<h2>Must know</h2>
<ul>
<li><strong>Databricks launches Lakebase for AI workloads</strong>: Databricks just dropped Lakebase, a new PostgreSQL-compatible database built for AI workloads. It’s a serverless OLTP database that separates compute from storage, integrating directly with their Data Intelligence Platform. This means transactional data can be queried by Spark and Databricks SQL without the usual ETL hassle, targeting real-time ML feature serving and AI agent memory. It’s GA on AWS, with Azure in preview, and GCP coming later this year. <a href="https://app.daily.dev/posts/56RHAShlB" 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. The core issue? Different embedding models create incompatible vector spaces, making migration a nightmare. This highlights a nasty form of vendor lock-in where your persisted data is tied to a specific model’s geometry, not just an API. The takeaway: treat embedding models as infrastructure and design for change from day one if you’re persisting embeddings at scale. <a href="https://app.daily.dev/posts/xhDI1zj4i" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux kernel 7.0 rc1 released, Linus muses on succession</strong>: The Linux kernel 7.0 rc1 is out, and Linus Torvalds says the major version bump is just because he gets confused by big numbers, not a huge feature. Still, it brings cemented Rust support, faster cache clearing, non-disruptive kernel updates, and updated AMD/Intel silicon support. Torvalds also mused about a future maintainer, referencing the community’s recently developed succession plan, which is always a fun thought experiment. <a href="https://app.daily.dev/posts/cYP8biu57" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>OpenClaw AI agent use cases and limitations</strong>: OpenClaw, an autonomous AI agent tool, offers use cases like automated React dependency maintenance, daily tech research briefings, and cloud cost monitoring, but struggles with Claude Code permission prompts. <a href="https://app.daily.dev/posts/WCj2dFu7Q" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Developer productivity lost to non-coding tasks</strong>: A developer’s 10-second code fix for <code>flash-attention</code> took over 10 hours due to non-coding tasks like fighting sandboxed environments, <code>CUDA</code> conflicts, and build system issues, highlighting that <code>LLM tools</code> don’t help with these common productivity sinks. <a href="https://app.daily.dev/posts/OBIqfWI7x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Shuru: local-first microVM sandbox for AI agents on macOS</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>VS Code extension ‘Fresh File Explorer’ released</strong>: A new VS Code extension, ‘Fresh File Explorer,’ replaces the standard file explorer to display recently modified files, integrating Git history and offering features like a recency heatmap and one-click ‘Resurrect’ 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 investigation found that <code>RwLock</code> performed five times 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 tiny critical sections. <a href="https://app.daily.dev/posts/rFMs5918e" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ubuntu 26.04 LTS to default to OpenJDK 25</strong>: Ubuntu 26.04 LTS will default to OpenJDK 25, aligning its LTS release cycle with OpenJDK and bringing improvements in memory efficiency, garbage collection, and new language features, while OpenJDK 21 remains available. <a href="https://app.daily.dev/posts/acjUVfZBa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>React confirm manager for async dialogs</strong>: The <code>react-confirm-manager</code> library offers promise-based, async, and customizable confirmation dialogs for React, allowing imperative triggering without JSX or hooks at call sites, and supporting scoped containers, animations, and custom rendering. <a href="https://app.daily.dev/posts/VascGpijE" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic’s Claude compiler video: more hype than reality?</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. So, another day, another AI company trying to make their LLM sound like Skynet. <a href="https://app.daily.dev/posts/vGYk0DbTO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Gemini 3.1 Pro underperforms in real-world coding</strong>: Google’s Gemini 3.1 Pro, despite strong benchmark scores, reportedly “underperformed” Claude Opus 4.6 in real-world coding and UI generation tasks, with issues like tool-calling failures and incomplete feature implementation. Apparently, benchmarks are just for show, and when it comes to actual work, Gemini is a “pure disappointment.” Who knew? <a href="https://app.daily.dev/posts/bIAFUawfa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Meta patents AI for deceased user simulation</strong>: Meta was granted a patent for an AI system that simulates a deceased user’s social media activity, including posting and chatting, using an LLM trained on their past data. Because nothing says “grief counseling” like a chatbot pretending to be your dead relative. Ethical concerns? Mental health concerns? Nah, just more engagement for the platform. <a href="https://app.daily.dev/posts/AeW0pHCcd" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
