<h2>Must know</h2>
<ul>
<li><strong>Linux kernel 7.0 rc1 released with succession commentary</strong>: Linus Torvalds dropped Linux kernel 7.0 rc1, not for any groundbreaking features, but because he apparently gets confused by big numbers. This release, however, does cement Rust support, brings faster cache clearing, and introduces non-disruptive kernel updates, alongside updated AMD/Intel silicon support and RISC-V/LoongArch improvements. Torvalds also mused about a future maintainer, hinting at the community’s recently developed succession plan. <a href="https://app.daily.dev/feed-by-ids?id=cYP8biu57&amp;id=8ssuMcdSU&amp;id=JXfUDc5F4&amp;id=QdVE1yqfx&amp;id=TcabAvodN&amp;id=cQun6gD6y&amp;id=zaDpYuaxU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google embedding model deprecation causes vendor lock-in</strong>: Google deprecated its <code>text-embedding-004</code> model, instantly bricking a production RAG-based movie recommender because different embedding models create incompatible vector spaces. This highlights a nasty form of vendor lock-in: your persisted embeddings are tied to a specific model’s geometry, not just an API. The fix involved re-embedding the entire dataset and a blue/green deployment, underscoring the need to treat embedding models as critical infrastructure and design for change. <a href="https://app.daily.dev/posts/xhDI1zj4i" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks launches lakebase for AI workloads</strong>: Databricks introduced Lakebase, a new PostgreSQL-compatible database specifically designed for AI workloads. It aims to merge operational database capabilities with their lakehouse architecture, offering a serverless OLTP database that separates compute from storage. Lakebase supports Postgres 17 with pgvector, instant data branching, and unified access controls, integrating directly with the Databricks Data Intelligence Platform for real-time ML feature serving and AI agent memory. <a href="https://app.daily.dev/posts/56RHAShlB" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>VS Code extension ‘fresh file explorer’ released</strong>: A new VS Code extension, Fresh File Explorer, replaces the standard file explorer to show recently modified files, integrating Git history and offering features like a recency heatmap, ‘Resurrect’ for deleted files, and pinning important items. <a href="https://app.daily.dev/posts/pQedWHBRB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI user experience patterns beyond chat interfaces</strong>: Atlassian proposes four practical UX patterns for embedding LLMs more naturally into workflows: AI Form Filling, Ghost Text Completion, Proactive Anticipatory Triggers, and Memory Modularity &amp; Workspace Context, moving beyond basic chat interfaces. <a href="https://app.daily.dev/posts/Ex0Q52GRH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>TSSLint 3.0 final major release with reduced dependencies</strong>: TSSLint 3.0, the final major version of the lightweight TypeScript semantic linting tool, removes the esbuild dependency, adds a TSL compatibility layer, and re-introduces <code>@tsslint/compat-eslint</code>, but is incompatible with the upcoming TypeScript 7 Go-based compiler. <a href="https://app.daily.dev/posts/314okFZnQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Redesigning a homepage in 20 minutes with figma mcp and ads mcp</strong>: A workflow chaining Figma MCP, Atlassian Design System MCP, and Rovo Dev CLI allowed a React application homepage to be redesigned in under 20 minutes, generating 1,381 lines of production-ready code and fixing accessibility issues. <a href="https://app.daily.dev/posts/ZWlO975fr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>SwiftUI list performance issues with geometryreader</strong>: Using <code>GeometryReader</code> inside every row of a SwiftUI <code>List</code> causes severe performance degradation due to fresh layout measurements on cell reuse, leading to CPU spikes and dropped frames during scrolling. <a href="https://app.daily.dev/posts/dGEGPWHMv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rust rwlock performs worse than mutex for read-heavy workloads</strong>: A hands-on investigation found that <code>RwLock</code> performed approximately 5 times worse than <code>Mutex</code> for a read-heavy LRU tensor cache in Rust due to cache-line ping-pong from atomic increments on read locks, suggesting <code>Mutex</code> might be better for very short critical sections. <a href="https://app.daily.dev/posts/rFMs5918e" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Virtual scrolling for billions of rows with hightable</strong>: HighTable, a React component, uses techniques like lazy loading, table slicing, infinite pixels, pixel-precise scroll, and two-step random access to render billions of rows in a browser table with high performance and accessibility, without fake scrollbars or canvas rendering. <a href="https://app.daily.dev/posts/GM4i1uj2n" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic’s claude compiler video: more marketing than magic</strong>: Anthropic released a marketing video showing Claude autonomously building a C compiler capable of compiling Linux, SQLite, Lua, and Doom. Critics, however, are calling out the ‘dishonest framing,’ suggesting the real goal was a stress test of Claude’s autonomous runtime, not a straightforward demonstration of compiler-building prowess. So, another day, another AI company trying to make their LLM sound like a wizard, when it’s probably just a very fast intern with a lot of boilerplate. <a href="https://app.daily.dev/posts/vGYk0DbTO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Meta patents ai for deceased user simulation: because grief wasn’t complicated enough</strong>: Meta was granted a patent for an AI system designed to simulate a user’s social media activity, including posting and chatting, using an LLM trained on their past data, for when a user is ‘absent’ from the platform. Because nothing says ‘moving on’ like your dead grandma’s AI bot still liking your posts. Clearly, Meta is prioritizing ‘engagement’ over, you know, basic human decency or the mental health implications of digital ghosts. <a href="https://app.daily.dev/posts/AeW0pHCcd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>China’s kung fu robots: the next big threat, or just a good show?</strong>: China showcased Unitree Robotics’ G1 humanoid robots doing martial arts at its Spring Festival Gala, with claims of high-speed cluster control and AI fusion localization. One article, with the subtle title ‘China’s Kung Fu Robots Just Changed Everything,’ suggests China is replicating its EV dominance playbook in robotics and that Western competitors will be left in the dust by 2028. Because a few backflips at a gala definitely mean global robotics supremacy, and not just a really impressive demo. <a href="https://app.daily.dev/feed-by-ids?id=UN40kBCVc&amp;id=hPdh434Wa" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
