<h2>Must know</h2>
<ul>
<li><strong>The collapse of MkDocs and the rise of Zensical</strong>: The MkDocs ecosystem, a staple for over 90,000 GitHub projects, is fracturing after the original founder returned from an eight-year hiatus to push a plugin-incompatible rewrite. This unilateral move sparked a community revolt and a brief PyPI takeover attempt by former maintainers. The community has now split into three competing successors, with Zensical, a ground-up rewrite by the Material for MkDocs team, currently leading in developer mindshare. <a href="https://app.daily.dev/feed-by-ids?id=GedRmjUpE&amp;id=ZimmvE6Mp" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GrapheneOS defies global age verification mandates</strong>: Privacy-focused GrapheneOS has publicly refused to comply with new age verification laws in Brazil and California, even at the cost of being banned from those markets. These regulations require operating systems to collect user age data and provide API signals to apps, which critics argue creates a permanent surveillance infrastructure. This hardline stance complicates GrapheneOS’s upcoming hardware partnership with Motorola, as the OS team prioritizes anonymity over commercial expansion. <a href="https://app.daily.dev/feed-by-ids?id=3oUfmyp1R&amp;id=f2c1Srdfh&amp;id=u4yVQPhrL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Flash MoE brings 397B parameter models to MacBooks</strong>: A new inference engine called Flash-MoE has demonstrated the ability to run the massive Qwen3.5-397B model on a standard MacBook Pro with only 48GB of RAM. By utilizing a pure C and Metal architecture that streams experts directly from SSD using parallel pread calls, it achieves usable speeds of over 4 tokens per second. This development, detailed on GitHub, suggests that the hardware barrier for running frontier-level Mixture-of-Experts models is collapsing faster than anticipated. <a href="https://app.daily.dev/posts/xsGaEEFUf" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Linux 7.0 rc5 arrives with fixes for legacy Radeon GPUs and Logitech peripherals</strong>: Linus Torvalds notes the cycle is calming down ahead of a stable April release. <a href="https://app.daily.dev/feed-by-ids?id=1tWnyX7nj&amp;id=JTYKWHg4j" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vercel Workflow adds automatic end to end encryption for all run data</strong>: The system uses per-deployment unique keys and AES-256-GCM to secure step arguments and return values. <a href="https://app.daily.dev/posts/0r6jV9CyY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Project Detroit reopens to integrate JavaScript and Python into the JVM</strong>: The initiative leverages the Foreign Function and Memory API to allow seamless polyglot execution on the Java runtime. <a href="https://app.daily.dev/posts/24VxbiILg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cilium hits ten year milestone with sixty percent CNI market share</strong>: The project is now positioning itself as the primary networking data plane for large-scale AI training clusters. <a href="https://app.daily.dev/posts/rtti5mtPR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Deno faces community scrutiny following mass layoffs and product pivots</strong>: Early investors are questioning the runtime’s trajectory after the lukewarm reception of JSR and Deno Deploy. <a href="https://app.daily.dev/posts/ONnCa8fkU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Aurora DSQL adds a browser based playground for serverless Postgres testing</strong>: The update includes new integrations for Prisma and VS Code alongside AI-assisted schema design tools. <a href="https://app.daily.dev/posts/Un5GpVc7x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>A developer turned a Logitech mouse DPI register into persistent data storage</strong>: The Rust-based hack exploits the HID++ protocol to store two bytes of cross-computer persistent data on the hardware. <a href="https://app.daily.dev/posts/uG0HL5oPv" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The AI agent security theater</strong>: We are currently being flooded with ‘AI-native’ security suites like Booz Allen’s Vellox that promise ‘machine-speed’ defense. In reality, unsupervised agents are already leaking secrets to NPM registries and falling for basic prompt injections, as highlighted at QCon London. Calling a tool ‘autonomous’ doesn’t fix the fact that we’re just giving LLMs more ways to fail at scale with higher permissions. <a href="https://app.daily.dev/feed-by-ids?id=Ur5RHzsG2&amp;id=lljljIGzG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The death of coding as an art form</strong>: A wave of philosophical hand-wringing is mourning the ‘diminished art’ of coding, claiming AI is turning software into ‘fast fashion.’ This narrative conveniently ignores that most enterprise CRUD apps were never ‘art’ to begin with; they were just expensive ways to move JSON around. If your sense of ‘craft’ is threatened by a glorified autocomplete, maybe the craft was actually just the friends we made along the way. <a href="https://app.daily.dev/feed-by-ids?id=AKKujKavD&amp;id=EaqYqE4h6" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vibe coding your way into a DevOps talent shortage</strong>: Marketing fluff is now claiming that APAC startups are ‘bypassing’ DevOps teams by using managed platforms like Vercel to scale AI agents. It’s a classic rebranding of vendor lock-in as a ‘strategic choice’ to solve a talent shortage that the platforms themselves helped create. You haven’t ‘bypassed’ DevOps; you’ve just outsourced your infrastructure’s soul to a monthly subscription you can’t easily leave. <a href="https://app.daily.dev/posts/s59UHWEkP" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
