<h2>Must know</h2>
<ul>
<li><strong>Google embedding model deprecation causes vendor lock-in</strong>: Google deprecated its <code>text-embedding-004</code> model, rendering thousands of stored vectors unusable for a production RAG-based movie recommender. This highlights a critical vendor lock-in issue: different embedding models produce incompatible vector spaces, making migration difficult. The author, writing on Medium, emphasizes that persisted embeddings couple data to a specific model’s geometry, not just an API, necessitating a design for change if persisting embeddings at scale. <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 engineered for AI workloads. This serverless OLTP database integrates with Databricks’ data lakehouse architecture, separating compute from storage and supporting features like instant data branching and pgvector. It’s currently GA on AWS, with Azure in public preview, aiming to bridge operational database needs with AI-driven analytics. <a href="https://app.daily.dev/posts/56RHAShlB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding agents lack innovative problem-solving</strong>: While AI coding agents like Claude Code demonstrate high productivity and speed, they fundamentally lack the ability to devise innovative solutions when facing obstacles. This limitation is a critical differentiator from human developers, suggesting that human developers who also lack creative problem-solving skills will struggle for relevance in the evolving landscape, according to a post on Medium. <a href="https://app.daily.dev/posts/hB7gw6iXo" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>OpenClaw Gateway offers AI agent orchestration dashboard</strong>: Mission Control is an open-source AI agent orchestration dashboard built with Next.js 14 and TypeScript, connecting to OpenClaw Gateway via WebSocket to create tasks, plan with AI, and manage specialized agents on a Kanban board. <a href="https://app.daily.dev/posts/l0KLAjYGc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Shuru provides local-first microVM sandbox for AI agents on macOS</strong>: Shuru is a new local-first microVM sandbox tool for macOS, utilizing Apple’s Virtualization.framework to run lightweight Linux VMs for secure, ephemeral, and reproducible environments for AI agent workloads without Docker. <a href="https://app.daily.dev/posts/TqgsBjEoV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux kernel 7.0 rc1 released with Rust support and performance boosts</strong>: Linus Torvalds released Linux kernel 7.0 rc1, which includes cemented Rust support, faster cache clearing, non-disruptive kernel updates, and performance improvements for RISC-V and LoongArch. <a href="https://app.daily.dev/posts/cYP8biu57" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New VS Code extension ‘Fresh File Explorer’ focuses on recent files</strong>: Fresh File Explorer is a new VS Code extension that replaces the standard file explorer, displaying recently modified files with Git history integration, color-coded recency, and a ‘Resurrect’ feature for deleted files. <a href="https://app.daily.dev/posts/pQedWHBRB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>JDK 24 introduces post-quantum cryptography and security enhancements</strong>: JDK 24 brings significant security enhancements, including post-quantum cryptography (ML-KEM, ML-DSA), a new KDF API, SHA-3 performance gains, and the permanent removal of the Security Manager. <a href="https://app.daily.dev/posts/TLN02dr7b" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rust RwLock can be slower than Mutex for read-heavy workloads</strong>: A hands-on investigation revealed that <code>RwLock</code> performed significantly worse than <code>Mutex</code> for a read-heavy LRU tensor cache in Rust due to cache-line ping-pong from atomic reader counter increments, 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>Database design should start with consequences, not tables</strong>: Good database design begins with understanding business domain consequences, advocating for normalizing to 3NF first, then denormalizing deliberately, and using constraints to embed business rules, as performance issues are often schema problems in disguise. <a href="https://app.daily.dev/posts/fI73pTtiz" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Gemini 3.1 Pro is a ‘pure disappointment’ in coding tasks</strong>: Despite strong benchmark scores, Google’s Gemini 3.1 Pro consistently underperformed Claude Opus 4.6 in real-world coding and UI generation tasks, according to a hands-on comparison. Apparently, it struggles with tool-calling failures, incomplete features, and complex prompts, leading to a ‘pure disappointment’ for anyone expecting actual utility beyond flat visual design. Maybe stick to benchmarks, Google. <a href="https://app.daily.dev/posts/bIAFUawfa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic’s Claude compiler video criticized for ‘dishonest framing’</strong>: Anthropic released a marketing video showcasing Claude autonomously building a C compiler capable of compiling Linux and Doom, but critics are calling foul, citing ‘dishonest framing.’ Apparently, the real goal was just a stress test of Claude’s autonomy, not a straightforward demonstration of compiler-building prowess. So, another day, another AI company trying to make their LLM sound more capable than it actually is. <a href="https://app.daily.dev/posts/vGYk0DbTO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Meta patents AI to simulate deceased users</strong>: Meta was granted a patent for an AI system that simulates deceased users’ social media activity, including posting and chatting, using an LLM trained on their past data. Because nothing says ‘ethical’ and ‘mentally healthy’ like a digital zombie of your loved ones posting on Facebook. Just what we needed, more ways for Meta to monetize our existence, even after we’re gone. <a href="https://app.daily.dev/posts/AeW0pHCcd" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
