<h2>Must know</h2>
<ul>
<li><strong>OpenAI kills Sora to focus on the super app</strong>: In a major strategic pivot, OpenAI is shutting down its Sora video generation tools, including the API and ChatGPT integration, effective immediately. CEO Sam Altman cited compute constraints and a need to eliminate ‘side quests’ to focus on core productivity and a new foundational model codenamed Spud. The company plans to consolidate its desktop tools and Codex into a single ‘super app’ to better serve business users. <a href="https://app.daily.dev/posts/55QkjBkbg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Arm enters the silicon fray with AGI CPU</strong>: Arm has announced its first production silicon, the AGI CPU, specifically architected for agentic AI workloads in data centers. Built on the Neoverse V3 platform with up to 136 cores, the chip supports PCIe Gen 6 and CXL 3.0 while claiming 2x performance per rack over x86 alternatives. Launch partners include Meta, OpenAI, and Cloudflare, with server configurations from Supermicro supporting up to 45,000 cores in a single liquid-cooled 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>Massive supply chain attack hits security tools and LiteLLM</strong>: A coordinated campaign by threat group TeamPCP has compromised the Checkmarx KICS GitHub Action and the popular LiteLLM package on PyPI. According to Wiz Research, the attackers exploited a GitHub Actions misconfiguration in Aqua Security’s Trivy scanner to exfiltrate secrets and inject credential-stealing malware. LiteLLM versions 1.82.7 and 1.82.8 are confirmed infected, potentially impacting 36% of cloud environments that use the library for LLM orchestration. <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>NVIDIA donates GPU DRA driver to CNCF</strong>: The donation moves ownership of the Kubernetes Dynamic Resource Allocation driver to the community, enabling smarter GPU sharing and multi-node NVLink scaling. <a href="https://app.daily.dev/posts/AoVkebmT8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vite 8 adopts Rust based Rolldown bundler</strong>: The new architecture replaces the esbuild and Rollup mix with a single Rust bundler, promising 10 to 30x faster production builds while keeping plugin compatibility. <a href="https://app.daily.dev/posts/QG6z68w6t" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare launches Dynamic Workers for AI agents</strong>: This new V8 isolate-based runtime executes AI-generated code in milliseconds with 100x faster startup than traditional containers. <a href="https://app.daily.dev/posts/eYSqS4YkC" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Node.js releases critical security updates</strong>: Versions 20.20.2 and 22.22.2 address high-severity vulnerabilities including prototype pollution in HTTP headers and TLS SNICallback errors. <a href="https://app.daily.dev/feed-by-ids?id=C8HkH2xnr&amp;id=Wjr34d3yd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kyverno graduates to full CNCF maturity</strong>: The Kubernetes-native policy engine has reached the highest project tier, now supporting Common Expression Language for admission control. <a href="https://app.daily.dev/feed-by-ids?id=2t2tcYM88&amp;id=leqEkFz9G" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LocalStack v3 adds TUI and Azure plans</strong>: The AWS emulation tool now features a revamped ‘lstk’ CLI with a terminal interface and confirmed development of Azure service emulation. <a href="https://app.daily.dev/posts/dSZeNsqCa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Safari 26.4 introduces masonry grid and WebTransport</strong>: The update brings native CSS Grid Lanes for masonry layouts and the WebTransport API for low-latency bidirectional HTTP/3 communication. <a href="https://app.daily.dev/posts/vXvtsrMcB" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The data center construction mirage</strong>: While NVIDIA prints money, an investigative critique reveals that only 33% of announced US data center capacity is actually under development. Hyperscalers are apparently stockpiling GPUs years before they have the power or floor space to plug them in, creating a massive capital misallocation bubble that makes the dot-com era look disciplined. <a href="https://app.daily.dev/posts/X3my4Sf8s" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Recursive AI code review is an Ouroboros</strong>: Brendan Burns and others are now suggesting we’ll eventually stop reviewing AI-generated code entirely, treating it like compiler output. This ignores the ‘Ouroboros Machine’ effect where AI-generated code reviewed by AI creates a dependency loop of compounding defects, which already results in 1.7x more bugs than human-written code. <a href="https://app.daily.dev/feed-by-ids?id=6DCcfVtqG&amp;id=eYsXfeXJi" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The FCC’s router ban is a manufacturing fantasy</strong>: The US FCC just banned new foreign-made consumer routers for ‘national security,’ seemingly forgetting that almost every router on earth—including those from Cisco and Netgear—is made abroad. Unless you’re buying a Starlink router made in Texas, this is essentially a market-stalling move that will just drive up prices while doing nothing for security. <a href="https://app.daily.dev/feed-by-ids?id=Ll2NLdArm&amp;id=N8TfQEV2B&amp;id=p4tK0Js3X" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
