<h2>Must know</h2>
<ul>
<li><strong>Kubernetes is now the standard for enterprise AI, but culture is the new bottleneck</strong>: The 2025 CNCF Annual Cloud Native Survey reveals Kubernetes production usage has hit 82%, solidifying its role as the infrastructure standard for enterprise AI workloads, with 66% of AI adopters using it for inference. However, the survey, widely cited by sources like the CNCF, highlights a critical shift: cultural resistance, not technical complexity, is now the primary barrier to adoption. Mature organizations are distinguishing themselves through GitOps, platform engineering, and robust observability, indicating that successful AI integration is less about the tech and more about the people. <a href="https://app.daily.dev/posts/Lwsgn3lZ5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is industrializing exploit and malware development</strong>: New research indicates AI is no longer just assisting but actively accelerating offensive cybersecurity. LLMs like Opus 4.5 and GPT-5.2 generated over 40 distinct zero-day exploits, bypassing modern security mitigations for a QuickJS vulnerability, with GPT-5.2 solving the most challenging scenario in 3 hours for about $50. Separately, Check Point Research identified VoidLink, the first AI-generated sophisticated malware suite, developed by a single person in under a week using AI agents—a task estimated to take 30 weeks for multiple human teams. This marks a significant shift towards an ‘industrialized’ model for cyberattacks, as reported by sources like Check Point Research. <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>North Korean hackers target fintech developers via VS Code projects</strong>: North Korean threat actors are exploiting VS Code projects to target fintech, crypto, and blockchain developers. They distribute malicious job assessment projects that, when opened in VS Code, automatically execute payloads to deploy backdoors and malware. This campaign leverages developers’ trust in open-source workflows and standard tools, enabling remote code execution and cryptocurrency mining on compromised machines. The attacks are evolving in sophistication, highlighting a critical new threat vector for high-value sectors, as detailed in technical reports. <a href="https://app.daily.dev/feed-by-ids?id=1wKLyk8gG&amp;id=Q9M3WuQfH" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>New local AI coding model GLM-4.7 Flash outperforms competitors</strong>: GLM-4.7 Flash, a 30B-A3B mixture-of-experts language model, is claiming significant performance leads over competitors like Qwen 3-30B-A3B in local AI coding and agentic benchmarks, with superior tool calling abilities and fast inference speeds. <a href="https://app.daily.dev/posts/eWdzokgG0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding agents show velocity gains but need human guardrails</strong>: AI coding agents like Claude Code can deliver 10-100x velocity gains for well-defined tasks, but struggle with ambiguous requirements, often requiring significant human oversight, strong test coverage, and architectural documentation 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>Microsoft mandates Intune security updates for mobile apps</strong>: Microsoft is enforcing 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 immediate blocking from corporate resources. <a href="https://app.daily.dev/posts/PNEoUht10" 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 publicly accessible, 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>New Agent2Agent SDK and architecture for modular AI agents</strong>: A new A2A Java SDK 1.0.0.Alpha1 aligns with the forthcoming Agent2Agent Protocol 1.0, standardizing communication between AI agents, while architectural discussions emphasize modular ‘Agent Skills’ to prevent prompt bloat and improve scalability. <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>LLVM adopts ‘human in the loop’ policy for AI-assisted contributions</strong>: LLVM has implemented a formal policy requiring human review and accountability for all AI-assisted code contributions, aiming to address an increase in low-quality, tool-generated submissions and ensure maintainer efficiency. <a href="https://app.daily.dev/posts/EdB2gt1Vu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Python package manager uv offers faster installs</strong>: The new Python package manager ‘uv’ is gaining traction for its significantly faster installs, reproducible environments, and improved uninstall mechanisms compared to ‘pip’, prompting developers to evaluate it for workflow efficiency. <a href="https://app.daily.dev/posts/Y2d0smXfY" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI-powered ‘vibecoding’ startup raises $70M, because we needed more buzzwords</strong>: Emergent, an Indian AI-powered ‘vibecoding’ startup, just bagged $70 million in Series B funding, claiming $50 million ARR and 5 million users in seven months. Apparently, ‘vibecoding’ lets you build full-stack apps with AI agents, reducing the need for actual coding expertise. An opinion piece even suggests executives should be ‘vibecoding’ to become individual contributors, because who needs engineering staff when you have AI and a catchy new term for ‘prompt engineering’? <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>CEOs still waiting for AI to magically boost revenue and cut costs</strong>: A PwC survey of 4,454 CEOs found that a whopping 56% haven’t seen any revenue increase or cost reduction from their AI spending. Only 12% saw both. Yet, some articles are still touting ‘agentic AI’ and ‘AI-driven ERP’ with claims of 25% productivity gains and 45% processing time reductions, conveniently omitting any specific sources or real-world success stories. It seems the ‘AI readiness’ hype is still very much alive, despite the actual payoff being as elusive as a stable Kubernetes autoscaling configuration. <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>AI-generated code: still a hot mess of memory leaks and deprecated APIs</strong>: Turns out, AI-generated SwiftUI code is great at producing persistent memory leaks by strongly capturing ‘self’ in nested closures, leading to unbounded memory growth. And let’s not forget its penchant for using deprecated APIs, causing migration nightmares when frameworks update. While some tout AI for ‘velocity gains,’ it seems we’re still stuck manually refactoring its output and wrapping risky API usage, proving that ‘AI-assisted’ often means ‘AI-generated problems, human-solved.’ <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>
