<h2>Must know</h2>
<ul>
<li><strong>Databricks launches new database for AI workloads</strong>: Databricks just dropped Lakebase, a PostgreSQL-compatible database built specifically for AI workloads. It’s designed to merge operational database features with their existing data lakehouse architecture, offering serverless OLTP, instant data branching, and unified access controls. This move aims to simplify real-time ML feature serving and AI agent memory, and it’s already generally available on AWS, with Azure support in preview. <a href="https://app.daily.dev/posts/56RHAShlB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google’s embedding model deprecation causes headaches</strong>: Google deprecated its <code>text-embedding-004</code> model, effectively bricking thousands of stored vectors for a production RAG-based recommender system. The core issue is that different embedding models create incompatible vector spaces, making migration a nightmare and highlighting a hidden form of vendor lock-in. The takeaway: if you’re persisting embeddings at scale, treat your embedding model as critical infrastructure and plan for inevitable changes. <a href="https://app.daily.dev/posts/xhDI1zj4i" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI’s dual impact on productivity and jobs</strong>: AI and agentic tools are still in their infancy, requiring significant human oversight, yet ignoring them is a risk due to market pressures. While some, like OpenAI’s Codex team, are deeply integrating AI for massive productivity gains, others find limited benefits due to bottlenecks in deployment pipelines. There’s also growing concern about AI-generated ‘slop’ overwhelming content and the potential for widespread white-collar job displacement, creating a non-cyclical feedback loop of continuous disruption. <a href="https://app.daily.dev/feed-by-ids?id=4gPWtRfOI&amp;id=8TOu6Vf1U&amp;id=C3J7JVQAT&amp;id=KAYKDJGSY&amp;id=LrCmvlStu&amp;id=Lv2cFkNVA&amp;id=QwH8nwgz6&amp;id=UT5jXEM9J&amp;id=frjHTf2Ap&amp;id=iVsrjcoM3" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Linux kernel 7.0 rc1 released with Rust support</strong>: Linus Torvalds released Linux kernel 7.0 rc1, a version bump purely for numerical reasons, which includes cemented Rust support, faster cache clearing, and non-disruptive kernel updates. <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>New AI user experience patterns emerge</strong>: Beyond basic chat interfaces, new AI UX patterns like form filling, ghost text completion, proactive triggers, and modular memory are being developed to integrate LLMs more naturally into workflows. <a href="https://app.daily.dev/posts/Ex0Q52GRH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI and citizen science converge for scientific discovery</strong>: A new report highlights the grand challenges and strategic priorities for combining AI with citizen science to accelerate scientific discovery, emphasizing novel human-AI interaction models and trustworthy systems. <a href="https://app.daily.dev/posts/KpgU5LzgK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenClaw gateway offers AI agent orchestration</strong>: Mission Control is an open-source dashboard for orchestrating AI agents, built with Next.js 14 and TypeScript, allowing users to create tasks, plan with AI, and track agent progress on a Kanban board. <a href="https://app.daily.dev/posts/l0KLAjYGc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Developer productivity hindered by 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 failures, highlighting how abstraction layers and tooling overhead consume significant time. <a href="https://app.daily.dev/posts/OBIqfWI7x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New VS Code extension for file navigation</strong>: Fresh File Explorer is a new VS Code extension that replaces the standard file explorer, displaying recently modified files with a heatmap, Git history integration, 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>PostgreSQL vector search alternatives explored</strong>: A new post discusses how to implement <code>pgvector</code>-like vector search functionality within PostgreSQL without relying on the <code>pgvector</code> extension, suggesting a DIY approach to vector search capabilities. <a href="https://app.daily.dev/posts/vUwhYnpqm" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic’s compiler video: more marketing than magic</strong>: Anthropic released a marketing video showing Claude supposedly building a C compiler capable of compiling Linux and Doom. Critics are calling it ‘dishonest framing,’ suggesting the real goal was to stress-test Claude’s autonomy, not to showcase a straightforward compiler-building feat. So, another day, another AI demo that’s less about groundbreaking tech and more about generating buzz. <a href="https://app.daily.dev/posts/vGYk0DbTO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Gemini 3.1 pro: pure disappointment, apparently</strong>: Despite impressive benchmark scores, Google’s Gemini 3.1 Pro reportedly ‘underperformed’ in real-world coding and UI generation tasks compared to Claude Opus 4.6. The unnamed ‘hands-on comparison’ claims Gemini struggled with tool-calling failures and incomplete features, leading to a ‘pure disappointment’ for those expecting its benchmark prowess to translate to actual utility. Maybe those benchmarks aren’t quite as ‘real-world’ as they seem, eh? <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 designed to simulate deceased users’ 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. The patent describes its use when a user is ‘absent’ from the platform, which sounds less like a feature and more like a Black Mirror episode waiting to happen. <a href="https://app.daily.dev/posts/AeW0pHCcd" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
