<h2>Must know</h2>
<ul>
<li><strong>AI-generated code is a security nightmare waiting to happen</strong>: Security experts are sounding the alarm about AI-generated code, warning it could lead to catastrophic failures if not properly reviewed. The concern is that AI models, trained on potentially vulnerable codebases, propagate weaknesses, especially when developers use them for security-critical systems without full comprehension. They recommend using battle-tested libraries and strongly typed languages like Rust to mitigate risks, as reported by security experts. This comes as LLVM and Oh My Zsh are implementing policies requiring human oversight and accountability for AI-assisted contributions, citing an increase in low-quality submissions. <a href="https://app.daily.dev/feed-by-ids?id=rlAZMkOYu&amp;id=EdB2gt1Vu&amp;id=tr2l7aEc5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic and Google AI models vulnerable to prompt injection</strong>: Anthropic’s Git MCP server was found to have critical vulnerabilities that enable prompt injection, unauthorized code execution, and LLM tampering. These flaws, affecting versions prior to 2025.12.18, could be exploited by chaining attacks across components. Separately, Google Gemini was vulnerable to manipulation via calendar invitations, where malicious instructions embedded in event data could lead to exfiltration of private meeting information. Both incidents highlight the growing risk of semantic threats in AI systems, where context manipulation, not traditional code exploits, is the attack vector. <a href="https://app.daily.dev/feed-by-ids?id=U6A8HrCeq&amp;id=rKbuKp9sI&amp;id=dHjNBMkFO&amp;id=j6bBmLvCW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubernetes is now the standard for AI infrastructure</strong>: Kubernetes has become the de facto standard for AI workloads, with 82% of organizations using it in production and 66% specifically for AI inference, according to the 2025 CNCF Annual Cloud Native Survey. The adoption barriers have shifted from technical complexity to cultural resistance, emphasizing the need for mature cloud-native practices like GitOps and platform engineering. However, this increased AI adoption is straining open-source infrastructure, raising sustainability concerns for ongoing community investment and maintainer support. <a href="https://app.daily.dev/posts/Lwsgn3lZ5" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Next.js 16 brings turbo builds and AI debugging</strong>: Next.js 16 is out, making Turbopack the default bundler for faster builds and hot reloading, introducing explicit Cache Components, and integrating AI-assisted debugging via Model Context Protocol (MCP) for automated fix suggestions. <a href="https://app.daily.dev/posts/91xiEmcVe" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub enhances security with code-to-cloud traceability</strong>: GitHub has rolled out new artifact metadata APIs for code-to-cloud traceability, allowing teams to connect build artifacts to their source and deployment data, and introduced SLSA Build Level 3 security with cryptographic attestations. <a href="https://app.daily.dev/posts/9XGazeUyY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-powered malware is here and it’s fast</strong>: Check Point Research identified VoidLink as the first fully AI-generated advanced malware, developed by a single individual in less than a week, producing a codebase equivalent to 30 weeks of multi-team effort. <a href="https://app.daily.dev/posts/nLTC7GTyu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New framework for user-space Linux block device drivers</strong>: ublk is a new framework enabling Linux block device drivers to be developed in user space, offering faster prototyping, easier debugging, and safer crash handling compared to traditional in-kernel drivers. <a href="https://app.daily.dev/posts/0oo0Z2s5H" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New open-source tool for managing code snippets</strong>: Snipplle is a new open-source platform built with Nuxt 4, TypeScript, and PostgreSQL for managing code snippets, offering syntax highlighting, collections, workspaces for collaboration, and CLI integration. <a href="https://app.daily.dev/posts/Lhaif3YU8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vercel increases cron job limits</strong>: Vercel has increased the limit for scheduled cron jobs to 100 per project across all plans (Hobby, Pro, Enterprise), removing previous per-team caps and simplifying configuration via vercel.json. <a href="https://app.daily.dev/posts/fn5DJh7a8" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI coding tools: the emperor has no clothes (or good code)</strong>: Turns out, those shiny new AI coding tools aren’t quite the silver bullet they’re hyped up to be. A Carnegie Mellon study found that while AI assistants might give you a 30-40% boost in the first month, those gains vanish by month three, leaving you with 30% more static analysis warnings and 40% higher code complexity. So much for ‘revolutionizing’ development; seems like we’re just trading one kind of technical debt for another, as reported by CMU and GitClear. <a href="https://app.daily.dev/feed-by-ids?id=BqO4RcudK&amp;id=EZUXhYwRZ&amp;id=LiH8HLrgt" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Tailwind Labs lays off 75% of developers, blames AI for lack of documentation visits</strong>: In a stunning display of ‘AI ate my homework,’ Tailwind Labs laid off 75% of its developers, attributing declining website traffic and revenue to AI-assisted coding. Apparently, developers are now so good at coding with AI that they no longer need to visit documentation sites. This raises the profound question: if AI automates away the need for documentation, who’s going to build the next generation of frameworks for AI to automate? The circle of life, or just a convenient excuse? <a href="https://app.daily.dev/posts/PBAzOX1yL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>CEOs still waiting for that AI magic money tree</strong>: A PwC survey of 4,454 CEOs reveals a shocking truth: 56% saw no increase in revenue or decrease in costs from their AI investments. Only a measly 12% saw both. Meanwhile, CEO confidence is at a five-year low, thanks to geopolitical risks, cyber threats, and, you guessed it, unreliable AI returns. It seems the ‘trillion-dollar AI deals’ are more about hype than actual dividends, leaving many executives wondering if they bought a lemon, as per PwC’s findings. <a href="https://app.daily.dev/posts/Kov24DV15" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
