<h2>Must know</h2>
<ul>
<li><strong>AI agents are getting real, but so are the headaches</strong>: AI agents are now capable of complex research, analyzing thousands of papers, and generating reports autonomously, boosting productivity. However, this comes with a new set of challenges, including ‘AI agent psychosis’—erratic behavior in production—and the risk of hallucinations, as highlighted by Changelog News and <a href="http://Monday.com" target="_blank" rel="noopener nofollow">Monday.com</a>. Companies like <a href="http://Monday.com" target="_blank" rel="noopener nofollow">Monday.com</a> are tackling these issues with server-side pre-computation and human-in-the-loop approvals, suggesting that while AI agents are powerful, they require significant engineering oversight to be reliable. <a href="https://app.daily.dev/feed-by-ids?id=AJzG7x8FZ&amp;id=AS53pzdrZ&amp;id=ATZPMmQn8&amp;id=NG1jOcpWT" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AMD’s secure VMs hit by new hardware vulnerability</strong>: A new hardware vulnerability, StackWarp (CVE-2025-29943), has been discovered in AMD Zen 1–5 CPUs, impacting the security of SEV-SNP-protected confidential virtual machines. This exploit allows privileged attackers to corrupt the stack pointer and hijack control flow, bypassing what was thought to be enhanced VM security. AMD has released microcode updates, with more firmware patches planned, but the incident underscores the ongoing challenge of securing confidential computing environments. <a href="https://app.daily.dev/posts/9DURDnZX3" 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. These sophisticated attacks bypassed multi-factor authentication, allowing attackers full access to compromised accounts. Socket.dev researchers reported the discovery, prompting Google to act, but not before thousands of users were affected, highlighting a critical vulnerability in browser extension security. <a href="https://app.daily.dev/posts/3ImoRkfEG" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>GitLab 18.8 introduces AI-orchestrated DevOps</strong>: GitLab 18.8 is out with a Duo Agent Platform that orchestrates AI agents for tasks like managing work items and triaging security vulnerabilities, aiming to automate parts of the software development lifecycle. <a href="https://app.daily.dev/posts/R9WyVAVPv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft open-sources XAML Studio</strong>: Microsoft has open-sourced XAML Studio under the .NET Foundation, aiming to foster community contributions for this lightweight prototyping tool for WinUI and UWP projects, 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>Intel boosts LLM support on Arc Graphics</strong>: Intel’s LLM-Scaler-Omni update enhances support for ComfyUI and SGLang on Arc Graphics, focusing on improving image, voice, and video generation capabilities within its AI ecosystem. <a href="https://app.daily.dev/posts/XBX3pzO5M" 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 version 23H2 to fix critical bugs preventing proper shutdown/restart and causing persistent credential prompts for Remote Desktop users. <a href="https://app.daily.dev/posts/u2UYa6XNM" 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 old browser support, and modernizing its codebase to ES modules, while still being used by a significant portion of websites. <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>AI disrupts Indian IT outsourcing and cloud costs</strong>: India’s top IT outsourcers are drastically slowing hiring due to AI adoption, while Microsoft Copilot is showing high resource consumption on Fabric platforms, indicating AI’s dual impact on workforce and cloud operational costs. <a href="https://app.daily.dev/feed-by-ids?id=DLuyoqNjG&amp;id=nKPk64Gpl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Wine 11.0 stable brings major technical upgrades</strong>: Wine 11.0 stable is out with over 6,300 changes, including NTSYNC kernel module support, a completed WoW64 architecture, and upgraded graphics stack with Vulkan 1.4.335 support, enhancing Windows application compatibility on Linux. <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>Sequoia Capital’s ‘bold’ move into AI: Or just hedging bets?</strong>: Sequoia Capital is investing in Anthropic’s $25 billion funding round, despite already holding stakes in OpenAI and xAI. The FT is calling this a break from established VC norms, but let’s be real, it looks more like a classic ‘don’t put all your eggs in one basket’ strategy, especially when the baskets are all in the same AI gold rush. Sam Altman’s past warnings about investors losing access to confidential info if they fund competitors seem to have been conveniently forgotten, or perhaps, just ignored. <a href="https://app.daily.dev/posts/yuiHTajaU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding tools: The ‘productivity’ boost that requires more human babysitting</strong>: Apparently, AI coding tools are ‘reshaping’ development by generating lower-quality code for complex tasks, requiring ‘substantial human correction’ and ‘technical knowledge.’ So, we’re trading one kind of work for another, more tedious kind, all while developers are told they need ‘managerial’ skills to prompt these digital interns. It’s almost like the tools are just amplifying our existing process debt, not actually solving anything fundamental, as one article so subtly points out. <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>NVIDIA’s alleged book piracy: Training AI on stolen goods, who’d have thought?</strong>: A lawsuit alleges NVIDIA executives contacted Anna’s Archive, a known pirate library, to access millions of copyrighted books for AI model training, despite warnings about the illegal content. It seems the ‘innovation’ in AI sometimes involves a shortcut through copyright law. While the lawsuit is still just allegations, it’s a stark reminder that the data fueling these advanced models often comes from ethically questionable sources, and apparently, some executives are perfectly fine with that. <a href="https://app.daily.dev/posts/2IcgLruSE" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
