<h2>Must know</h2>
<ul>
<li><strong>Anthropic adds auto mode to Claude Code</strong>: Anthropic has introduced a new permissions system for Claude Code that uses a classifier to automatically approve safe tool calls while blocking destructive actions like mass file deletion. This research preview for Team plan users aims to balance developer velocity with security by escalating to human prompts only when the model repeatedly hits safety blocks. The feature is compatible with the latest Claude 4.6 Sonnet and Opus models, according to Anthropic’s technical release. <a href="https://app.daily.dev/posts/Hgz3is1rZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA open sources GPU orchestration for Kubernetes</strong>: At KubeCon Europe, NVIDIA donated its Dynamic Resource Allocation (DRA) Driver to the CNCF, transferring ownership of critical GPU sharing and multi-node NVLink interconnect logic to the community. This move, backed by AWS and Google Cloud, enables smarter hardware reconfiguration and confidential computing via Kata Containers. NVIDIA also launched Grove, a new Kubernetes API specifically for orchestrating large-scale AI workloads, as reported by CNCF and NVIDIA. <a href="https://app.daily.dev/posts/AoVkebmT8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Supply chain attack hits LiteLLM via GitHub Actions</strong>: A misconfigured CI/CD pipeline in the Trivy project allowed attackers from the TeamPCP group to exfiltrate PyPI publishing tokens and inject malware into LiteLLM versions v1.82.7 and v1.82.8. The breach highlights a growing trend of targeting open-source security tools to gain enterprise credentials, leading to the immediate removal of the affected packages from PyPI. Security researchers at RSAC 2026 noted this as a primary example of the risks inherent in automated agentic supply chains. <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 inference</strong>: The new sandbox project from IBM and Google Cloud enables KV-cache-aware routing and independent scaling of prefill and decode phases across heterogeneous accelerators. <a href="https://app.daily.dev/posts/tdsbXMmJT" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Node.js 20.20.2 fixes seven security vulnerabilities</strong>: The latest LTS release addresses critical CVEs including timing-safe comparison flaws in Web Crypto and prototype pollution in HTTP headers. <a href="https://app.daily.dev/posts/C8HkH2xnr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft uses Uno Platform to unify WinUI development</strong>: The Windows Community Toolkit team successfully migrated to a single codebase supporting WinUI 2, WinUI 3, and WebAssembly via Uno’s cross-platform abstraction. <a href="https://app.daily.dev/posts/Cmh5tPdzd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Snyk launches security agents for the MCP ecosystem</strong>: The new platform includes automated agents to map AI Bill of Materials and enforce guardrails within Claude Code, Cursor, and Devin pipelines. <a href="https://app.daily.dev/posts/iRGLDmExe" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Epic Games cuts 20 percent of workforce</strong>: CEO Tim Sweeney announced layoffs of over 1,000 employees to pivot focus toward Unreal Engine 6 and Fortnite seasonal content. <a href="https://app.daily.dev/posts/G1SUSytCF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Capcom bans generative AI assets in game production</strong>: Following backlash over AI-altered models in a tech demo, the studio committed to using AI only for internal development efficiency rather than final player-facing assets. <a href="https://app.daily.dev/posts/enr9iPrED" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>SlowReact offers hands-on performance debugging</strong>: A new open-source repository provides intentionally broken React 19 components to help developers master the Profiler and identify costly re-renders. <a href="https://app.daily.dev/posts/y44kwac6A" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The fight AI with AI marketing trap</strong>: Security vendors at RSAC are tripping over themselves to sell ‘autonomous defense agents,’ but ThreatLocker’s Danny Jenkins correctly calls the narrative ‘pure garbage.’ No amount of generative magic can magically determine software intent; you’re better off with boring, tangible controls like blocking untrusted binaries than a hallucinating chatbot playing digital security guard. <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>The text to SQL scandal is coming</strong>: MotherDuck predicts a massive ‘text-to-SQL scandal’ in 2026 as enterprises realize their LLM-generated queries are riddled with subtle logic errors. While vendors promise humans will ‘write very little SQL,’ the reality is that data engineers will just spend their time debugging increasingly complex, agent-generated spaghetti code that lacks proper context. <a href="https://app.daily.dev/posts/X1JY9WQX8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI productivity is a statistical ghost</strong>: Despite the hype around AI coding agents, empirical analysis of PyPI data shows no broad-based explosion in developer output. The only measurable ‘productivity’ increase is in the release frequency of AI-related packages, which is more likely a result of VC funding and desperate attention-seeking than actual engineering efficiency. <a href="https://app.daily.dev/feed-by-ids?id=zpUTGTPYX&amp;id=RvQ2iIf5k" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
