<h2>Must know</h2>
<ul>
<li><strong>Chrome zero-day vulnerability under active exploitation</strong>: Google has released an urgent update for Chrome (versions 145.0.7632.75/76 for Windows/Mac and 144.0.7559.75 for Linux) to patch CVE-2026-2441, a high-severity use-after-free bug in its CSS font feature value map parsing. This is Chrome’s first zero-day of 2026, and Google confirmed it’s being actively exploited in the wild. The vulnerability allows arbitrary code execution within the browser sandbox, making immediate updates critical. <a href="https://app.daily.dev/posts/ISqaRw1RF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft Copilot Chat security flaw exposes 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 anxieties about AI tools accessing sensitive corporate data, especially given previous Copilot security concerns and the European Parliament blocking built-in AI features on work devices. <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 disclosed</strong>: OX Security disclosed critical vulnerabilities in four popular VS Code extensions: Live Server (72M downloads), Code Runner (37M), Markdown Preview Enhanced (8.5M), and Microsoft’s Live Preview (11M). These flaws, affecting 128 million installs, 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. Update Live Preview to 0.4.16+ and disable unused extensions. <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 impact remains elusive for most firms</strong>: An 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, aligning with similar findings from PwC and Deloitte. <a href="https://app.daily.dev/posts/rfmYnNc4e" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mistral AI acquires Koyeb to expand compute ambitions</strong>: Mistral AI acquired cloud startup Koyeb, integrating its serverless deployment platform into Mistral Compute, signaling a strategic shift towards full-stack AI infrastructure and targeting enterprise customers with data residency and hybrid deployment needs. <a href="https://app.daily.dev/posts/s077dNRb2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Asahi Linux progresses on Apple M3 and Apple Silicon support</strong>: Asahi Linux reports significant progress on Apple M3 hardware, with key peripherals like keyboard, touchpad, WiFi, and USB3 working, and DisplayPort Alt Mode via USB-C now functional in experimental branches after years of reverse engineering. <a href="https://app.daily.dev/posts/9Q140qigi" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google launches Lyria 3 AI music generation in Gemini</strong>: Google integrated Lyria 3, an AI music generation model, into Gemini Tools and the Gemini app, allowing users to create 30-second songs with lyrics from text, photos, or videos, complete with AI-generated cover art and SynthID watermarking. <a href="https://app.daily.dev/posts/eN6VW0Hvt" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Sarvam AI launches new models and expands device integration</strong>: Indian AI startup Sarvam AI launched new open-source 30B and 105B parameter LLMs, along with text-to-speech, speech-to-text, and vision models, and is expanding into consumer devices like Nokia feature phones, cars, and smart glasses with lightweight edge AI models. <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>Palo Alto Networks acquires Koi Security for AI agent governance</strong>: Palo Alto Networks acquired Koi Security for approximately $400 million, aiming to integrate Koi’s platform for monitoring and governing software components, browser extensions, and AI models into Prisma AIRS and Cortex XDR to address growing AI agent security risks. <a href="https://app.daily.dev/feed-by-ids?id=3Hx9gllgC&amp;id=U1HM18ytf&amp;id=70bwD2IsO&amp;id=Hx9SCREZp&amp;id=IOwwjUreT&amp;id=L7iJSIIDS&amp;id=VukD3HNtV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 kernel updates include LoongArch features and driver removal</strong>: The Linux 7.0 kernel is set to include new features for the LoongArch architecture, such as 128-bit atomic cmpxchg support and memfd_secret system call wiring, while also removing the obsolete IBM Mwave ACP Modem driver from the 1990s. <a href="https://app.daily.dev/feed-by-ids?id=1hfw6pdps&amp;id=nmU7F9kzS" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic’s C compiler claims: more marketing than reality</strong>: Anthropic released a marketing video claiming Claude built a C compiler “from scratch” with “no human intervention.” 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. So, a 100,000-line functional compiler is impressive, but the marketing spin is pure fantasy. <a href="https://app.daily.dev/posts/0csgEkRMy" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic Sonnet 4.6: cheaper per token, but a token muncher</strong>: Anthropic launched Claude Sonnet 4.6, claiming it matches Opus 4.6 performance at a fraction of the cost. However, independent analysis by Artificial Analysis found Sonnet 4.6 uses over 4x more tokens than its predecessor and more than Opus 4.6 due to “adaptive thinking.” So, while the per-token price is lower, your actual bill might be higher. Classic bait and switch, but with tokens. <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>AI agents: the new solution for supply chain delivery failures</strong>: An experiment used Claude Opus 4.6, connected to a distribution-tracking database, to diagnose supply chain delivery failures, replacing static dashboards for root-cause analysis. Because, apparently, we’ve reached a point where AI agents are needed to tell us why our packages are late, and that’s somehow more efficient than, you know, fixing the actual supply chain issues. <a href="https://app.daily.dev/feed-by-ids?id=gdtsbiLsr&amp;id=qB0VtU9Kz" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
