AI Engineer

Cassidy's blog

CNCF

InfoQ

InfoWorld

LangChain

<h2>Must know</h2>
<ul>
<li><strong>AI coding tools are a mixed bag for productivity</strong>: AI coding tools like Copilot and Claude Code are widely used, but their impact on developer productivity is complex. While some developers see measurable gains, others find themselves spending more time correcting AI-generated code than it saves, especially for nuanced tasks. Experienced open-source developers, according to METR research, often see minimal productivity gains, highlighting that AI amplifies existing process debts rather than replacing foundational engineering skills. This suggests that while AI can accelerate prototyping, human expertise remains crucial for complex problem-solving and maintaining code quality. <a href="https://app.daily.dev/feed-by-ids?id=3DI7vnha4&amp;id=ATZPMmQn8&amp;id=DdiXGTqPN&amp;id=L2hSWhbgY&amp;id=LJqUceZbE&amp;id=Y8gY63ibF&amp;id=aGFVTrinz&amp;id=aKVj0i6Bm&amp;id=j91dGf3TP&amp;id=tJZLXej6S" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitLab 18.8 introduces AI-orchestrated DevOps</strong>: GitLab 18.8 is out, bringing a Duo Agent Platform that orchestrates AI agents across the software development lifecycle. This includes a Planner Agent for managing work items and a Security Analyst Agent that uses natural language processing to triage vulnerabilities, even automatically dismissing irrelevant ones. A Central AI Catalog is also included for agent discovery and management, aiming to automate security and planning within DevOps pipelines, according to GitLab’s official release announcement. <a href="https://app.daily.dev/posts/R9WyVAVPv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AMD CPUs vulnerable to StackWarp attack</strong>: A new hardware vulnerability, StackWarp (CVE-2025-29943), affects AMD Zen 1–5 CPUs, including EPYC 7003, 8004, 9004, and 9005 series. This flaw allows privileged attackers to exploit an undocumented control bit to manipulate the CPU stack engine, corrupting the stack pointer and hijacking control flow in SEV-SNP-protected confidential virtual machines. AMD has released microcode updates, with more firmware patches planned for April 2026, and advises disabling hyperthreading on critical systems as a mitigation. <a href="https://app.daily.dev/posts/9DURDnZX3" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Microsoft open-sources XAML Studio</strong>: Microsoft has open-sourced XAML Studio, a lightweight prototyping tool for WinUI and UWP projects, under the .NET Foundation, with a redesigned UI and deeper WinUI 3 integration coming in version 2.0. <a href="https://app.daily.dev/posts/hUJrXK08w" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Malicious Chrome extensions steal enterprise session tokens</strong>: Five Chrome extensions, disguised as productivity tools, were found stealing authentication session tokens from enterprise HR and ERP platforms like Workday and NetSuite, bypassing MFA and allowing full account takeover. <a href="https://app.daily.dev/posts/3ImoRkfEG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>jQuery 4.0.0 released after a decade</strong>: jQuery 4.0.0, released for its 20th anniversary, focuses on removing deprecated features, dropping support for older browsers, and modernizing its codebase to ES modules, while reducing library size and improving XSS protection. <a href="https://app.daily.dev/feed-by-ids?id=5SOCKxGlX&amp;id=dThF1vPz2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenCode offers open-source AI coding alternative</strong>: OpenCode is an open-source, CLI-first AI coding agent designed as a free, privacy-focused alternative to proprietary tools, allowing users to link any AI model via API key or access curated models through OpenCode Zen. <a href="https://app.daily.dev/feed-by-ids?id=15Mjf12HP&amp;id=3RoDG910s&amp;id=KwgMObiSh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Windows 11 23H2 emergency patch for shutdown and RDP bugs</strong>: Microsoft released an emergency update (KB5077797) for Windows 11 23H2 to fix critical bugs preventing proper shutdown/restart and causing persistent Remote Desktop credential prompts, which were introduced by January’s Patch Tuesday. <a href="https://app.daily.dev/posts/u2UYa6XNM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents cause ‘psychosis’ in production</strong>: Changelog News raises concerns about ‘AI agent psychosis,’ describing erratic or unreliable behavior by AI agents in production, while <a href="http://Monday.com" target="_blank" rel="noopener nofollow">Monday.com</a> engineers are mitigating LLM hallucinations in their Slack bot by server-side pre-computation and human-in-the-loop approvals. <a href="https://app.daily.dev/feed-by-ids?id=AJzG7x8FZ&amp;id=AS53pzdrZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Wine 11.0 stable release brings major upgrades</strong>: Wine 11.0 stable is out with over 6,300 changes, including NTSYNC kernel module support for better performance, a completed WoW64 architecture, and upgraded graphics with EGL as the default OpenGL backend and Vulkan 1.4.335 support. <a href="https://app.daily.dev/posts/5LDn1X8et" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>NVIDIA allegedly sought pirated books for AI training</strong>: Apparently, NVIDIA executives, those paragons of innovation, allegedly contacted Anna’s Archive, a well-known pirate library, to get millions of copyrighted books for AI model training. The lawsuit claims they were warned it was illegal but proceeded anyway, also using other notorious pirate sites. Because why pay for data when you can just, you know, ‘innovate’ around copyright laws? Truly thinking different. <a href="https://app.daily.dev/posts/2IcgLruSE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI security market projected to be worth a trillion dollars</strong>: Venture capitalists are now projecting the AI security market to be worth between $800 billion and $1.2 trillion by 2031. Because nothing says ‘we’re solving real problems’ like a market projection that’s roughly the GDP of Spain, based on the idea that AI agents might try to blackmail you. Witness AI even got $58 million to tackle ‘shadow AI.’ Sounds like a gold rush for consultants and VCs, not necessarily for actual security. <a href="https://app.daily.dev/feed-by-ids?id=2XMeIQvIr&amp;id=6gbkOathj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-driven ERP: the future is now, or maybe later</strong>: A survey says 70% of executives think traditional ERP isn’t the future, yet 97% are satisfied with their current systems. So, they’re happy but want something else? Now, 33% are eyeing ‘autonomous, AI-driven systems’ for ERP, while only 5% of AI pilots deliver measurable value. Sounds like a lot of talk about ‘composable and sovereign AI architectures’ to justify more consulting fees, rather than actual, proven solutions. <a href="https://app.daily.dev/feed-by-ids?id=QJmBgvguS&amp;id=zDjVyaxS1" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
