 #AIDebugging #PromptTracing #SoftwareDevelopment #TechTrends
Source: The New Stack
Zziwa Raymond Ian
2026-06-11 14:30
Unlocking AI's potential in software delivery can lead to delegating up to 40% of tickets. 🤖 While AI excels in coding and other areas like planning and QA, current manual processes hinder its full capabilities. Redesigning the software delivery process is essential to prioritize AI agents. Key elements include providing context, setting guardrails, and ensuring visibility for engineers. This approach involves five phases: Planning, In Review, In Development, Preview, and Deploy. To support...
Source: The New Stack
Zohar Einy
2026-06-11 14:00
Async agents in cloud-native systems present a significant challenge: trust in their outputs. Ido Pesok highlights that as these agents operate autonomously, the need for self-verification becomes crucial. When developers step back, the agents must ensure their own accuracy, as reliance on external verification diminishes. A clean test run may not guarantee success in a complex environment, where dependencies and unexpected changes can lead to failures. Understanding this verification gap is...
Source: The New Stack
Arjun Iyer
2026-06-11 14:00
🌐 AI is evolving rapidly, shifting how businesses operate across sectors. The rise of agentic AI is set to transform technology even further, with a projected market growth from $9.14 billion to $139.19 billion by 2034. 📈 This shift demands new cloud infrastructure to support complex, autonomous systems. Companies need to prepare for this future by investing in agentic-ready solutions today. 🛠️ #AI #CloudComputing #TechInnovation #AgenticAI #FutureOfWork
Source: The New Stack
Kevin Cochrane
2026-06-11 13:00
Upgrading applications with AI coding agents presents challenges, especially for large projects. 🌐 While upgrading to Spring Boot 4 using natural language commands is appealing, the process can be time-consuming and error-prone. Developers may face multiple iterations to resolve coding issues, taking 1-2 days or more. ⏳ The Spring Petclinic example highlights these challenges, where unexpected changes and deprecated methods led to failed upgrades. 🛠️ Understanding these complexities is...
Source: The New Stack
Raquel Pau
2026-06-11 12:42
WeAreDevelopers is expanding its reach! 🌍 This Berlin-based conference, now in its 11th year, will debut in the US and India in 2026, hosting events in San Jose and Bengaluru. Co-founder Sead Ahmetovic highlights the importance of giving developers a platform to celebrate their achievements and share their stories. 📢 Join the conversation and empower the developer community! #WeAreDevelopers #TechConference #SoftwareDevelopment #DeveloperCommunity
Source: The New Stack
Nick Lucchesi
2026-06-11 12:00
🚀 Large language models are now vital in software development, generating code and infrastructure rapidly. However, trust in their output is a challenge due to the quality of training data. 🛠️ SonarSweep addresses this by improving the data used to train models, aiming to reduce bugs and vulnerabilities in AI-generated code. 📊 Models learn from both good and bad examples, which impacts code quality in real-world applications. Careful curation of training data is essential for reliable...
Source: The New Stack
Joe Tyler
2026-06-10 17:27
Engineers are facing challenges due to overwhelming observability data. 📊 While increased visibility into systems is beneficial, it often leads to confusion rather than quicker solutions. The abundance of data can result in chasing false leads and wasted time. To address this, AI agents are emerging as a solution, helping to parse information efficiently and suggest remediation pathways. 🤖 Join Datadog’s Vignesh Palaniappan on June 30 to explore these insights and discover how to enhance your...
Source: The New Stack
Alex Wilhelm
2026-06-10 17:18
🚀 Google has introduced DiffusionGemma, an experimental model that is 4x faster than previous Gemma versions. This model utilizes diffusion to generate text in parallel, refining it step-by-step. It can produce over 1,000 tokens per second on a single Nvidia H100 GPU. While it excels in speed, it does have some limitations in quality compared to Gemma 4. Find out more on HuggingFace! #Google #AI #DiffusionGemma #MachineLearning #Innovation
Source: The New Stack
Frederic Lardinois
2026-06-10 17:11
🚀 Anthropic has launched Fable 5, a new Mythos-class model that promises longer autonomous operation and improved programming skills. While early reviews highlight its efficiency over Opus 4.8, users are concerned about the short usage window and strict guardrails. 💻 Fable 5 is available for Claude Pro and enterprise users, but it comes at a cost of $10 per million input tokens and $50 for output tokens. 💸 Community feedback indicates mixed feelings, with some praising its performance while...
Source: The New Stack
Meredith Shubel
2026-06-10 17:01
Boris Cherny, head of Claude Code at Anthropic, has shifted from prompting AI directly to "writing loops." This approach, also known as loop engineering, focuses on designing systems for coding agents to operate autonomously. Peter Steinberger and Google engineer Addy Osmani emphasize that this method transforms coding agents into long-running execution systems, enhancing their efficiency and capabilities. Loop engineering combines scheduled execution, isolated workspaces, and persistent...
Source: The New Stack
Janakiram MSV
2026-06-10 16:46
AWS recently announced the general availability of its Graviton5 processor, which powers two new Amazon EC2 instances: M9g and M9gd. These instances are designed for general-purpose workloads and high-speed local storage, respectively. 🖥️ The Graviton5 features a doubled core count of 192, promising up to 25% better compute performance than its predecessor. Notably, these instances utilize the sixth-generation Nitro System, which includes the new Nitro Isolation Engine. 🔒 The Nitro Isolation...
Source: The New Stack
Frederic Lardinois
2026-06-10 16:40
Last week, Microsoft disabled 73 GitHub repositories after a malware attack aimed at stealing developers’ credentials. The incident originated from a malicious commit uploaded to the durabletask repository. Despite the shutdown, Microsoft has not disclosed how many developers were impacted or specific details about the breach. GitHub stated the repos violated its terms of service but did not provide further clarification. This event highlights ongoing security concerns in the industry, as...
Source: The New Stack
Meredith Shubel
2026-06-10 14:00
🚀 Databricks has launched OpenSharing, the successor to its Delta Sharing protocol. This new update supports Apache Iceberg REST Catalog clients and includes on-premises storage vendors like Everpure, MinIO, and Qumulo. OpenSharing enhances secure data sharing by extending capabilities to AI models and unstructured data, helping organizations share assets effectively. Since its debut in 2021, Delta Sharing has become widely adopted, with thousands of partners benefiting from its open...
Source: The New Stack
Frederic Lardinois
2026-06-10 13:00
🚀 Ramp is introducing Applied AI Solutions, a service that deploys forward deployed engineers (FDEs) into finance teams to enhance automation in accounts payable, procurement, and more. Unlike traditional AI tools, this high-touch service focuses on understanding the unique context of each enterprise. Ramp aims to bridge the gap between AI ambition and results, ensuring that finance decisions are informed and governed effectively. #AI #Finance #TechInnovation #Automation #BusinessSolutions
Source: The New Stack
Matthew Burns