<h2>Must know</h2>
<ul>
<li><strong>Anthropic and Google AI models vulnerable to prompt injection</strong>: Anthropic’s Git MCP server was found to have critical vulnerabilities (CVE-2025-68143, CVE-2025-68144, CVE-2025-68145) allowing prompt injection and remote code execution, as reported by security researchers. Similarly, Google Gemini was vulnerable to manipulation via malicious calendar invitations, where embedded instructions could exfiltrate private meeting data. Both incidents highlight the growing risk of semantic attacks on AI systems, emphasizing the need for robust security beyond traditional software exploits. <a href="https://app.daily.dev/feed-by-ids?id=U6A8HrCeq&amp;id=dHjNBMkFO&amp;id=j6bBmLvCW&amp;id=rKbuKp9sI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubernetes becomes core AI infrastructure, faces new challenges</strong>: Kubernetes is now the standard for AI workloads, with 66% of organizations using it for AI inference, according to the 2025 CNCF Annual Cloud Native Survey. However, adoption barriers have shifted from technical complexity to cultural resistance and sustainability concerns due to increased AI workloads. MetalBear’s new ‘mirrord for CI’ tool aims to streamline Kubernetes testing by allowing concurrent CI tests on shared clusters, reducing setup time and costs for enterprise users. <a href="https://app.daily.dev/feed-by-ids?id=Lwsgn3lZ5&amp;id=s2plqXXTq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical vulnerabilities found in Chainlit AI framework</strong>: Chainlit, a popular open-source Python framework for AI chatbots, contained critical vulnerabilities (CVE-2026-22218, CVE-2026-22219) that allowed arbitrary file reads and server-side request forgery. These flaws could expose API keys, credentials, and cloud tokens, putting sensitive enterprise data at risk. Patched in version 2.9.4, the disclosure emphasizes the security risks of integrating third-party frameworks into enterprise AI systems. <a href="https://app.daily.dev/posts/I9vckVFsM" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI-generated code raises security and quality concerns</strong>: Studies show AI coding assistants increase code generation but also lead to higher static analysis warnings and code complexity, with developers reporting frequent bugs and code duplication in AI-generated code. <a href="https://app.daily.dev/feed-by-ids?id=BqO4RcudK&amp;id=EZUXhYwRZ&amp;id=LiH8HLrgt&amp;id=YH0ajyAf5&amp;id=rlAZMkOYu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Linux kernel patches boost performance and hardware support</strong>: Intel engineers released a Linux kernel patch improving NVMe performance by up to 15% on high core-count systems, while Linux 6.19 includes critical ATA subsystem fixes for power management and Linux 7.0 will add Intel GPU firmware update support for non-x86 architectures. <a href="https://app.daily.dev/feed-by-ids?id=BIhibPcDf&amp;id=CI1vi8cFn&amp;id=LPWAW20mV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks unveils enterprise AI governance framework</strong>: Databricks launched a new AI Governance Framework with 43 considerations across five pillars, aiming to help enterprises deploy AI projects successfully by addressing unclear ownership and compliance gaps. <a href="https://app.daily.dev/posts/DO9dLR2KM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Open-source tools for AI agent development and monitoring emerge</strong>: RubyLLM::Instrumentation is a new gem for monitoring LLM operations in Ruby apps, while Temporal and Braintrust integrated to enhance AI agent observability and reliability, and xaa.dev launched a playground for secure Cross App Access testing. <a href="https://app.daily.dev/feed-by-ids?id=B1cAlIUwy&amp;id=JOl3Mddyk&amp;id=LmGT7Q5vx&amp;id=pQZ55h2AV&amp;id=HeOTx2Sl7&amp;id=vZ2X1Grfn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New tools for web development and database management</strong>: RelateDB offers an offline browser-based schema designer, ScyllaDB launched high-performance vector search, and ClickHouse 25.12 brings performance and UX improvements, including 50x faster <code>distinctJSONPaths</code> and 6x faster <code>DISTINCT</code> queries on <code>LowCardinality</code> columns. <a href="https://app.daily.dev/feed-by-ids?id=q95L7aBiO&amp;id=F9apuNvT1&amp;id=bbohKtF7X" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Malware campaigns target developers via VS Code extensions and LinkedIn</strong>: The Evelyn Stealer malware campaign is targeting developers through malicious VS Code extensions, exfiltrating credentials and crypto wallets, while a separate phishing campaign uses LinkedIn direct messages to deliver remote access trojans via DLL sideloading. <a href="https://app.daily.dev/feed-by-ids?id=IOUku4koZ&amp;id=hCAXdmj5Q" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vercel increases cron job limits, Docker SOCI for faster container startup</strong>: Vercel increased cron job limits to 100 per project across all plans, and Grab implemented Docker SOCI lazy loading, achieving 4x faster image pull times and 30-40% reduction in container startup times in production. <a href="https://app.daily.dev/feed-by-ids?id=fn5DJh7a8&amp;id=nl6ArTdyL" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI-generated malware: the next big thing, apparently</strong>: Check Point Research is sounding the alarm about VoidLink, supposedly the ‘first fully AI-generated advanced malware,’ created by a single developer in less than a week. It’s got eBPF rootkits and cloud enumeration modules, because apparently, AI can now churn out 88,000 lines of malicious code faster than you can say ‘zero-day.’ The framing is, of course, alarmist, suggesting a ‘shift from theoretical AI-derived cyber threats to actualized incidents.’ Because, you know, we needed more things to worry about. <a href="https://app.daily.dev/posts/nLTC7GTyu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI will make your essays ‘obsolete’ but don’t worry, critical thinking is still in</strong>: An article suggests generative AI makes traditional essays ‘less relevant’ and that education must adapt by focusing on ‘flexibility, creativity, and critical thinking.’ It compares this to calculators replacing manual computation, conveniently ignoring that calculators don’t write your entire math homework for you. So, while AI might make your kids better at ‘debate’ and ‘advanced reasoning,’ they probably won’t be able to write a coherent sentence without a prompt. <a href="https://app.daily.dev/posts/gEyepJU8w" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is plateauing, but don’t tell the web dev job market</strong>: The web development job market is ‘recovering’ with a 12-13% increase in listings, while ‘improvements to large AI models appear to be plateauing.’ Yet, the article still insists AI won’t replace web developers, comparing it to autopilot. So, AI is both stagnating and not a threat, but you still need to master prompt engineering and ‘deep expertise in advanced web technologies’ because AI ‘struggles in these areas.’ Make up your mind, already. <a href="https://app.daily.dev/posts/OjN0nkHHH" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
