<h2>Must know</h2>
<ul>
<li><strong>NVIDIA and CNCF standardize GPU resource allocation</strong>: At KubeCon Europe 2026, NVIDIA donated its GPU Dynamic Resource Allocation (DRA) Driver to the CNCF, moving away from the legacy integer-based resource model. This shift enables Kubernetes workloads to request specific GPU topologies, NVLink-connected slices, and multi-node memory domains via ResourceClaims. The move, backed by AWS and Google Cloud, aims to provide a vendor-neutral standard for smarter GPU sharing and hardware reconfiguration. <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>Arm launches AGI CPU for agentic data centers</strong>: Arm has unveiled its first production silicon, the AGI CPU, featuring up to 136 Neoverse V3 cores on an Armv9.2 architecture. Designed specifically for agentic AI workloads, the chip supports DDR5-8800 memory with up to 6TB capacity and 96 lanes of PCIe Gen 6. Partners like Supermicro and Lenovo are already showcasing liquid-cooled designs capable of packing over 45,000 cores into a single 200kW rack. <a href="https://app.daily.dev/feed-by-ids?id=S5lBW0zAU&amp;id=7HHpd3WhV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Open-source supply chain hit by coordinated TeamPCP attack</strong>: A major security breach has compromised Checkmarx’s KICS, multiple VS Code plugins, and the popular LiteLLM package on PyPI. The attackers, identified as TeamPCP, exploited a GitHub Actions misconfiguration in Aqua Security’s Trivy scanner to exfiltrate privileged tokens and inject infostealer malware. Wiz Research reports that LiteLLM is present in 36% of cloud environments, putting SSH keys and cloud credentials at significant risk. <a href="https://app.daily.dev/feed-by-ids?id=FpC0RIGlf&amp;id=GU6PZXcza&amp;id=I6Q0A8ukt&amp;id=rQV4oxnV0" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Anthropic adds auto mode to Claude Code</strong>: A new classifier-based permissions layer allows Claude to autonomously execute safe tool calls while escalating destructive actions to human approval. <a href="https://app.daily.dev/posts/Hgz3is1rZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Node.js releases critical security updates</strong>: Versions 20.20.2 and 24.14.1 address multiple CVEs, including high-severity fixes for HTTP/2 flow control and Web Cryptography timing attacks. <a href="https://app.daily.dev/feed-by-ids?id=C8HkH2xnr&amp;id=EJVcsmjRO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare Dynamic Workers bypass containers for AI</strong>: A new V8 isolate-based runtime allows executing AI-generated TypeScript in milliseconds with 100x faster startup than traditional VMs. <a href="https://app.daily.dev/posts/eYSqS4YkC" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Tinybird implements in-place ALTER TABLE</strong>: The deployment engine now uses native ClickHouse ALTER statements for schema changes, reducing 200GB table migrations from minutes to under 5 seconds. <a href="https://app.daily.dev/posts/iQGOf5SQT" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kyverno graduates to top-level CNCF status</strong>: The Kubernetes-native policy engine has reached full maturity, now handling 20k admission requests per minute at scale for users like LinkedIn. <a href="https://app.daily.dev/posts/leqEkFz9G" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LocalStack v3 revamps CLI and adds TUI</strong>: The AWS emulation tool consolidated its editions into a single binary and introduced a new terminal UI for easier resource management. <a href="https://app.daily.dev/posts/dSZeNsqCa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Netflix Live Origin handles 65 million concurrents</strong>: Engineering details reveal a custom Nginx-based ‘Live Origin’ server using millisecond-grain caching and redundant regional pipelines for massive scale. <a href="https://app.daily.dev/posts/yovKQxg1J" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The data center construction mirage</strong>: While every hyperscaler is screaming about AI capacity, an investigative critique reveals that only 33% of announced US data center projects are actually under development. NVIDIA is reportedly selling GPUs years ahead of when the power grids can even support them, creating a massive gap between ‘sales’ and actual operational compute. It’s a capital misallocation bubble where the hardware is arriving long before the buildings have roofs. <a href="https://app.daily.dev/posts/X3my4Sf8s" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI kills Sora to focus on the super app</strong>: After months of hype and Hollywood demos, OpenAI is shuttering Sora because it’s apparently a ‘side quest’ that eats too much compute. They’re pivoting to a ‘super app’ codenamed Spud, proving once again that shipping a viral video is much easier than maintaining a sustainable API. If you spent the last year building a workflow around Sora, congratulations, you’ve been ‘pivoted’ out of existence. <a href="https://app.daily.dev/posts/55QkjBkbg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The 10x productivity lie</strong>: A longitudinal study from DX found that AI tools are delivering a measly 10% productivity boost, a far cry from the 10x ‘god-mode’ promised by VCs. Meanwhile, PR review times have jumped 91% because humans are now drowning in a sea of AI-generated ‘slop’ that looks correct but fails at the architectural boundaries. We’ve successfully automated the easy part of coding and made the hard part—verification—twice as slow. <a href="https://app.daily.dev/feed-by-ids?id=zpUTGTPYX&amp;id=RvQ2iIf5k&amp;id=665kbWaCc" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
