<h2>Must know</h2>
<ul>
<li><strong>AI agents are changing how we build software</strong>: AI agents are rapidly integrating into development workflows, with Apple’s Xcode 26.3 now natively supporting Claude Code/Codex and Google launching a Developer Knowledge API to feed AI tools up-to-date documentation. This shift is enabling new paradigms like Netlify’s ‘Agent Runners’ for AI-powered deployment, allowing non-technical users to make changes via AI prompts. However, this also brings concerns about security, with critical flaws found in Anthropic’s Claude Desktop Extensions allowing remote code execution, and a ‘lethal trifecta’ vulnerability identified when agents handle untrusted data, access secrets, and communicate externally. <a href="https://app.daily.dev/feed-by-ids?id=jUcS4YRSW&amp;id=AVkBPwGIz&amp;id=7UymufP7l&amp;id=roR6t4ZAo&amp;id=3BxyRjmYn&amp;id=4NFnRjHAX&amp;id=Ik4RG9gCW&amp;id=GfBuGFYT2&amp;id=yRw17CGPv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Heroku is officially dead, Salesforce pivots to AI</strong>: Salesforce has moved Heroku into a ‘sustaining engineering’ phase, halting new enterprise contract sales and effectively signaling its end-of-life. This move is interpreted as Salesforce pivoting its strategy towards AI-led growth, leaving existing Heroku customers to treat it as legacy infrastructure and plan migrations. This follows a trend of cloud platforms adapting to AI workloads, as seen with Google’s massive $185 billion investment in AI infrastructure and data centers, and CNCF data showing Kubernetes evolving for AI. <a href="https://app.daily.dev/feed-by-ids?id=13PGZg8yI&amp;id=smfFrov84&amp;id=832MyHBnu&amp;id=iGOyLrEVQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux kernel permanently adopts Rust</strong>: Linux 7.0 officially concludes the ‘Rust experiment’ by making Rust a permanent part of the kernel project. This decision signals a long-term commitment to the language, encouraging investment in Rust training and development within the Linux ecosystem. This comes as GNU Linux-libre 6.19 also released, focusing on removing proprietary firmware dependencies from the mainline kernel, further emphasizing open-source principles. <a href="https://app.daily.dev/feed-by-ids?id=k0qXAqGCH&amp;id=HW1jv2rLq" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Transformers.js v4 brings WebGPU and major speedups</strong>: Transformers.js v4 preview is now available on NPM, featuring a new C++ WebGPU runtime for cross-platform support, delivering 4x speedup for BERT models and 10x faster builds. <a href="https://app.daily.dev/posts/blzYLzH95" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents develop C compiler for Linux kernel</strong>: Anthropic researchers used 16 Claude Opus 4.6 AI agents to build a Rust-based C compiler that successfully compiles the Linux kernel, though the generated code is not production-ready due to inefficiency. <a href="https://app.daily.dev/feed-by-ids?id=qhPl6UlEO&amp;id=9jEdtbewr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google Cloud Monitoring enhances OTLP metrics support</strong>: Google Cloud Monitoring now supports OpenTelemetry Protocol (OTLP) for metrics, enabling provider-agnostic telemetry pipelines and direct SDK-to-Cloud Monitoring ingestion. <a href="https://app.daily.dev/posts/REiFVQA5G" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>SwiftUI @Observable macro causes performance issues</strong>: The @Observable macro in SwiftUI can cause performance issues and race conditions due to ‘overobserving’ and asynchronous tasks directly mutating models, requiring careful architectural solutions. <a href="https://app.daily.dev/feed-by-ids?id=bOK3npOxM&amp;id=iwu0LXDGt" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New DNS-PERSIST-01 validation method simplifies certificates</strong>: DNS-PERSIST-01, a new domain control validation method, will allow Certificate Authorities to issue certificates based on a single, static DNS record, significantly reducing complexity for certificate automation. <a href="https://app.daily.dev/posts/7hFHkMPwg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Valkey performance enhancements post-Redis fork</strong>: Valkey, a fork of Redis, is undergoing significant performance improvements in versions 8 and 9, including a redesigned hash table architecture and reduced memory overhead, while maintaining backwards compatibility with Redis 7.2. <a href="https://app.daily.dev/posts/BJPWCn9Zl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub Apps get fine-grained permissions for Enterprise Teams APIs</strong>: GitHub Apps can now access Enterprise Teams API endpoints using fine-grained permissions, allowing administrators to programmatically manage teams at scale with improved security. <a href="https://app.daily.dev/posts/5IGPhAfLF" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic’s Claude Co-work plugin ‘wipes out’ $400 billion in tech stocks</strong>: Anthropic released Claude Co-work with industry-specific plugins, and according to one uncorroborated source, this ‘wiped out $400 billion in tech stocks.’ Apparently, the mere existence of an AI assistant with plugins is enough to send the entire SaaS market into a tailspin. Because, you know, a single product update from one company totally dictates the valuation of an entire industry. Sure, Jan. <a href="https://app.daily.dev/posts/QnkkrfiEd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated code: 96% don’t trust it, 48% don’t verify it</strong>: A survey found that 96% of engineers don’t fully trust AI-generated code, yet only 48% bother to verify it before committing. This is peak efficiency, folks: distrusting your tools but using them blindly anyway. What could possibly go wrong when half the developers are just shrugging and pushing potentially buggy, insecure AI output straight to production? It’s not like code quality or security matters anymore, right? <a href="https://app.daily.dev/posts/QxTS3kJG9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI makes developers 56% faster and 19% slower, simultaneously</strong>: Research on AI coding assistants shows contradictory results: GitHub claims 56% faster task completion, while another study indicates 19% slower overall times due to new overhead like reviewing AI output and prompting. So, AI makes you both faster and slower at the same time. It’s a quantum state of productivity, where you’re simultaneously crushing deadlines and missing them. Just another day in the AI hype cycle, where reality is optional. <a href="https://app.daily.dev/posts/t4jX0rY0U" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
