<h2>Must know</h2>
<ul>
<li><strong>Chrome zero-day under active exploitation</strong>: Google has released an urgent update for Chrome (145.0.7632.75/76) to patch CVE-2026-2441, a high-severity use-after-free vulnerability in its CSS font feature parsing that is actively being exploited in the wild. This marks Chrome’s first zero-day of the year, allowing for potential arbitrary code execution within the browser sandbox. The fix involves replacing reference-based iteration with a copy to prevent dangling pointers, a class of bug that Rust’s ownership model aims to prevent. <a href="https://app.daily.dev/posts/ISqaRw1RF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft Copilot chat security flaw exposed confidential emails</strong>: Microsoft confirmed a security bug (CW1226324) in Copilot Chat that allowed the AI to bypass confidentiality labels and Data Loss Prevention (DLP) policies, summarizing emails from users’ Sent Items and Drafts folders without authorization. The flaw, active since late January, highlights broader enterprise anxiety about AI tools accessing sensitive corporate data, with Microsoft rolling out a fix in early February. This incident adds to previous Copilot security concerns and the European Parliament’s decision to block built-in AI features on work devices due to similar confidentiality issues. <a href="https://app.daily.dev/posts/jSmIdhAQU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical vulnerabilities in popular VS Code extensions</strong>: OX Security disclosed critical vulnerabilities in four popular VS Code extensions—Live Server, Code Runner, Markdown Preview Enhanced, and Microsoft’s Live Preview—totaling 128 million installs. These flaws enable file theft, remote code execution, and local network reconnaissance. While Microsoft silently patched Live Preview, three other maintainers never responded, leaving millions of developers at risk. Developers are advised to update Live Preview, disable unused extensions, and avoid browsing untrusted sites with localhost servers running. <a href="https://app.daily.dev/posts/IuDICegHN" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI agent development, security, and governance accelerate</strong>: Enterprises are rapidly adopting and developing agentic AI, focusing on autonomous AI agents for various tasks, with new tools like StrongDM ID emerging to provide identity for AI agents and Cencurity offering security gateways to proxy LLM traffic. <a href="https://app.daily.dev/feed-by-ids?id=2UmMyMAEf&amp;id=32FCqrqN0&amp;id=4fmmBjp20&amp;id=Ck3BiV0To&amp;id=bbSNghGeS&amp;id=fvUXvxfty&amp;id=hYwz99Mp8&amp;id=i8naoDZlv&amp;id=nicRNl9Cx&amp;id=pJKpsRETp" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>India’s AI push and global AI growth</strong>: India is making a significant push in AI, with companies like Sarvam launching open-source multilingual LLMs and Adani planning a $100 billion investment in AI datacenters, while Canva’s revenue growth is increasingly driven by LLM referral traffic. <a href="https://app.daily.dev/feed-by-ids?id=8IlGUVGaS&amp;id=JaYsn9TtR&amp;id=NDuZu47ZF&amp;id=TI4zH4l6O&amp;id=W4mWB5Hp1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI model performance, cost, and sovereign AI initiatives</strong>: Anthropic’s Claude Sonnet 4.6 is marketed as cheaper and improved, but independent benchmarks suggest it’s a ‘token muncher,’ while India and Japan are making strides in sovereign AI with NVIDIA’s support, and Google DeepMind questions the depth of LLM moral reasoning. <a href="https://app.daily.dev/feed-by-ids?id=i4O6sQBIw&amp;id=j12nF8mSg&amp;id=jj2Kauaop&amp;id=lUz1VvbAr&amp;id=y0N7mzotz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI’s impact on software development and engineering roles</strong>: Spotify’s CEO claims senior developers haven’t written code since December 2025, supervising AI instead, while Martin Fowler warns AI can accelerate technical debt, and ‘Outcome Engineering’ proposes reframing software around impact rather than code. <a href="https://app.daily.dev/feed-by-ids?id=7hHra6Ai0&amp;id=WUJRo6v7h&amp;id=enHqyOYiB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Tailscale peer relays general availability</strong>: Tailscale announced the general availability of Peer Relays, allowing users to deploy high-throughput relay nodes on any Tailscale device to improve network connectivity and performance, now supporting static endpoints and enhanced observability. <a href="https://app.daily.dev/posts/Q0uzlfbeJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks launches custom agents</strong>: Databricks launched Custom Agents (formerly Agent Framework), enabling developers to build, test, and deploy production-quality AI agents as fully managed Databricks Apps on serverless compute, with features like framework-agnostic development and built-in memory. <a href="https://app.daily.dev/posts/QJn2roYcs" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA run:ai GPU fractioning for LLM inference performance</strong>: NVIDIA and Nebius AI Cloud benchmarked GPU fractioning using NVIDIA Run:ai, demonstrating that fractional GPU scheduling is a production-ready strategy for maximizing GPU ROI, achieving significant throughput and concurrent user capacity for LLM inference. <a href="https://app.daily.dev/posts/LsiRfCVWP" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI productivity impact: more hype than reality</strong>: An NBER survey of nearly 6,000 executives found over 80% reported no measurable impact from AI on productivity or employment over the past three years. Yet, these same executives anticipate AI will eliminate 1.75 million jobs by 2028. So, AI isn’t making us more productive now, but it’s definitely coming for your job later. Makes perfect sense. <a href="https://app.daily.dev/posts/rfmYnNc4e" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic’s C compiler: a marketing masterpiece</strong>: Anthropic released a marketing video claiming Claude built a C compiler ‘from scratch’ with ‘no human intervention’ that could compile Linux. Turns out, it used 37 years of GCC test suites, an online GCC oracle, required human restarts, cost $20,000 in API calls, and couldn’t even boot Linux. But hey, ‘dishonest marketing language’ is just part of the AI game, right? <a href="https://app.daily.dev/posts/0csgEkRMy" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Pentagon vs anthropic: a very dramatic misunderstanding</strong>: The Pentagon is threatening to label Anthropic a national security risk because they <em>inquired</em> about Claude’s alleged use in a military operation in Venezuela. Apparently, asking questions about your AI being used in a conflict zone is now considered ‘condemnation’ by the DoD. Who knew curiosity was such a threat to national security? <a href="https://app.daily.dev/posts/9eGp6hlhZ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
