<h2>Must know</h2>
<ul>
<li><strong>NVIDIA open sources GPU drivers and AI orchestration for CNCF</strong>: At KubeCon Europe, NVIDIA donated its Dynamic Resource Allocation (DRA) Driver to the CNCF, effectively handing over GPU sharing and multi-node NVLink interconnect management to the Kubernetes community. This move, backed by AWS and Google Cloud, includes new open-source projects like Grove for AI workload orchestration and NVSentinel for GPU fault remediation. It signals a shift toward making specialized AI hardware a first-class, standardized citizen in cloud-native infrastructure. <a href="https://app.daily.dev/posts/AoVkebmT8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic adds auto mode to Claude Code for autonomous execution</strong>: Anthropic introduced a new permissions system for its CLI tool that uses a classifier to automatically approve safe tool calls while blocking destructive actions like mass deletions. This research preview for Team plan users aims to reduce human-in-the-loop friction during multi-file refactors without resorting to dangerous ‘skip permissions’ flags. If the model hits repeated blocks, it escalates to a human prompt, attempting to balance agentic autonomy with safety. <a href="https://app.daily.dev/posts/Hgz3is1rZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Supply chain attack hits LiteLLM via compromised GitHub Actions</strong>: A misconfigured CI/CD pipeline in the Trivy project allowed attackers to exfiltrate a PyPI publishing token, leading to the injection of credential-stealing malware into LiteLLM versions v1.82.7 and v1.82.8. The threat group, dubbed TeamPCP, is systematically targeting open-source security tools and developer infrastructure to harvest enterprise credentials. This incident highlights the extreme fragility of the AI toolchain, where a single leaked token can poison a dependency used by thousands of agentic workflows. <a href="https://app.daily.dev/feed-by-ids?id=bf7GKaXaH&amp;id=xiExzcA7d" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>CNCF adopts llm-d for distributed Kubernetes inference</strong>: The framework introduces KV-cache-aware routing and prefill/decode disaggregation to scale LLM workloads across heterogeneous accelerators. <a href="https://app.daily.dev/posts/tdsbXMmJT" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Safari 26.4 introduces CSS Grid Lanes for masonry layouts</strong>: The update also ships WebTransport for low-latency HTTP/3 communication and the Keyboard Lock API for fullscreen web apps. <a href="https://app.daily.dev/posts/vXvtsrMcB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>DSPy adoption remains low despite solving core AI engineering pains</strong>: Teams continue to reinvent DSPy’s typed I/O and prompt management patterns from scratch, leading to unmaintainable ad hoc complexity. <a href="https://app.daily.dev/posts/KkRBBl10a" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Epic Games lays off 20 percent of workforce to fund Unreal Engine 6</strong>: CEO Tim Sweeney cited declining Fortnite engagement and legal costs as drivers for cutting over 1,000 employees. <a href="https://app.daily.dev/posts/G1SUSytCF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Snyk launches security agents to map AI Bill of Materials</strong>: The new platform includes automated agents to discover MCP servers and enforce governance guardrails in Claude Code and Cursor. <a href="https://app.daily.dev/posts/iRGLDmExe" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloud native community swells to 20 million developers</strong>: A new report finds 39% of all developers are now cloud native, with 7.3 million specifically focused on AI workloads. <a href="https://app.daily.dev/posts/6gZyGZFiz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Slug GPU text rendering algorithm enters the public domain</strong>: After a 10-year retrospective, the patent for the high-performance algorithm has been dedicated to the public with MIT-licensed shaders. <a href="https://app.daily.dev/posts/EUFNfYm8U" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The fight AI with AI security narrative</strong>: ThreatLocker’s CEO recently called the ‘fight AI with AI’ marketing pitch ‘pure garbage,’ noting that AI fundamentally cannot determine software intent. While every RSAC vendor is peddling ‘autonomous SOC agents,’ the reality is that we’re just adding more non-deterministic layers to already broken security models. It’s a great way to burn VC cash while ignoring basic controls like blocking untrusted binaries. <a href="https://app.daily.dev/feed-by-ids?id=f6diiN9F1&amp;id=k3pmCJTXy" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Generative AI for Beginners on .NET 10</strong>: Microsoft is already pushing a ‘Generative AI for Beginners’ course rebuilt for .NET 10, because apparently, we’ve mastered the fundamentals of a two-year-old tech so well we need a version for a framework that isn’t even out yet. It’s the classic enterprise move: wrap a fast-moving research field in heavy abstractions like <a href="http://Microsoft.Extensions.AI" target="_blank" rel="noopener nofollow">Microsoft.Extensions.AI</a> before anyone actually knows what they’re building. If you need a dependency injection pattern for your prompt engineering, you might have bigger problems. <a href="https://app.daily.dev/posts/4bgn03WFr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The text to SQL scandal of 2026</strong>: Industry ‘future casters’ are predicting a massive text-to-SQL scandal where LLMs hallucinate data queries, yet they claim it won’t slow adoption because humans will ‘write very little SQL’ anyway. We are effectively cheering for the death of the only verifiable interface to our data in favor of ‘agent-driven architectures’ that we already know will fail. It’s a bold strategy to replace precise logic with vibes and then just plan for the inevitable disaster. <a href="https://app.daily.dev/posts/X1JY9WQX8" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
