<h2>Must know</h2>
<ul>
<li><strong>Google and Moonshot AI push agentic vision with new models</strong>: Google introduced Agentic Vision for Gemini 3 Flash, transforming static image processing into interactive investigation with visual reasoning and code execution, as reported by Google. Moonshot AI also released Kimi K2.5, an open-source visual agentic model for coding and research automation, featuring a Mixture of Experts architecture and an Agent Swarm capability. These developments aim to make visual AI practical for production systems and enhance agent workflows. <a href="https://app.daily.dev/feed-by-ids?id=KIgS0LnaZ&amp;id=2GP3EQ58p" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft updates Quantum Development Kit and launches AI Observability</strong>: Microsoft announced major updates to its Quantum Development Kit (QDK), integrating it with VS Code and GitHub Copilot for AI-assisted quantum programming, and adding specialized toolkits for chemistry and error correction. Simultaneously, Dynatrace launched its AI Observability app, supporting agentic AI frameworks like Amazon Bedrock AgentCore and LangChain, providing end-to-end visibility for agent interactions and unifying telemetry with OpenTelemetry and OpenLLMetry, according to Dynatrace. <a href="https://app.daily.dev/feed-by-ids?id=0i847vEah&amp;id=d1KKTrRCA" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubernetes telemetry exploit allows full cluster compromise</strong>: A security researcher discovered an exploit in Kubernetes’ <code>nodes/proxy GET</code> call that allows privileged command execution and full cluster compromise without audit trails. Despite the severe impact, the Kubernetes team classified this as “intended behavior.” A fix is planned for Kubernetes 1.36, but administrators are advised to audit RBAC policies and restrict <code>kubelet</code> port access in the interim, as detailed in the security researcher’s findings. <a href="https://app.daily.dev/posts/bOiz0Chx6" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Google BigQuery integrates SQL-native AI inference</strong>: Google BigQuery now supports SQL-native managed inference, allowing data teams to deploy and run over 180,000 Hugging Face and Vertex AI models directly via SQL statements, eliminating the need for separate ML infrastructure. <a href="https://app.daily.dev/posts/eM9NtN7mD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>SERA open-source coding agent announced</strong>: SERA, an open-source coding agent, achieves state-of-the-art performance on SWE-Bench Verified at 54.2% and is affordable to train, costing approximately $400 for reproduction. <a href="https://app.daily.dev/posts/y5jjtMeVw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Laravel Boost v2 integrates community AI skills</strong>: Laravel Boost v2 introduces new functionality for integrating community AI skills, allowing users to install AI skills via a <code>boost:add-skill</code> command that automatically syncs across all agents. <a href="https://app.daily.dev/posts/0wx3OGQ9R" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rust Foundation releases 2025 annual report and strategic plan</strong>: The Rust Foundation released its 2025 Annual Report and a 2026-2028 strategic plan focusing on stable infrastructure, sustainable maintainer support, and responsible adoption growth for the Rust Project. <a href="https://app.daily.dev/posts/DRpplNBlF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Klint Rust-based linting tool proposed for Linux kernel</strong>: RFC patches have been posted to integrate Klint, a Rust-based linting tool, into the Linux kernel to detect issues like incorrect prelude imports and excessive stack frame sizes. <a href="https://app.daily.dev/posts/RgFFWztiy" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microservices pitfalls and modular monolith recommendation</strong>: A new article argues that microservices primarily address organizational problems and recommends starting with a well-structured modular monolith, using event storming to identify proper service boundaries. <a href="https://app.daily.dev/posts/aM21BxeGj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Formae expands Terraform challenger to more clouds</strong>: Platform Engineering Labs expanded Formae, an Infrastructure as Code platform using Apple’s Pkl language, to support Google Cloud, Azure, Oracle Cloud, and OVHcloud, offering an alternative to Terraform with improved state management through autonomous agents. <a href="https://app.daily.dev/posts/qaAQ3pbgp" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic CEO argues for minimal AI regulation, conveniently</strong>: Anthropic CEO Dario Amodei penned a 22,000-word essay arguing for minimal AI regulation, focusing on theoretical superintelligent AI risks and advocating for restricting China’s chip access. Critics, however, suggest this is just typical corporate resistance, ignoring present concerns like wealth inequality and job displacement, especially since Anthropic isn’t expected to be profitable until 2028. Funny how the ‘existential risk’ narrative always seems to align with a company’s bottom line. <a href="https://app.daily.dev/posts/amOI7bITg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Enterprise AI agents face ‘trough of disillusionment’</strong>: Despite all the hype, enterprise AI agent implementations are struggling to move from lab to production, with Redis CEO reporting fewer production agents than anticipated. Apparently, aggregating unstructured data and replicating human judgment is harder than just slapping ‘AI’ on it. Enterprises are now in the ‘trough of disillusionment’ regarding AI ROI, but don’t worry, they’re still investing in these ‘next-generation infrastructure’ platforms. Because nothing says progress like throwing more money at a problem that isn’t quite working yet. <a href="https://app.daily.dev/posts/7PYYjzg77" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Moltbot: the viral AI assistant that’s a security nightmare</strong>: Moltbot, the open-source personal AI assistant, went viral with 44,000 GitHub stars, but it’s also a critical security risk. It operates locally, uses Cloudflare, and autonomously manages tasks, but security experts warn of prompt injection attacks, plaintext data storage, and its GitHub account was even hijacked by crypto scammers. So, you get a cool AI assistant that can manage your life, but also potentially give it away to the highest bidder. What a deal! <a href="https://app.daily.dev/posts/zkO1JWwHu" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
