<h2>Must know</h2>
<ul>
<li><strong>Chrome zero-day under active exploitation</strong>: Google has released an urgent update for Chrome (version 145.0.7632.75/76) 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 the year and is already being actively exploited in the wild. The bug, written in C++, can lead to arbitrary code execution within the browser sandbox, making immediate updates critical for all users. <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 has disclosed critical vulnerabilities in four widely used VS Code extensions: Live Server, Code Runner, Markdown Preview Enhanced, and Microsoft’s Live Preview, collectively impacting over 128 million installations. These flaws enable file theft, remote code execution, and local network reconnaissance. While Microsoft silently patched Live Preview, the maintainers of the other three extensions have not responded, leaving millions of developers exposed. Developers are advised to update Live Preview and disable unused extensions. <a href="https://app.daily.dev/posts/IuDICegHN" 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. This flaw enabled Copilot Chat to summarize emails from users’ Sent Items and Drafts folders without authorization, active since late January. While a fix is rolling out, this incident adds to previous Copilot security concerns and highlights broader enterprise anxiety about AI tools accessing sensitive corporate data, especially given the European Parliament’s recent ban on similar features. <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>AI model performance and cost analysis</strong>: Anthropic’s new Claude Sonnet 4.6, marketed as 40% cheaper, is identified by independent benchmarks as a ‘token muncher’ due to its adaptive thinking, potentially negating per-token cost savings for heavy use cases. <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 agent development, security, and governance accelerate</strong>: Enterprises are rapidly adopting agentic AI, with new tools like StrongDM ID providing identity for AI agents and Cencurity offering security gateways to protect against data leakage and risky code patterns. <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>AI’s impact on software development and engineering roles</strong>: Spotify’s CEO claims senior developers haven’t written code since December 2025, instead supervising AI-generated code, while Martin Fowler warns that AI can accelerate existing strengths and weaknesses, potentially creating ‘debt accelerators’ in unhealthy codebases. <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>New developer tools for cloud deployment and streaming</strong>: Deno Deploy introduced <code>deno sandbox</code> for instant Linux microVM provisioning, TinyIce offers a lightweight Icecast2-compatible audio streaming server, and Konta brings GitOps workflows to Docker Compose on low-resource VPS environments. <a href="https://app.daily.dev/feed-by-ids?id=7Ar3LQ3uM&amp;id=BHwbWnzAf&amp;id=QRFLXVuPF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Tailscale peer relays general availability</strong>: Tailscale has made Peer Relays generally available, 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>Tailscale integrates with huntress and fleet for automated device posture enforcement</strong>: Tailscale now integrates with Huntress EDR and Fleet MDM, enabling device posture attributes and management state to be used in Tailscale access policies, automatically restricting access for non-compliant devices. <a href="https://app.daily.dev/feed-by-ids?id=L7iJSIIDS&amp;id=VukD3HNtV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks launches custom agents</strong>: Databricks has 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 CI/CD integration. <a href="https://app.daily.dev/posts/QJn2roYcs" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI productivity impact: much ado about nothing?</strong>: A survey of nearly 6,000 executives found that over 80% of firms 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 doing much now, but it’s definitely coming for your job later. Got it. <a href="https://app.daily.dev/posts/rfmYnNc4e" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic’s C compiler: marketing hype vs reality</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, 16 AI agents autonomously running for two weeks to produce a 100,000-line functional compiler is still pretty neat, if only they’d been honest about it. <a href="https://app.daily.dev/feed-by-ids?id=0csgEkRMy&amp;id=1CyjIJfh9&amp;id=6spQIGtcM&amp;id=dhghGjegM&amp;id=poO8tUjKL&amp;id=v3dp2qPCw&amp;id=vTvxCCQVI&amp;id=z9Rc6eFF4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated code: the new ‘instant legacy’</strong>: Stack Overflow’s 2025 survey shows AI tool usage at 84% but developer trust dropped to 29%, with main branch success rates at a five-year low. Microsoft even republished an AI-generated, error-ridden Git branching diagram. Apparently, AI makes code generation effortless, but it’s just accelerating the creation of ‘instant legacy code’ that’s hard to understand or debug. Progress! <a href="https://app.daily.dev/feed-by-ids?id=alUW9sJXE&amp;id=guYQJk0m5&amp;id=okTNHJRhB&amp;id=sc19eWuhU&amp;id=vbdTHZjVN&amp;id=xQwTKeb7K&amp;id=ySpIGlBP7&amp;id=ygeEuk5Vz" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
