<h2>Must know</h2>
<ul>
<li><strong>Google’s embedding model deprecation causes vendor lock-in</strong>: Google just pulled the plug on its <code>text-embedding-004</code> model, effectively bricking thousands of stored vectors for a production RAG-based movie recommender. This highlights a nasty vendor lock-in problem: different embedding models create incompatible vector spaces, making migration a nightmare. The takeaway? Treat your embedding model like critical infrastructure and design for change, or face the consequences of Google’s whims. <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 rolled out Lakebase, a new PostgreSQL-compatible database specifically for AI workloads. It aims to merge operational database features with their lakehouse architecture, offering a serverless OLTP database that separates compute from storage. With support for Postgres 17, pgvector, and integration with the Databricks Data Intelligence Platform, it’s clearly targeting real-time ML feature serving and AI agent memory, currently GA on AWS with Azure in preview. <a href="https://app.daily.dev/posts/56RHAShlB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux kernel 7.0 rc1 and the AI content deluge</strong>: The Linux kernel 7.0 rc1 is out, and Linus Torvalds says the version bump is just because he gets confused by big numbers, not a major feature. Still, it brings cemented Rust support, faster cache clearing, non-disruptive kernel updates, and updated AMD/Intel silicon support. Meanwhile, the Lobsters community is pushing for an ‘AI generated’ flag to combat the flood of low-effort, LLM-authored content, and maintainers like Jeff Geerling are drowning in AI-generated pull requests, forcing GitHub to offer options to restrict them. <a href="https://app.daily.dev/feed-by-ids?id=cYP8biu57&amp;id=LrCmvlStu&amp;id=UT5jXEM9J" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Developer productivity lost to non-coding tasks</strong>: A developer’s 10-second code fix took 10 hours due to non-coding tasks like fighting sandboxed environments, CUDA conflicts, and build system issues, highlighting that LLM tools don’t help with these bottlenecks. <a href="https://app.daily.dev/posts/OBIqfWI7x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Fresh file explorer extension for VS Code</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 and one-click restore for deleted files. <a href="https://app.daily.dev/posts/pQedWHBRB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rust RwLock underperforms mutex in read-heavy workloads</strong>: A hands-on comparison found Rust’s <code>RwLock</code> performed 5x worse than <code>Mutex</code> for a read-heavy LRU cache due to cache-line ping-pong, 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>Uber and OpenAI reveal infrastructure traffic control</strong>: Uber’s “Global Rate Limiter” and OpenAI’s “Access Engine” are new infrastructure-level traffic control platforms, with Uber’s handling 80 million requests per second using “probabilistic shedding.” <a href="https://app.daily.dev/posts/OVoWP2oNj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Shuru: local-first microVM sandbox for AI agents</strong>: Shuru is a new local-first microVM sandbox for AI agent workloads on macOS, using Apple’s Virtualization.framework to provide secure, ephemeral, and reproducible Linux environments without Docker. <a href="https://app.daily.dev/posts/TqgsBjEoV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong><a href="http://Lobste.rs" target="_blank" rel="noopener nofollow">Lobste.rs</a> attempts sqlite migration</strong>: The <code>lobste.rs</code> community attempted a live migration from MariaDB to SQLite but quickly reverted due to severe performance degradation, though they plan to retry after addressing the issues. <a href="https://app.daily.dev/posts/RLQeRSeL0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Phishing attack on X account</strong>: A developer lost access to their X account due to a sophisticated phishing attack, highlighting the importance of scrutinizing sender addresses, checking URLs, and using authenticator apps for 2FA. <a href="https://app.daily.dev/posts/cFsXn3pdx" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic’s claude compiler video: pure marketing magic</strong>: Anthropic released a marketing video showing Claude autonomously building a C compiler capable of compiling Linux and Doom. Critics are calling it “dishonest framing,” suggesting the real goal was a stress test of Claude’s autonomy, not a straightforward compiler-building demo. Because, you know, AI building a compiler is totally believable without any human intervention. <a href="https://app.daily.dev/posts/vGYk0DbTO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Gemini 3.1 pro: benchmark king, real-world dud</strong>: Google’s Gemini 3.1 Pro, despite its impressive benchmark scores, apparently “underperformed” Claude Opus 4.6 in real-world coding and UI tasks. The article’s title, “Gemini 3.1 PRO…pure disappointment,” really drives home the point. It’s almost like benchmarks don’t always translate to actual utility, who would’ve thought? <a href="https://app.daily.dev/posts/bIAFUawfa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Neiry’s bio-drones: pigeons with cameras and ‘privacy compliance’</strong>: Neiry, a deeptech company, is testing “bio-drones”—pigeons with neural interfaces, solar panels, and cameras—to guide them along preset routes. They claim “privacy compliance” by filtering identifiable details. Because nothing says privacy like a pigeon with a brain implant and a camera flying over your house, right? And they’re targeting Brazil and India, because why not start with the largest populations? <a href="https://app.daily.dev/posts/m7v8MA0Dh" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
