<h2>Must know</h2>
<ul>
<li><strong>Cloud security startup raises big bucks for its “inside-out” approach</strong>: Upwind Security, founded by the team behind <a href="http://Spot.io" target="_blank" rel="noopener nofollow">Spot.io</a>, secured a hefty $250 million in Series B funding, valuing the company at $1.5 billion. They’ve achieved 900% year-over-year revenue growth by focusing on an “inside-out” runtime cloud security platform that analyzes internal signals, a departure from traditional agentless scanning. The funding, led by Bessemer Venture Partners, will fuel expansion into AI security and developer-focused misconfiguration prevention, as reported by TechCrunch. <a href="https://app.daily.dev/posts/LGgCBPA2D" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>India offers massive tax breaks to lure global AI workloads</strong>: India is offering a tax holiday through 2047 for foreign cloud providers running AI workloads from Indian data centers, provided services are sold outside the country. This move aims to attract major investments from tech giants like Amazon, Google, and Microsoft, who have already committed over $100 billion to Indian infrastructure. However, challenges like power shortages and high electricity costs could hinder scaling data center capacity, according to TechCrunch. <a href="https://app.daily.dev/posts/UMOBKmYfu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Open-source SwiftUI agent skill aims to fix AI-generated code</strong>: A new open-source SwiftUI Agent Skill has been released to help AI coding assistants build and refactor SwiftUI views. It includes guidelines to prevent common pitfalls like deprecated <code>onChange</code> modifiers and performance issues, aiming to embed SwiftUI expertise into AI agents upfront and reduce technical debt. This tool could significantly improve the quality of AI-generated SwiftUI code. <a href="https://app.daily.dev/posts/LlEDXwb6h" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Cloud run gets nvidia rtx 6000 gpus for big AI models</strong>: Google Cloud Run now supports NVIDIA RTX PRO 6000 Blackwell GPUs, enabling serverless deployment of large AI models (up to 70B+ parameters) without cluster management. <a href="https://app.daily.dev/posts/Vh4SzEb3w" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New open-source tool for llm inference on kubernetes</strong>: Kthena, a new open-source sub-project of Volcano, is designed for LLM inference orchestration on Kubernetes, addressing challenges like low GPU utilization and multi-model management. <a href="https://app.daily.dev/posts/HONeqoore" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Aws adds new foundation models to sagemaker jumpstart</strong>: AWS has added DeepSeek OCR, MiniMax M2.1, and Qwen3-VL-8B-Instruct models to SageMaker JumpStart, expanding options for document processing, multilingual coding, and vision-language tasks. <a href="https://app.daily.dev/posts/wEnEYbP9z" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-coded tool for local ai code review</strong>: GitHuman, a new local-first tool for reviewing AI-generated code changes, was entirely written by Claude Code and self-reviewed, shifting the code review checkpoint to the staging area. <a href="https://app.daily.dev/posts/M78k9F0lW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New framework aligns llm judges with human feedback</strong>: MemAlign is a new framework that aligns LLM judges with human feedback using a dual-memory system, learning from just 2-10 natural language examples to achieve competitive quality. <a href="https://app.daily.dev/posts/gmGqWGOak" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Developer integrates claude ai with grav cms</strong>: A developer created an open-source MCP plugin for Grav CMS, enabling Claude AI to directly manage blog content via a JSON-RPC API with security features. <a href="https://app.daily.dev/posts/fBtDP8BjG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Nvidia unveils cosmos policy for advanced robot control</strong>: NVIDIA introduced Cosmos Policy, a robot control system fine-tuned from the Cosmos Predict-2 world foundation model, achieving high success rates on robotics benchmarks. <a href="https://app.daily.dev/posts/kQ1bPWgKA" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Tesla robotaxis crash more than humans, still opaque about it</strong>: Tesla’s robotaxi fleet in Austin recorded 9 crashes over 500,000 miles, resulting in a crash rate approximately 3x worse than human drivers for all crashes and 9x worse for police-reported incidents. This is despite having safety monitors in every vehicle, and Tesla continues to redact all crash narratives in its NHTSA reports, unlike competitors like Waymo who operate truly driverless vehicles with a better safety record. But hey, who needs transparency when you have Elon’s tweets? <a href="https://app.daily.dev/posts/M8WkrOIKM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft walks back windows 11 ai overload</strong>: Microsoft is reportedly scaling back its aggressive AI integration in Windows 11 after significant user backlash, reviewing Copilot integrations in apps like Notepad and Paint, and pausing new Copilot buttons. Windows Recall is also under review, with a potential rename or evolution. Apparently, users didn’t appreciate having AI shoved down their throats, proving that sometimes, even Microsoft has to listen to the peasants. <a href="https://app.daily.dev/posts/mR53bImT0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>South africans are apparently super into ai, google says</strong>: A Google-Ipsos study claims South Africans are adopting AI at rates significantly above global averages, with 70% using AI chatbots for upskilling and 86% finding it helpful for learning. They also report 90% want to learn more about AI and 75% see workplace value. While 63% trust tech companies with AI oversight, only 58% trust government. It’s almost as if people are desperate for any tool that promises to make their lives easier, regardless of whether it actually delivers. <a href="https://app.daily.dev/posts/UIjeWpTql" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
