<h2>Must know</h2>
<ul>
<li><strong>Chrome zero-day under active exploitation</strong>: Google just patched a high-severity zero-day vulnerability, CVE-2026-2441, in Chrome that’s actively being exploited. It’s a use-after-free bug in the CSS font feature parsing, meaning a bad actor could execute arbitrary code in your browser. Update your Chrome immediately, as this is the first zero-day of the year for the browser, according to “they hacked CSS.” <a href="https://app.daily.dev/posts/ISqaRw1RF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft Copilot Chat exposed confidential emails</strong>: Microsoft confirmed a security bug in Copilot Chat that allowed the AI to bypass confidentiality labels and Data Loss Prevention policies, summarizing emails from users’ Sent Items and Drafts without authorization. The flaw was active since late January, and while a fix is rolling out, Microsoft hasn’t disclosed how many business customers were affected. This highlights ongoing enterprise anxiety about AI tools handling sensitive corporate data. <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 widely used VS Code extensions: Live Server, Code Runner, Markdown Preview Enhanced, and Microsoft’s Live Preview, totaling 128 million installs. These flaws could lead to file theft, remote code execution, and local network reconnaissance. While Microsoft silently patched Live Preview, many maintainers haven’t responded, so update your extensions and disable unused ones. <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 productivity gains remain elusive for most businesses</strong>: A new NBER survey of nearly 6,000 executives found over 80% reported no measurable impact from AI on productivity or employment, despite 69% using some form of AI. <a href="https://app.daily.dev/posts/rfmYnNc4e" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic’s new Sonnet 4.6 model claims big performance for less</strong>: Anthropic released Claude Sonnet 4.6, claiming it matches or exceeds Opus 4.6 performance on enterprise benchmarks at a significantly lower price, though independent analysis suggests it uses over 4x more tokens, potentially negating cost savings. <a href="https://app.daily.dev/feed-by-ids?id=3TzVcx3Vs&amp;id=i4O6sQBIw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub Copilot expands with Claude Opus 4.6 and Windows support</strong>: GitHub Copilot now integrates Claude Opus 4.6 across various IDEs and has extended its autonomous coding agent to support Windows development environments. <a href="https://app.daily.dev/feed-by-ids?id=49jCLhThU&amp;id=xntn7IAsM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mistral AI acquires Koyeb to boost compute ambitions</strong>: Mistral AI made its first acquisition, buying cloud startup Koyeb to integrate its serverless platform into Mistral Compute, signaling a move towards full-stack AI infrastructure for enterprise clients. <a href="https://app.daily.dev/posts/s077dNRb2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Sarvam AI launches new models and expands to consumer devices</strong>: Indian AI startup Sarvam AI released new open-source 30B and 105B parameter LLMs and is integrating lightweight edge AI models into Nokia feature phones, cars, and smart glasses. <a href="https://app.daily.dev/feed-by-ids?id=JaYsn9TtR&amp;id=NDuZu47ZF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents are struggling with enterprise data</strong>: A study by IBM Research and UC Berkeley found that LLM-based agents frequently fail in enterprise IT workflows due to “Incorrect Verification,” with stronger models failing more cleanly but still requiring architectural interventions like summarizer agents for reliability. <a href="https://app.daily.dev/feed-by-ids?id=FtPdxfWs5&amp;id=Iw0Phxscr&amp;id=K5A0Ez19y&amp;id=L8eI95l3u&amp;id=V9sOWEokm&amp;id=WXyNkLy8T&amp;id=XYiUFiLYo&amp;id=cyEsB6IQv&amp;id=dfEz8EdS5&amp;id=dy6cgx4MB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New npm CLI release enhances security</strong>: npm CLI v11.10.0 introduces bulk OIDC trusted publishing configuration and a new <code>--allow-git</code> flag for <code>npm install</code> to explicitly control git dependencies, addressing a security vulnerability. <a href="https://app.daily.dev/posts/3xMuRnyG7" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic’s C compiler claims are a bit much</strong>: Anthropic is out there claiming Claude built a C compiler “from scratch” with “no human intervention.” Turns out, it used 37 years of GCC test suites, an online oracle, cost $20,000 in API calls, and needed human restarts. It’s a cool technical feat, but maybe dial back the marketing hype a notch, Anthropic. <a href="https://app.daily.dev/feed-by-ids?id=0csgEkRMy&amp;id=6spQIGtcM&amp;id=dhghGjegM&amp;id=kcB1LfSD2&amp;id=poO8tUjKL&amp;id=rsFWrvtAL&amp;id=v3dp2qPCw&amp;id=z9Rc6eFF4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is not shipping faster, it’s creating bottlenecks</strong>: CircleCI’s 2026 State of Software Delivery report says AI isn’t making us ship faster; it’s actually creating bottlenecks at the review and integration stages. While code generation is up, main branch throughput is down, and recovery times are climbing. So much for AI being the magic bullet for developer velocity. <a href="https://app.daily.dev/posts/okTNHJRhB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated code is flooding open source with ‘slop’</strong>: Godot engine maintainers are calling out the “draining and demoralizing” influx of low-quality, AI-generated pull requests. GitHub is even getting blamed for incentivizing this “AI slop.” Apparently, AI is great at generating code, not so great at generating useful contributions. <a href="https://app.daily.dev/feed-by-ids?id=8WhnSQAtX&amp;id=HouoEDJq7&amp;id=QnWXK2Mm8&amp;id=YaGme2ewn" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
