<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 deletions. This research preview for Team plan users aims to reduce manual approval friction without resorting to the risky skip-permissions flag. It is currently compatible with the latest Claude 4.6 Sonnet and Opus models. <a href="https://app.daily.dev/posts/Hgz3is1rZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA and partners donate GPU and inference projects to CNCF</strong>: At KubeCon Europe, NVIDIA donated its Dynamic Resource Allocation (DRA) Driver to the CNCF to standardize GPU sharing and multi-node NVLink interconnects within Kubernetes. Simultaneously, IBM, Red Hat, and Google Cloud donated llm-d, a framework for distributed LLM inference that disaggregates prefill and decode phases. These moves signal a major industry shift toward making heterogeneous accelerator orchestration a first-class cloud-native citizen. <a href="https://app.daily.dev/feed-by-ids?id=AoVkebmT8&amp;id=tdsbXMmJT" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Supply chain attack hits LiteLLM via GitHub Actions</strong>: A misconfigured GitHub Actions pipeline in the Trivy project allowed attackers to exfiltrate a PyPI publishing token, leading to the compromise of LiteLLM versions v1.82.7 and v1.82.8. The malicious packages contained credential-stealing malware, highlighting a targeted campaign by the group TeamPCP against popular AI developer infrastructure. Security researchers at RSAC 2026 noted this as a critical example of the growing risks in autonomous agent supply chains. <a href="https://app.daily.dev/feed-by-ids?id=bf7GKaXaH&amp;id=xiExzcA7d&amp;id=tRBxFZkzF" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Snyk launches AI agent security and Evo AI-SPM</strong>: New tools provide automated discovery of AI Bill of Materials and policy guardrails for agents in Claude Code, Cursor, and Devin. <a href="https://app.daily.dev/posts/iRGLDmExe" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>DSPy adoption remains low despite solving core AI engineering patterns</strong>: The framework’s abstractions for typed I/O and composable modules are often reinvented poorly by teams struggling with ad hoc prompt management. <a href="https://app.daily.dev/posts/KkRBBl10a" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>MotherDuck predicts the collapse of the modern data stack</strong>: The shift toward agent-driven architectures is expected to commoditize ETL and BI vendors as LLMs take over SQL generation and data modeling. <a href="https://app.daily.dev/posts/X1JY9WQX8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Epic Games lays off 20 percent of workforce</strong>: CEO Tim Sweeney cited declining Fortnite engagement and high legal costs while pivoting focus toward Unreal Engine 6 development. <a href="https://app.daily.dev/posts/G1SUSytCF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>In-browser AI inference gains traction with WebGPU and WebNN</strong>: Architectural privacy and reduced latency are driving the use of Transformers.js and local models like Gemini Nano directly in Chrome. <a href="https://app.daily.dev/posts/FmvIxbZwO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>FivFold introduces full-stack kits with AI context generation</strong>: The new scaffolding platform includes an <a href="http://AGENTS.md" target="_blank" rel="noopener nofollow">AGENTS.md</a> generator to help coding assistants better understand project-specific backend boilerplate. <a href="https://app.daily.dev/posts/hPqZU29Ud" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloud native developer community hits 20 million</strong>: A new CNCF report indicates that 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>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The fight AI with AI security narrative</strong>: At RSAC, ThreatLocker’s leadership called the ‘fight AI with AI’ marketing ‘pure garbage,’ noting that AI still can’t reliably determine software intent. Vendors are flooding the zone with ‘autonomous’ security agents that often just add unvalidated complexity to already noisy environments. <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>PwC’s autonomous consultant agents</strong>: PwC is launching a platform where AI agents perform consulting work directly for clients, which sounds like a great way to charge premium rates for LLM outputs. They claim ‘professionals review outputs in the background,’ which is consultant-speak for ‘we hope the hallucination isn’t expensive.’ <a href="https://app.daily.dev/posts/9Lxa6AWQ6" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Aviation metaphors for project management</strong>: A new methodology called ‘Flights’ wants you to call your tasks ‘crates’ and your owners ‘captains’ to avoid Agile fatigue. It’s a classic case of rebranding the same old Jira tickets with a fresh coat of ‘aviation metaphor’ paint instead of actually fixing broken workflows. <a href="https://app.daily.dev/posts/PRTXr8G4W" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
