<h2>Must know</h2>
<ul>
<li><strong>Kubernetes is now the standard for enterprise AI, culture is the new bottleneck</strong>: Kubernetes production usage has hit 82% among organizations, solidifying its role as the standard infrastructure for enterprise AI workloads, with 66% of AI adopters using it for inference. The 2025 CNCF Annual Cloud Native Survey, widely cited by The Register, reveals that cultural resistance, not technical complexity, is now the primary barrier to adoption. Mature organizations are leveraging GitOps and platform engineering to scale Kubernetes, highlighting a shift in focus from technical hurdles to organizational dynamics. <a href="https://app.daily.dev/posts/Lwsgn3lZ5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding agents: impressive speed, critical guardrails needed</strong>: AI coding agents like Claude Code and GitHub Copilot are showing significant velocity gains (10-100x) for well-defined tasks, with Cursor even building a web browser in Rust with AI agents in a week. However, these tools struggle with ambiguous requirements, can introduce vulnerabilities, and require extensive human oversight. According to InfoQ, robust guardrails like automated quality checks, mandatory peer reviews, strong test coverage, and up-to-date architectural documentation are essential to prevent technical debt and quality regressions. <a href="https://app.daily.dev/feed-by-ids?id=EqMZfdLoq&amp;id=GPLZOgOgP&amp;id=J24eplQdr&amp;id=cSbwL1BS5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-driven exploit development is industrializing</strong>: Experiments show LLMs like Opus 4.5 and GPT-5.2 can generate over 40 distinct exploits for zero-day vulnerabilities, bypassing modern security mitigations like ASLR and CFI. GPT-5.2 even solved a complex scenario in 3 hours for about $50, indicating exploit generation is now limited by token throughput, not human expertise. Check Point Research also identified VoidLink, the first AI-generated sophisticated malware suite, developed in a fraction of the time a human team would take, signaling a major shift in offensive cybersecurity capabilities. <a href="https://app.daily.dev/feed-by-ids?id=WkT9zuzY1&amp;id=nLTC7GTyu" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>New AI models: GLM-4.7 Flash for local coding</strong>: GLM-4.7 Flash, a 30B-A3B mixture-of-experts model, is designed for local AI coding tasks and reportedly outperforms competitors like Qwen 3-30B-A3B on coding and agentic AI benchmarks. <a href="https://app.daily.dev/posts/eWdzokgG0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic fixes critical Git server vulnerabilities</strong>: Anthropic patched three critical vulnerabilities in its Git MCP server (CVE-2025-68145, CVE-2025-68144, CVE-2025-68143) that could have allowed remote code execution via prompt injection, highlighting security risks in AI-integrated systems. <a href="https://app.daily.dev/posts/U6A8HrCeq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>EU launches decentralized vulnerability database</strong>: The GCVE has launched <a href="http://db.gcve.eu" target="_blank" rel="noopener nofollow">db.gcve.eu</a>, a public, decentralized cybersecurity vulnerability database integrating data from over 25 sources, aiming to reduce European reliance on US-based systems like CVE. <a href="https://app.daily.dev/posts/i2uCJ3RGE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft Intune mandates security updates for mobile apps</strong>: Microsoft will enforce mandatory Intune Mobile Application Management security updates from January 19, 2026, requiring iOS and Android apps to update to the latest Intune SDK versions or face being blocked from accessing corporate resources. <a href="https://app.daily.dev/posts/PNEoUht10" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent communication standardizing with A2A Java SDK</strong>: The A2A Java SDK 1.0.0.Alpha1 has been released, aligning with the forthcoming Agent2Agent Protocol 1.0 specification to standardize communication between AI agents, featuring Java records, Protocol Buffers, and enhanced API documentation. <a href="https://app.daily.dev/feed-by-ids?id=LmGT7Q5vx&amp;id=PMa93QJZs&amp;id=ZcWzztsx0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Next.js 16 brings Turbo Builds and AI debugging</strong>: Next.js 16 stabilizes Turbopack as the default bundler for faster builds, introduces explicit Cache Components, and integrates AI-assisted debugging via Model Context Protocol (MCP) in DevTools for unified logs and suggested fixes. <a href="https://app.daily.dev/feed-by-ids?id=91xiEmcVe&amp;id=GHFDClpAg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>One Identity Manager 10.0 enhances security and automation</strong>: One Identity released version 10.0 of Identity Manager, introducing risk-based governance, identity threat detection and response (ITDR) capabilities, and AI-assisted reporting via natural language querying for large-scale enterprise identity management. <a href="https://app.daily.dev/posts/xsUICDLZg" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI will make you a 100x developer, just don’t ask it to write good code</strong>: Apparently, AI coding tools can make you 100 times more productive, but only if you give them ‘precise specifications’ and ‘structured boundaries.’ So, it’s not magic, it’s just a very expensive, very fast intern that still needs constant supervision and a ‘context architect’ to hold its hand. Who knew the secret to AI success was just better management of your digital minions? <a href="https://app.daily.dev/feed-by-ids?id=EqMZfdLoq&amp;id=GPLZOgOgP&amp;id=J24eplQdr&amp;id=cSbwL1BS5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vibecoding: the new executive superpower, or just more AI hype?</strong>: Emergent, an Indian AI startup, just bagged $70 million for ‘vibecoding,’ a platform that supposedly lets executives build full-stack apps without ‘advanced coding expertise.’ Meanwhile, an opinion piece suggests ‘every exec should be vibecoding’ to become an individual contributor. Because nothing says robust, scalable software like a CEO ‘vibecoding’ a hackathon platform. What could possibly go wrong? <a href="https://app.daily.dev/feed-by-ids?id=AllrxKqHW&amp;id=McILqauKu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI will save your career, by making you a better prompt engineer</strong>: The latest wisdom from the tech gurus is that AI will accelerate your career, but only if you stop focusing on ‘basic coding skills’ and become a ‘context architect’ or ‘prompt engineer.’ So, instead of writing code, you’ll be writing instructions for the AI that writes the code. It’s like we’re evolving from craftsmen to glorified project managers for our silicon overlords. Progress! <a href="https://app.daily.dev/feed-by-ids?id=j0xF0WiVr&amp;id=yCkgCoa6x" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
