<h2>Must know</h2>
<ul>
<li><strong>Microsoft patches actively exploited Windows 0-day</strong>: Microsoft released an urgent patch for CVE-2026-20805, a Windows information disclosure vulnerability actively exploited in the wild. This flaw allows attackers to leak memory addresses, potentially bypassing ASLR as part of larger attack chains. The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has mandated federal agencies patch this by February 3, underscoring its critical nature. This is part of Microsoft’s January 2026 Patch Tuesday, which addressed 112 CVEs, including several critical remote code execution vulnerabilities in Office. <a href="https://app.daily.dev/posts/Kwy6WAbQP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub Copilot deprecates older AI models</strong>: GitHub is deprecating three AI models used by Copilot—Claude Opus 4.1, GPT-5, and GPT-5-Codex—effective February 17, 2026. Users must update their workflows to integrate the new replacements: Claude Opus 4.5, GPT-5.2, and GPT-5.1-Codex, respectively. Enterprise administrators might need to manually enable these new models in Copilot settings to ensure continuous service, as the old models will be automatically removed. This highlights the rapid evolution and necessary adaptation in AI-powered coding tools. <a href="https://app.daily.dev/feed-by-ids?id=ImHNcBPhJ&amp;id=ZoKLGVJ9V&amp;id=uw83EGG5r" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical RCE vulnerabilities found in AI/ML Python libraries</strong>: Security researchers discovered remote code execution (RCE) vulnerabilities in popular AI/ML Python libraries from Nvidia (NeMo), Salesforce (Uni2TS), and Apple (FlexTok). These flaws stem from unsafe use of Hydra’s instantiate() function, allowing malicious code embedded in model metadata to execute when loaded. With tens of millions of downloads, particularly via Hugging Face, these vulnerabilities posed a significant supply chain risk, though no in-the-wild exploitation has been reported. Patches have been issued, and Meta has updated Hydra documentation with security warnings. <a href="https://app.daily.dev/feed-by-ids?id=4tbUFTlCZ&amp;id=c7PnFTJdg&amp;id=kZzcz2YmW" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Node.js releases security fixes for DoS vulnerability</strong>: Node.js versions 20.20.0 and 25.3.0 include critical security fixes for a denial-of-service (DoS) vulnerability related to stack space exhaustion, particularly affecting users of async_hooks, React Server Components, and Next.js. <a href="https://app.daily.dev/feed-by-ids?id=DwsoMgMKH&amp;id=ImHNcBPhJ&amp;id=ZoKLGVJ9V&amp;id=uw83EGG5r" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Advanced VoidLink Linux malware targets cloud and containers</strong>: Check Point Research identified VoidLink, an advanced Linux malware written in Zig, designed to target cloud and container environments like Docker and Kubernetes, featuring 37+ modular plugins for credential harvesting and evasion. <a href="https://app.daily.dev/feed-by-ids?id=uJaIfhgHH&amp;id=QWUKVWdsK&amp;id=x9wl564Rc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents in business platforms vulnerable to critical exploits</strong>: ServiceNow’s Virtual Agent and Superhuman AI’s email platform were found to have critical vulnerabilities, allowing full system control or email exfiltration via weak authentication and prompt injection, respectively, highlighting risks in integrating AI agents without robust security. <a href="https://app.daily.dev/feed-by-ids?id=2A2nfRTHq&amp;id=EDFS33r2W&amp;id=KcLhdNkuO&amp;id=qah3DxopU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New terminal manager for AI coding agents released</strong>: ‘Agent of Empires’ (aoe), a Rust-based terminal session manager, has been released to help developers manage multiple AI coding agent sessions, offering a TUI dashboard and automatic status detection for tools like Claude Code and OpenCode. <a href="https://app.daily.dev/posts/AqojY4mSV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ZLUDA brings CUDA compatibility to AMD and Intel GPUs</strong>: ZLUDA, an open-source project, now offers full llama.cpp support and Windows compatibility, allowing CUDA applications to run on AMD and Intel GPUs without code modification, benefiting developers running ML/AI workloads outside the NVIDIA ecosystem. <a href="https://app.daily.dev/feed-by-ids?id=ImHNcBPhJ&amp;id=ZoKLGVJ9V&amp;id=uw83EGG5r" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Laravel 12.47.0 introduces new developer features</strong>: Laravel 12.47.0 brings several enhancements, including the @includeIsolated Blade directive for isolated template includes, a Cache::withoutOverlapping() method for preventing concurrent cache operations, and enhanced enum support. <a href="https://app.daily.dev/posts/SQ5jDUcaB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux Mint 22.3 “Zena” released with GUI upgrades</strong>: Linux Mint 22.3 “Zena” is out, featuring a redesigned, customizable application menu, an on-screen keyboard, enhanced Wayland support, and new System Information and Administration tools, all based on Ubuntu 24.04.3 LTS. <a href="https://app.daily.dev/posts/ecK1ViBJk" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI verification bottleneck: developers still doing the heavy lifting</strong>: Apparently, AI code generation is so amazing that developers are now spending more time verifying its unreliable output than they save in writing it. The New Stack reports that 96% of developers don’t trust AI-generated code, leading to increased technical debt and correction work. So, the ‘productivity boost’ just means shifting toil from creation to endless babysitting. Progress! <a href="https://app.daily.dev/feed-by-ids?id=SVMT3qR0j&amp;id=1lxfdifYi" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Shadow AI: because we needed more ways for data to leak</strong>: Nordic APIs warns about ‘Shadow AI,’ which is apparently the new ‘Shadow API’ but with extra steps for disaster. These unsanctioned AI tools are ingesting new data and operating with weak identity controls, leading to ‘identity collapse’ where nobody knows who did what. So, while everyone’s busy trying to secure their APIs, employees are just casually plugging in AI tools that could expose everything. What could possibly go wrong? <a href="https://app.daily.dev/posts/yF1D00gr0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>40% of AI agents will fail, but don’t worry, it’s your fault</strong>: Gartner predicts 40% of agentic AI projects will be canceled by 2027, not because the AI is bad, but because companies lack ‘guardrails.’ Apparently, we need to treat AI agents like human employees with strict governance, technical constraints, and human oversight. So, after all the hype about autonomous AI, it turns out they still need a babysitter and a manager. Who knew? <a href="https://app.daily.dev/feed-by-ids?id=GdT63S7BD&amp;id=QnCK3ryB4&amp;id=SaUBaotgn&amp;id=g50WJmopi&amp;id=y1lywwnED" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
