<h2>Must know</h2>
<ul>
<li><strong>Axios and LiteLLM hit by sophisticated supply chain attacks</strong>: A major security breach has compromised the Axios npm package (versions 1.14.1 and 0.30.4) and the LiteLLM Python library, deploying a cross-platform Remote Access Trojan linked to North Korean threat actor UNC1069. The attack exploited a hijacked maintainer account and compromised GitHub Actions credentials to inject malicious postinstall scripts that exfiltrate data within seconds of installation. Security researchers from Unit 42 and Google recommend pinning Axios to version 1.14.0 or earlier and auditing all CI/CD pipelines for unauthorized credential use. <a href="https://app.daily.dev/feed-by-ids?id=0Uj2ass08&amp;id=AZHl01Fi2&amp;id=zQFdPLgVT&amp;id=Y2vRayEM1&amp;id=YKPH3yXt5&amp;id=dfUjvXabl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare Workers launch dynamic V8 isolate loader</strong>: Cloudflare has entered open beta for its Dynamic Worker Loader, a feature designed specifically for executing AI-generated code within secure V8 isolates. By moving away from traditional containerization, Cloudflare claims 100x faster boot times and significantly better memory efficiency for short-lived agentic tasks. The system integrates with their new ‘Code Mode’ concept, which uses typed TypeScript APIs to reduce LLM token consumption by up to 81% compared to sequential tool calling. <a href="https://app.daily.dev/posts/MtPJm1761" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Raspberry Pi prices spike amid global LPDDR4 shortage</strong>: Raspberry Pi has implemented significant price hikes across its entire lineup, with the 16GB Pi 5 now reaching $299.99 due to a sevenfold increase in LPDDR4 DRAM costs. Founder Eben Upton attributed the shortage to major manufacturers like Samsung and SK Hynix prioritizing high-margin AI data center supply over consumer-grade memory. To mitigate the impact on hobbyists, a new 3GB variant of the Pi 4 has been introduced as a lower-cost entry point at $83.75. <a href="https://app.daily.dev/posts/lQotqYWmF" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>CaP-X framework brings agentic AI to physical robotics</strong>: The team behind Voyager released an open-source toolkit for LLM-based control of robot arms and humanoids, featuring auto-synthesized skill libraries. <a href="https://app.daily.dev/posts/wBCnRfhCg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Zerobox provides lightweight cross-platform process sandboxing</strong>: Built in Rust, this tool uses macOS Seatbelt and Linux Bubblewrap to wrap any command with strict file, network, and credential controls. <a href="https://app.daily.dev/posts/GdAmTRjL8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ESLint v10 finalizes the transition to flat config</strong>: The latest major release removes the legacy eslintrc system entirely and improves JSX identifier tracking to reduce false positives in monorepos. <a href="https://app.daily.dev/posts/PY1Nl23uV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Arch Installer 4.0 migrates to Textual TUI framework</strong>: The overhaul replaces the old curses library with a persistent async Python UI, preventing freezes during background installation tasks. <a href="https://app.daily.dev/posts/5UR81e2PY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>React Native WebGPU enables direct Metal and Vulkan access</strong>: A new library allows React Native apps to perform parallel GPU computations and on-device ML inference using modern Compute Shaders. <a href="https://app.daily.dev/posts/PszBtntIw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Debezium 3.5 introduces parallel multithreaded snapshots</strong>: The update significantly speeds up initial data loads and adds Oracle 26ai compatibility alongside improved transaction metadata for SQL Server. <a href="https://app.daily.dev/posts/frQ51l2Un" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Wayland Protocols 1.48 adds long-awaited session management</strong>: The new XDG Session Management protocol allows clients to restore window positioning and state across sessions. <a href="https://app.daily.dev/posts/60DR37TXb" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>NVIDIA’s LLM-generated Linux driver ‘preview’</strong>: NVIDIA claims their latest preview driver for the DRM color pipeline was written ‘nearly entirely’ by Claude Sonnet and Opus. While they insist it was under ‘strong human direction,’ the April 1st release date suggests this is just a high-effort prank to see which tech journalists will swallow the ‘AI wrote my kernel driver’ narrative. If true, we look forward to the first AI-generated kernel panic being blamed on a hallucinated pointer. <a href="https://app.daily.dev/posts/z6fJhldHb" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The ‘vibe coding’ productivity trap</strong>: A recent Scott Logic experiment claims AI ‘doubled’ development velocity, yet developers admitted they lost the ‘joy of craft’ and struggled with basic tasks when the AI was removed. We’re essentially celebrating the fact that we can generate 1.9x more code that we don’t fully understand, creating a generation of ‘senior’ engineers who are effectively just highly-paid prompt editors. It’s not productivity; it’s just high-velocity technical debt generation. <a href="https://app.daily.dev/feed-by-ids?id=4U6dgQJaX&amp;id=2lJyJCSw3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Perplexity’s $1.6M labor saving ‘miracle’</strong>: Perplexity AI is bragging that its enterprise platform completed ‘3.25 years of human work’ in just four weeks. This metric is pure marketing fiction that assumes every token generated is equivalent to a minute of human thought, ignoring the massive overhead of verification and the inevitable cleanup of autonomous ‘hallucinations.’ It’s the same inflated automation promise we’ve seen for decades, just rebranded with a more expensive API key. <a href="https://app.daily.dev/feed-by-ids?id=FOKINOtLJ&amp;id=Teg1nySKy" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
