<h2>Must know</h2>
<ul>
<li><strong>AI coding agents are exposing cloud infrastructure weaknesses</strong>: Autonomous AI agents, operating at machine speed, are revealing critical shortcomings in current cloud infrastructure, particularly in networking segmentation, identity security, and cost controls. According to InfoWorld, traditional cloud setups lack the fine-grained policies and real-time monitoring needed to prevent agents from causing outages or runaway expenses. This necessitates a fundamental shift in governance and architecture, rather than treating AI as a mere incremental addition. <a href="https://app.daily.dev/feed-by-ids?id=1w09x1vo7&amp;id=1w09x1vo7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google is restricting its programmable search engine</strong>: Google is restructuring its Programmable Search Engine, effective January 1, 2027. The free version will be limited to 50 domains, forcing users needing broader search capabilities to migrate to Vertex AI Search or contact Google for custom solutions. This change, announced on Google’s Programmable Search Engine Blog, impacts webmasters and developers, with no explanation provided for the strategic shift. <a href="https://app.daily.dev/posts/DC2sDltji" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft gave BitLocker keys to the FBI</strong>: In a first publicly known instance, Microsoft provided BitLocker recovery keys to the FBI to decrypt three laptops during a fraud investigation. This occurred under a legal warrant, leveraging the default setting where BitLocker keys are uploaded to Microsoft’s cloud. Security experts warn this practice, while aiding data recovery, increases privacy risks and potential exposure if Microsoft’s cloud is compromised. <a href="https://app.daily.dev/posts/6Aq315WBk" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Vercel optimizes deployments with smarter function caching</strong>: Vercel has updated its deployment system to skip function uploads when code hasn’t changed, reducing build times by 400ms-5s by injecting deployment-specific environment variables at runtime instead of build time. This optimization is automatic for Python, Go, Ruby, Rust, and standalone Vercel Functions, with Next.js support coming soon. <a href="https://app.daily.dev/posts/sygra8Fax" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ESLint v10.0.0-rc.1 released with TypeScript improvements</strong>: ESLint v10.0.0-rc.1 is out for testing, featuring built-in TypeScript type definitions for Espree and ESLint Scope, replacing external Definitely Typed versions, along with fixes to the no-shadow rule and RuleTester improvements. <a href="https://app.daily.dev/posts/APrgmrJOW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 6.19 disables NEXT_BUDDY scheduler due to performance issues</strong>: The Linux 6.19 kernel is disabling its NEXT_BUDDY scheduler feature by default after performance regressions were observed in MySQL, SPECjbb, DayTrader, and Nginx HTTPS server workloads. <a href="https://app.daily.dev/posts/iDlOb4czh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PyView brings Phoenix LiveView to Python</strong>: PyView enables Python developers to build real-time web applications without client-side JavaScript, using a server-driven architecture that renders HTML on the server and pushes only changed DOM fragments via WebSockets. <a href="https://app.daily.dev/posts/2huH9oN7d" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rails 8.1.2 released with performance and stability fixes</strong>: Rails 8.1.2 is out, featuring a 2x performance boost for String#squish, stream error reporting to Rails.error API, and improved console reload! functionality, alongside various bug fixes and updates. <a href="https://app.daily.dev/feed-by-ids?id=Ji6vBmphI&amp;id=KjzktGkL0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NetBox Labs releases AI coding agent best practices</strong>: NetBox Labs has published a best practices repository for AI coding agents like Claude Code and Cursor, offering guidance on authentication, API usage, and integration patterns to help engineers build better NetBox integrations. <a href="https://app.daily.dev/posts/oJ6EYnP2o" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic introduces Agent Skills for specialized AI work</strong>: Anthropic’s new Agent Skills paradigm extends AI agent capabilities by packaging domain expertise as organized collections of files, enabling rapid deployment across various verticals and supporting executable scripts as tools. <a href="https://app.daily.dev/posts/lvPOiATWV" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI will make you a 10x developer, just not in enterprises</strong>: Apparently, AI coding tools are a godsend for small teams with shiny new tech stacks, delivering ‘significant speed boosts.’ But if you’re stuck in an enterprise with a ‘legacy codebase,’ AI-generated code just dumps more context-less garbage into your already overflowing technical debt pile. So, AI is great, just not where most of the actual work happens. Got it. <a href="https://app.daily.dev/feed-by-ids?id=yJ8WtVNig&amp;id=yJ8WtVNig" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is the new abstraction layer, just don’t forget the fundamentals</strong>: We’re told AI is the ‘latest abstraction layer’ in software development, letting us build ‘more sophisticated applications’ with ‘less code.’ But don’t get too excited, because you still need ‘fundamental coding principles’ to make ‘architectural decisions’ and ‘oversee AI-generated code.’ So, AI abstracts away the grunt work, but you still need to be smart enough to tell it when it’s wrong. Revolutionary! <a href="https://app.daily.dev/posts/XUOoVy4Ig" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding tools are great, if you’re already an expert</strong>: Anthropic’s latest report confirms what we all suspected: ‘prompt sophistication’ correlates almost perfectly with ‘response quality’ from Claude. Turns out, AI doesn’t ‘level the playing field’; it just ‘multiplies’ the expertise of those who already have it. So, if you’re smart, AI makes you smarter. If you’re not, well, keep trying those ‘surface-level prompt tricks.’ <a href="https://app.daily.dev/feed-by-ids?id=O5Ad5zdYS&amp;id=O5Ad5zdYS" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
