<h2>Must know</h2>
<ul>
<li><strong>AI coding agents are fast but still need human babysitters</strong>: AI coding agents like Claude Code can deliver significant velocity gains (10-100x) for well-defined tasks, but they struggle with ambiguous requirements, often repeating code and omitting tests, according to developer accounts. While Cursor’s AI agents built a web browser in Rust with over a million lines of code in a week, the output was ‘mostly correct’ but not production-ready. GitHub Copilot users see up to 55% faster coding but also 322% more vulnerabilities and lower delivery stability, as reported by research. The consensus is that strong guardrails like automated quality checks, mandatory peer reviews, and robust test coverage are essential to prevent technical debt and quality regressions, meaning humans are still very much in the loop. <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>Kubernetes is now the standard for enterprise AI, but culture is the new bottleneck</strong>: Kubernetes production usage has hit 82% among surveyed organizations, establishing it as the go-to infrastructure for enterprise workloads, especially AI, with 66% of AI adopters using it for inference. The 2025 CNCF Annual Cloud Native Survey, widely cited, reveals that cultural resistance, not technical complexity, is now the primary barrier to adoption. Mature organizations are distinguishing themselves with GitOps and platform engineering, while sustainability concerns are growing for open-source infrastructure due to increasing AI workloads. <a href="https://app.daily.dev/posts/Lwsgn3lZ5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>North Korean hackers are targeting fintech developers via VS Code projects</strong>: North Korean threat actors are exploiting VS Code projects to target developers in cryptocurrency, blockchain, and fintech. 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 significant shift in the threat landscape for high-value sectors. <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>AI is industrializing exploit and malware development</strong>: LLMs like Opus 4.5 and GPT-5.2 generated over 40 exploits for a zero-day QuickJS vulnerability, bypassing modern security mitigations, with GPT-5.2 solving the most challenging scenario in 3 hours for about $50. Check Point Research also found VoidLink, the first 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 multiple human teams. <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>LLVM and Oh My Zsh adopt ‘human in the loop’ for AI contributions</strong>: LLVM and Oh My Zsh maintainers are implementing policies requiring human oversight for AI-assisted code contributions, mandating that contributors review, understand, and take accountability for all submitted code, regardless of the tools used, to combat the rise of low-quality AI-generated submissions and manage maintainer workload. <a href="https://app.daily.dev/feed-by-ids?id=EdB2gt1Vu&amp;id=tr2l7aEc5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New EU decentralized vulnerability database launched</strong>: The EU has launched <a href="http://db.gcve.eu" target="_blank" rel="noopener nofollow">db.gcve.eu</a>, a public cybersecurity vulnerability database designed to reduce reliance on US systems like CVE, featuring a decentralized approach that integrates data from over 25 sources and allows autonomous assignment of vulnerability identifiers. <a href="https://app.daily.dev/posts/i2uCJ3RGE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft updates React Native for Windows, still lagging mobile</strong>: Microsoft released React Native Windows version 0.81 with Hermes debugger support and Fabric architecture, but it still lags behind mobile versions (0.83) and has limited third-party package support, raising questions about Microsoft’s dual investment in React Native and MAUI. <a href="https://app.daily.dev/feed-by-ids?id=7jYV4dEG1&amp;id=qcm6RjfK7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Python package manager ‘uv’ challenges pip</strong>: The new Python package manager ‘uv’ offers faster installs, reproducible environments, and better uninstall mechanisms compared to the standard ‘pip’, prompting developers to evaluate based on project needs and workflow efficiency. <a href="https://app.daily.dev/posts/Y2d0smXfY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux graphics and desktop environments get updates</strong>: Mesa 26.0 now supports Qualcomm Adreno Gen 8 via a Vulkan driver, enabling Linux compatibility for new Snapdragon X2 laptop SoCs, while System76 continues updating its Rust-based COSMIC desktop, and MX Linux 25.1 introduces user-selectable dual init systems (systemd or SysVinit). <a href="https://app.daily.dev/feed-by-ids?id=6eKQwpvId&amp;id=B79z7fEbH&amp;id=hEaaVnp21" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Agent2Agent SDK and AI agent architecture advances</strong>: The A2A Java SDK 1.0.0.Alpha1 has been released, aligning with the forthcoming Agent2Agent Protocol 1.0 for standardized AI agent communication, featuring a migration to Java records, Protocol Buffers for data interchange, and a redesigned AgentCard for capability advertisement. <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>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI-powered code reviewers: The next big thing, or just more busywork?</strong>: Apparently, building an AI-powered code reviewer in 80 hours is a thing now, but the real takeaway is that the AI part was only 10% of the effort. The other 90%? Dealing with API rate limits, latency, and the AI’s tendency to hallucinate. So, while we’re told AI will revolutionize code reviews, it seems we’re mostly just shifting the complexity from writing code to babysitting the AI that writes code. Progress! <a href="https://app.daily.dev/posts/rTTGdX9v4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vibecoding: Because ‘coding’ sounds too much like ‘work’</strong>: Emergent, an Indian AI startup, just bagged $70 million for ‘vibe-coding,’ a platform that lets users build full-stack apps with AI agents, supposedly reducing the need for actual coding expertise. An opinion piece even suggests executives should be ‘vibecoding’ to build internal tools without bothering engineers. So, if you thought coding required skill and effort, think again—just catch a vibe, and AI will handle the rest. What could possibly go wrong with a generation of ‘vibecoders’ who’ve never debugged a segmentation fault? <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 hype cycle: Still going strong, still full of hot air</strong>: Another day, another AI model release, another round of exaggerated claims about revolutionary capabilities and job displacement. Apparently, this cycle repeats every two weeks, with initial euphoria quickly replaced by disappointment as limitations emerge. Social media is, of course, fueling this fire, pushing unrealistic predictions about software engineering becoming obsolete. So, if you’re an engineer, don’t worry, your job is safe—at least until the next AI model comes out and we have to go through this charade all over again. <a href="https://app.daily.dev/feed-by-ids?id=j0xF0WiVr&amp;id=yCkgCoa6x" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
