<h2>Must know</h2>
<ul>
<li><strong>Meta releases Muse Spark multimodal reasoning model</strong>: Meta Superintelligence Labs has launched Muse Spark, a natively multimodal model built from scratch that introduces a ‘Contemplating’ reasoning mode. According to Meta, the model uses parallel sub-agents and a ‘thought compression’ technique to achieve 10x compute efficiency over Llama 4 Maverick. While it leads in chart understanding and health benchmarks, it remains closed source for now, marking a shift from Meta’s traditional open-source strategy. <a href="https://app.daily.dev/posts/t1hgYSJW3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Chrome 147 introduces element scoped view transitions</strong>: Chrome Developers announced that version 147 now supports <code>Element.startViewTransition()</code>, allowing transitions on specific DOM subtrees rather than the entire document. This update enables multiple concurrent and nested transitions while keeping the rest of the page interactive, effectively solving layering issues with fixed-position content. The feature graduated to stable after extensive collaboration with the CSS Working Group to refine pseudo-element containment and overflow clipping. <a href="https://app.daily.dev/posts/upDe1yfse" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical Flatpak sandbox escape and host file access vulnerabilities</strong>: Flatpak 1.16.4 has been released to address three critical CVEs, including a complete sandbox escape (CVE-2026-34078) that allowed app-controlled symlinks to access arbitrary host paths. Another fix (CVE-2026-34079) prevents unauthorized file deletion on the host filesystem caused by unvalidated <code>ld.so</code> cache cleanup. Security researchers urge immediate updates via GitHub or system package managers to prevent potential full host code execution. <a href="https://app.daily.dev/posts/3itpGbR5u" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Railway migrates production frontend from Next.js to Vite and TanStack Router</strong>: The move reduced build times from 10 minutes to under 2 minutes by ditching Next.js server-first patterns for a more performant client-side layout model. <a href="https://app.daily.dev/posts/nCgXh3cQx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Docusaurus 3.10 stabilizes Faster features and adopts npm Trusted Publishing</strong>: The release marks Rspack and SWC integrations as stable and introduces a new experimental VCS API for significantly faster builds on large documentation sites. <a href="https://app.daily.dev/posts/jJ6FZ3fB4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Tailwind CSS 4.2 adds first class webpack plugin and muted palettes</strong>: The update eliminates manual PostCSS config for Webpack users and claims a 3.8x improvement in recompilation speeds across all integrations. <a href="https://app.daily.dev/posts/WyR2VLoTK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Espressif ESP32 S31 switches to dual core RISC V with WiFi 6</strong>: The new microcontroller departs from Xtensa architecture, adding a gigabit Ethernet MAC and 60 GPIO pins for high-performance IoT applications. <a href="https://app.daily.dev/posts/WXfCaFFQL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Storybook 10.3 introduces Model Context Protocol support for React</strong>: The integration allows AI agents to autonomously discover components, write stories, and run tests using the standardized MCP interface. <a href="https://app.daily.dev/posts/Q845rq2f6" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Malicious npm package @velora-dex/sdk delivers macOS backdoor</strong>: Version 9.4.1 executes a payload at import time that bypasses ignore-scripts to install a persistent RAT via launchctl; immediate rotation of credentials is required. <a href="https://app.daily.dev/posts/J4kaWYZmN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Atollic library brings Astro style islands to htmx applications</strong>: This new tool allows developers to hydrate Solid or React components within htmx-managed routes, solving state locality issues without a full framework rewrite. <a href="https://app.daily.dev/feed-by-ids?id=KbHNEMepg&amp;id=VmhloeOrm" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The AGI cancer cure PR shield</strong>: OpenAI is back at it, claiming AGI will magically cure cancer despite the minor inconveniences of missing medical data and the laws of biology. Satirical critiques point out that a chatbot can’t exactly speed up a five-year clinical trial or simulate human protein folding at scale. It’s a convenient narrative to distract from the fact that AI investment is actually cannibalizing the biotech funding needed for real research. <a href="https://app.daily.dev/posts/jvyfQ2Jpi" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic’s manufactured existential dread marketing</strong>: By claiming Claude Mythos is ‘too dangerous to release’ because it found a 27-year-old bug, Anthropic is following the classic GPT-2 hype playbook. Critics argue this anthropomorphizing of models as ‘escaping sandboxes’ is just a clever way to charge enterprise partners for ‘restricted’ access. If the model is truly a digital god of destruction, maybe don’t give it an email client and a Twitter account during ‘safety testing.’ <a href="https://app.daily.dev/feed-by-ids?id=EqYd854LH&amp;id=gSjjwv0bW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The $5 trillion Terafab pipe dream</strong>: Elon Musk’s new $25 billion joint venture with Intel aims for a terawatt of AI compute, but analysts at Bernstein suggest the actual bill is closer to $5 trillion. Between the mass exodus of xAI co-founders and a history of ‘optimistic’ timelines, this looks like another exercise in vertical integration theater. Intel joining as a foundry partner is a nice headline, but you can’t build a semiconductor empire on vibes and 1.8nm promises alone. <a href="https://app.daily.dev/posts/R7BXZGv2f" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
