<h2>Must know</h2>
<ul>
<li><strong>AI security risks: exposed models and critical vulnerabilities</strong>: Researchers found over 175,000 Ollama AI hosts exposed online without safety guardrails, enabling potential resource hijacking and remote execution. A critical 1-click RCE vulnerability was discovered in OpenClaw, an AI assistant used by over 100,000 developers, allowing authentication token theft and arbitrary command execution. SpaceX also updated Starlink’s privacy policy to use customer data for AI model training, raising significant privacy concerns, especially with a potential merger with xAI. <a href="https://app.daily.dev/feed-by-ids?id=2SdHHyeRJ&amp;id=YUSdAM3D6&amp;id=d76FYhLFT&amp;id=iEIaVGrhd&amp;id=yXHHUa4yo&amp;id=4nlzVZMEL&amp;id=ALGXJz29U&amp;id=fuL6WfKHt&amp;id=oAoGgXAAn&amp;id=zgJdriYuf" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI’s paradoxical impact on developer productivity</strong>: While AI tools accelerate code generation, they may paradoxically reduce overall productivity by enabling cheap complexity and increasing organizational chaos. The real bottlenecks in software development are integration, security, and operations, not coding speed. Unconstrained AI can lead to fragmented systems that are expensive to maintain. The proposed solution is platform engineering and ‘golden paths’ to constrain AI-generated code within company ecosystems and security guidelines. <a href="https://app.daily.dev/feed-by-ids?id=k9rFokMvG&amp;id=CUMsxmpoK&amp;id=H1aWwdchu&amp;id=KYToiOjQs&amp;id=TCb4dVj23" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloud native ecosystem sees retirements and new releases</strong>: The cloud native ecosystem is undergoing significant changes, with the Ingress NGINX project announcing its retirement and the Kubernetes Dashboard project being archived. On a brighter note, Cluster API v1.12 was released, bringing in-place updates, and a new Checkpoint/Restore Working Group has been formed. These shifts highlight the dynamic nature of cloud infrastructure, requiring constant adaptation. <a href="https://app.daily.dev/posts/EdFbm39PB" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>New NVIDIA NVFP4 quantization for LLMs released by Red Hat</strong>: Red Hat released NVFP4-quantized versions of popular LLMs, leveraging NVIDIA’s 4-bit floating-point format for efficient deployment with near-baseline accuracy, offering 1.5-1.8x smaller storage than FP8 and native hardware acceleration on Blackwell GPUs. <a href="https://app.daily.dev/posts/OPe22PWoI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>MLflow integrates new evaluation metrics for LLM agents and RAG systems</strong>: MLflow 3.9.0 now integrates over 50 evaluation metrics from DeepEval, RAGAS, and Arize Phoenix, providing a unified API to assess LLM agents and RAG systems for conversational quality, retrieval, hallucination, and safety. <a href="https://app.daily.dev/posts/zVcgznCs0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>DataDog releases pg_tracing for PostgreSQL distributed tracing</strong>: DataDog developed <code>pg_tracing</code>, a PostgreSQL extension for server-side distributed tracing that generates spans for database queries and operations, supporting PostgreSQL 14-16 and OpenTelemetry integration. <a href="https://app.daily.dev/posts/Xfr0u4GSj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>WebAssembly celebrates 10 years, expands beyond browser</strong>: WebAssembly (WASM) marked its 10th anniversary, evolving from a browser performance solution to a foundational technology for edge platforms, serverless runtimes, and embedded systems, now showing promise for AI workloads at the edge. <a href="https://app.daily.dev/posts/XjOBMuNCm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New open-source Kubernetes dashboard ‘Radar’ released</strong>: A new open-source Kubernetes dashboard called Radar, built as a single Go binary with an embedded React frontend, provides real-time cluster visibility, Helm management, and traffic visualization without cloud logins or cluster-side agents. <a href="https://app.daily.dev/feed-by-ids?id=DiEunA7dc&amp;id=a3wJg58hN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Wix introduces xEngineer model for AI-driven development</strong>: Wix launched its ‘xEngineer’ model, integrating AI into engineering roles to broaden ownership across the development lifecycle, aiming to evolve responsibilities rather than reduce headcount, particularly in DevOps and SRE. <a href="https://app.daily.dev/posts/1ESLU5xBs" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Minimalist coding agent design principles emerge</strong>: A new coding agent built from scratch prioritizes simplicity with a unified LLM API, minimal toolset (read, write, edit, bash), and a sub-1000 token system prompt, claiming comparable performance to feature-heavy alternatives on Terminal-Bench 2.0. <a href="https://app.daily.dev/posts/myqj9ttHJ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI-washing: blaming AI for layoffs</strong>: Apparently, AI caused over 50,000 layoffs in 2025. But a Forrester report suggests many companies are just ‘AI-washing,’ blaming AI for cuts actually driven by financial pressures or pandemic over-hiring. Turns out, most companies announcing these AI-related layoffs don’t even have mature AI applications ready to replace those roles. So, AI is the new scapegoat for bad business decisions, how novel. <a href="https://app.daily.dev/posts/yXHHUa4yo" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI’s ChatGPT: web search is a feature, not a bug (unless it costs money)</strong>: OpenAI makes it surprisingly difficult to enable web search in ChatGPT, even though the model frequently hallucinates without it. The UX actively discourages search usage, with no default setting or keyboard shortcut. One might suspect this isn’t about accuracy, but rather about saving a few pennies on search queries. Who needs facts when you have a chatbot that confidently makes things up? <a href="https://app.daily.dev/posts/0hvZGEXEr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated video: Darren Aronofsky and TIME prove it’s still mostly garbage</strong>: Darren Aronofsky and TIME, with Google DeepMind’s tech, released an AI-generated video series commemorating the American Revolution. Unsurprisingly, it was met with widespread criticism and negative reception. This just confirms that AI-generated visuals continue to fail at resonating with audiences. Maybe they should try hiring actual artists instead of letting algorithms churn out soulless content. <a href="https://app.daily.dev/posts/vgyq27hp2" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
