<h2>Must know</h2>
<ul>
<li><strong>Chrome zero-day under active exploitation</strong>: Google released an urgent Chrome update to patch CVE-2026-2441, a high-severity use-after-free vulnerability in its CSS font feature parsing. This is Chrome’s first zero-day of 2026 and is actively being exploited in the wild, allowing arbitrary code execution within the browser sandbox. The fix involves replacing a reference-based iteration with a copy to prevent modification during traversal, a class of bug that Rust’s ownership model would prevent. <a href="https://app.daily.dev/posts/ISqaRw1RF" 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, including Live Server (72M downloads) and Code Runner (37M), affecting over 128 million installs. These flaws enable file theft, remote code execution, and local network reconnaissance. While Microsoft silently patched its Live Preview extension, three other maintainers have not responded, leaving millions of developers exposed to supply chain risks. <a href="https://app.daily.dev/posts/IuDICegHN" 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 (CW1226324) that allowed the AI to bypass confidentiality labels and Data Loss Prevention 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 AI features on work devices. <a href="https://app.daily.dev/posts/jSmIdhAQU" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Mistral AI acquires Koyeb for compute ambitions</strong>: Mistral AI acquired cloud startup Koyeb, integrating its serverless deployment platform into Mistral Compute to build full-stack AI infrastructure for enterprise customers, signaling a move towards vertical integration and a potential sovereign European alternative to US hyperscalers. <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 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>Asahi Linux progresses on Apple M3 and DisplayPort Alt Mode</strong>: Asahi Linux reports significant progress on Apple M3 hardware, with key peripherals working, and DisplayPort Alt Mode via USB-C now functional in an experimental branch after years of reverse engineering, though a general M3 release awaits GPU driver work and speaker support. <a href="https://app.daily.dev/posts/9Q140qigi" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Deno Deploy introduces instant Linux microVMs</strong>: Deno Deploy launched <code>deno sandbox</code>, a new subcommand that provisions cloud-hosted Linux microVMs and SSH-connects in under a second, featuring a secrets mechanism that injects values at the network level without exposing raw secrets to sandbox code. <a href="https://app.daily.dev/posts/7Ar3LQ3uM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Tailscale Peer Relays now generally available</strong>: Tailscale Peer Relays are now generally available, offering customer-deployed, high-throughput relay nodes on any Tailscale device with improved performance, static endpoint support for restrictive cloud environments, and deeper observability, available on all plans including the free tier. <a href="https://app.daily.dev/posts/QRFLXVuPF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>KDE Plasma 6.6 released with new features</strong>: KDE Plasma 6.6 introduces OCR functionality in Spectacle, a new Plasma Keyboard, a first-run wizard, grayscale color blindness filter, smoother animations for high-refresh-rate displays, Wi-Fi QR code scanning, and automatic brightness adjustment. <a href="https://app.daily.dev/posts/klCS43HPr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft Project Silica develops glass data storage</strong>: Microsoft’s Project Silica is developing glass-based data storage technology using laser-etched glass chips to preserve information for millennia, aiming to store data equivalent to 2 million books in a single chip with extreme durability. <a href="https://app.daily.dev/posts/qmONeI2Gr" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic’s C compiler: AI built from scratch, or just really good prompt engineering?</strong>: Anthropic is touting Claude building a C compiler ‘from scratch’ with ‘no human intervention,’ capable of compiling Linux and Doom. What they conveniently gloss over is the $20,000 in API calls, 37 years of GCC test suites, human restarts, and the fact it couldn’t actually boot Linux. It’s a great achievement for AI agents, 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>AI productivity gains: the emperor has no clothes</strong>: A survey of nearly 6,000 executives found over 80% reported no measurable impact from AI on productivity or employment, with 89% seeing no gains in the last three years. Yet, they anticipate AI will eliminate 1.75 million jobs by 2028. So, AI isn’t helping 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 Sonnet 4.6: cheaper tokens, more tokens, same old story</strong>: Anthropic claims its new Sonnet 4.6 matches or exceeds Opus 4.6 performance at a fraction of the price. Sounds great, right? Except independent benchmarks found Sonnet 4.6 uses over 4x more tokens than its predecessor and more than Opus 4.6 due to ‘adaptive thinking.’ So, you’re paying less per token, but you’re using a whole lot more tokens. Classic AI economics. <a href="https://app.daily.dev/feed-by-ids?id=3TzVcx3Vs&amp;id=i4O6sQBIw" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
