<h2>Must know</h2>
<ul>
<li><strong>Kubernetes is now the standard for enterprise AI</strong>: The 2025 CNCF Annual Cloud Native Survey reveals Kubernetes production usage has hit 82%, making it the standard infrastructure for enterprise workloads, especially AI. A significant 66% of AI adopters use Kubernetes for inference. Interestingly, the survey, widely cited by sources like the CNCF, indicates that cultural resistance, not technical complexity, is now the primary barrier to adoption. Mature organizations are leveraging GitOps and platform engineering to scale their Kubernetes deployments. <a href="https://app.daily.dev/posts/Lwsgn3lZ5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-driven exploit and malware development is accelerating</strong>: Large language models like Opus 4.5 and GPT-5.2 are now generating sophisticated exploits for zero-day vulnerabilities, bypassing modern security mitigations such as ASLR and CFI. According to experimenters, GPT-5.2 solved a challenging scenario in just three hours for about $50. Check Point Research also identified VoidLink, the first known AI-generated sophisticated malware suite, developed by a single person using AI agents in under a week—a task estimated to take 30 weeks for human teams. Both studies, published on January 20, 2026, confirm AI is industrializing offensive cybersecurity. <a href="https://app.daily.dev/feed-by-ids?id=WkT9zuzY1&amp;id=nLTC7GTyu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft mandates Intune security updates for mobile apps</strong>: Starting January 19, 2026, Microsoft will enforce mandatory Intune Mobile Application Management security updates. Enterprises must update iOS and Android apps to the latest Intune SDK versions (20.8.0+ for iOS, 5.0.6726.0+ for Android Company Portal) or risk having their apps blocked from accessing corporate resources. This update introduces new security controls like PIN requirements and biometric authentication, and non-compliant apps will face immediate outages, impacting critical applications like Outlook and Teams. <a href="https://app.daily.dev/posts/PNEoUht10" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Next.js 16 boosts performance and AI debugging</strong>: Next.js 16 stabilizes Turbopack for faster builds, introduces explicit ‘use cache’ directive for caching, and integrates AI-assisted debugging via Model Context Protocol (MCP) in DevTools, allowing AI agents to suggest fixes. <a href="https://app.daily.dev/posts/91xiEmcVe" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic vulnerabilities exposed remote code execution risks</strong>: Anthropic’s Git MCP server had three critical vulnerabilities (CVE-2025-68145, -68144, -68143) that allowed prompt injection and potential remote code execution, 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>AI coding agents show velocity gains but need guardrails</strong>: AI coding agents like Claude Code can deliver 10–100x velocity gains for well-defined tasks, but struggle with ambiguous requirements, requiring human oversight, strong test coverage, and architectural documentation to prevent technical debt. <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>New Agent2Agent SDK standardizes AI agent communication</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 redesigned AgentCards. <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>Google Gemini prompt injection raises enterprise AI security concerns</strong>: Security researchers found a vulnerability in Google Gemini where malicious calendar invitations could exploit the model’s automatic data ingestion, leading to prompt injection attacks and potential manipulation of AI behavior in enterprise settings. <a href="https://app.daily.dev/posts/dHjNBMkFO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>EU launches decentralized vulnerability database</strong>: 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 and provide autonomous vulnerability identifier assignment. <a href="https://app.daily.dev/posts/i2uCJ3RGE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Taiwan and US sign AI and semiconductor supply chain pact</strong>: Taiwan and the United States finalized a strategic trade agreement focused on semiconductors and AI supply chains, with Taiwan committing $250 billion in investments and credit guarantees for US production, and Taiwanese chip makers like TSMC receiving lowered tariffs. <a href="https://app.daily.dev/posts/suaVSewS2" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI investment: all hype, no ROI?</strong>: PwC surveyed 4,454 CEOs, and a whopping 56% reported no revenue increase or cost decrease from their AI spending. Only 12% saw both. Meanwhile, some unnamed ‘studies’ are still claiming AI-driven ERP leads to 25% productivity gains and 45% processing time reduction. Guess those CEOs just aren’t ‘vibecoding’ hard enough, or maybe, just maybe, the hype machine is still running on fumes. <a href="https://app.daily.dev/feed-by-ids?id=Kov24DV15&amp;id=ogb2Wd6xs&amp;id=rEdiBiFbV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vibecoding: the new executive superpower, or just more AI snake oil?</strong>: Emergent, an Indian AI ‘vibe-coding’ startup, just bagged $70 million, claiming 5 million users and $50 million ARR. An opinion piece then suggests every exec should be ‘vibecoding’ to build apps without engineers. So, now executives can just ‘vibe’ their way to full-stack web apps? Sounds like a great way to generate a lot of ‘vibes’ and very little production-ready code. <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-generated SwiftUI code: because who needs best practices?</strong>: Apparently, AI-generated SwiftUI code is causing ‘persistent memory leaks’ by ignoring basic closure ownership rules. Engineers then have to ‘refactor code by adding weak references’ to clean up the AI’s mess. So, we’re paying AI to write code that we then have to pay engineers to fix. Peak efficiency, folks. Peak efficiency. <a href="https://app.daily.dev/feed-by-ids?id=0Dxh8ahUn&amp;id=NOlmy6yd2&amp;id=yTmR3XbQ9" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
