<h2>Must know</h2>
<ul>
<li><strong>MongoDB vulnerability exposes sensitive data</strong>: A critical vulnerability, CVE-2025-14847 (MongoBleed), in MongoDB allows unauthenticated attackers to read sensitive data from heap memory due to improper handling of zlib-compressed traffic. All MongoDB versions since 2017 are affected, with active exploitation reported and an estimated 87,000 servers exposed worldwide. While MongoDB Atlas instances are patched, self-hosted deployments require immediate manual updates, as highlighted by security researchers. <a href="https://app.daily.dev/posts/FYcryJBfH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI’s ChatGPT health raises privacy alarms</strong>: OpenAI has launched ChatGPT Health in beta, allowing users to integrate medical records for personalized insights, partnering with b.well Connected Health. However, privacy experts are concerned the service is not HIPAA-covered, potentially commoditizing user data for healthcare and insurance companies. The service is unavailable in the EU/UK due to privacy law constraints, raising questions about OpenAI’s aggressive monetization strategy amid projected financial losses, according to various reports. <a href="https://app.daily.dev/feed-by-ids?id=0lod4TIIq&amp;id=FGI4Oyps1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Visual Studio Code introduces agent skills for Copilot</strong>: Visual Studio Code 1.108 brings an experimental ‘Agent Skills’ feature, allowing developers to extend GitHub Copilot’s capabilities with custom instructions, scripts, and resources. These skills, organized in <code>.github/skills</code> directories, load automatically based on context, enabling users to customize or augment Copilot’s functionality directly. This aims to enhance developer productivity by making AI assistance more adaptable to specific workflows. <a href="https://app.daily.dev/posts/xn90KJxf1" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Ollama 0.14 lets LLMs run bash commands</strong>: Ollama 0.14-rc2 introduces an experimental agent loop mode, enabling LLMs to execute bash commands and perform web searches on your system with interactive user approval and built-in safety mechanisms. <a href="https://app.daily.dev/posts/b5gNbsldC" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents shift debugging to traces, not code</strong>: With AI agents, the source of truth moves from code to runtime traces because decision logic resides within the model, making trace analysis crucial for debugging, testing, and performance optimization. <a href="https://app.daily.dev/posts/4345MlXhv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Claude Code 2.1 brings major workflow improvements</strong>: Claude Code 2.1 introduces forked sub-agents for parallel task execution, automatic skill hot reloading, session teleportation, Chrome browser integration for terminal testing, and persistent memory across sessions. <a href="https://app.daily.dev/posts/WChBmOscl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Neo4j integrates with Dify for AI workflows</strong>: Neo4j now offers a plugin for Dify, an open-source platform for building AI workflows, allowing integration of knowledge graphs into agentic workflows and enabling Cypher query execution from natural language. <a href="https://app.daily.dev/posts/xt2S0tUFX" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Spree e-commerce platform hit by IDOR vulnerabilities</strong>: Two critical IDOR vulnerabilities (CVE-2026-22589 and CVE-2026-22588) in Spree e-commerce platform allow unauthenticated and authenticated attackers to access guest user PII and other users’ address data, respectively, with patches available in recent versions. <a href="https://app.daily.dev/feed-by-ids?id=OLsziBWC5&amp;id=clMuODSiY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Budgie 10.10 officially migrates to Wayland</strong>: The Budgie 10.10 desktop environment has officially transitioned from X11 to Wayland, marking a significant milestone and allowing the development team to focus on Budgie 11. <a href="https://app.daily.dev/posts/aF4NpjHCh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 improves Rust + LTO kernel builds</strong>: Linux kernel 7.0 is set to include improvements for building kernels with Rust and Link-Time Optimization (LTO), enhancing performance and code generation for Rust components. <a href="https://app.daily.dev/posts/eeFdKxuyc" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Tailwind CSS faces AI-induced existential crisis</strong>: Apparently, AI is so good at consuming documentation that it’s killing off the business model for projects like Tailwind CSS, leading to layoffs and revenue decline. The company even rejected GitHub’s request for ‘LLM-friendly’ docs, because why make it easier for the machines that are eating your lunch? It’s a real head-scratcher: AI drives adoption, but then starves the very projects it relies on. Maybe we should just train LLMs to pay for open source. <a href="https://app.daily.dev/posts/f3fYMXuKZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Learning C++ in 2026 is a ‘mistake’</strong>: According to Stefan Mischook, if you’re learning C++ in 2026, you’re making a ‘mistake’ because manual memory management is just a bug-and-vulnerability factory. He suggests pivoting to ‘modern’ languages like JavaScript, C#, or Java for better productivity. Because, you know, those languages are famously bug-free and never have performance issues that require dropping down to something like, oh, C++. <a href="https://app.daily.dev/posts/trj7gVKm4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Raspberry Pi + RTX 5090: the gaming combo nobody asked for</strong>: Someone actually tried to game on a Raspberry Pi 5 hooked up to an RTX 5090. Unsurprisingly, modern games are ‘unplayable’ due to x86 emulation overhead, yielding a whopping 15 FPS in Cyberpunk 2077. But hey, Portal 2 from 2011 runs great! It’s a fascinating experiment in proving that you can’t just throw a powerful GPU at an ARM board and expect miracles, especially when the software stack is still playing catch-up. <a href="https://app.daily.dev/posts/41amF20BK" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
