<h2>Must know</h2>
<ul>
<li><strong>Microsoft faces backlash over AI data center expansion</strong>: Microsoft is responding to widespread community opposition across 24 states regarding its AI data center projects. The company has pledged a “community-first” approach, promising to pay full electricity grid costs, minimize water use, and create local jobs, as reported by multiple sources. This initiative comes after withdrawing from a Wisconsin project and facing resistance in Michigan and Ohio, highlighting the growing environmental and social impact concerns of hyperscale AI infrastructure. <a href="https://app.daily.dev/posts/jqtBl3Jnb" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical vulnerabilities found in AI and developer tools</strong>: ServiceNow’s Virtual Agent chatbot had a critical vulnerability allowing full system control with just an email, due to weak authentication and insecure AI features. Similarly, Superhuman AI and Grammarly were susceptible to prompt injection, enabling email exfiltration without user interaction. Both companies have issued patches, but these incidents, reported by security researchers, underscore significant security risks in integrating agentic AI into enterprise and productivity platforms. <a href="https://app.daily.dev/feed-by-ids?id=2A2nfRTHq&amp;id=EDFS33r2W&amp;id=KcLhdNkuO&amp;id=qah3DxopU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic commits $1.5 million to Python open source security</strong>: Anthropic is investing $1.5 million over two years into the Python Software Foundation (PSF) to bolster security across the Python ecosystem, including CPython and PyPI. This funding aims to support automated malware detection for PyPI packages and contribute to the PSF Developer in Residence program, benefiting millions of Python users by mitigating supply-chain attacks, according to announcements from high-reliability organizations. <a href="https://app.daily.dev/posts/yCHi0TB7q" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Rust infrastructure team recap and plans</strong>: The Rust Infrastructure Team successfully migrated to a new CI bot, moved <a href="http://crates.io" target="_blank" rel="noopener nofollow">crates.io</a> and <a href="http://docs.rs" target="_blank" rel="noopener nofollow">docs.rs</a> to Fastly CDN, and expanded Infrastructure as Code coverage, with Q1 2026 plans including <a href="http://docs.rs" target="_blank" rel="noopener nofollow">docs.rs</a> modernization and establishing external hardware CI policy. <a href="https://app.daily.dev/posts/m2OVQmP6S" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Advanced Linux malware targets cloud and containers</strong>: Check Point Research identified VoidLink, an advanced malware framework written in Zig, designed to target Linux cloud and container environments like Docker and Kubernetes, featuring 37+ modular plugins for credential harvesting and evasion. <a href="https://app.daily.dev/feed-by-ids?id=uJaIfhgHH&amp;id=QWUKVWdsK&amp;id=x9wl564Rc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Lyft’s ML platform handles millions of predictions per second</strong>: Lyft developed LyftLearn Serving, an internal ML platform that supports millions of predictions per second using a microservice-based architecture with Flask/Gunicorn, Kubernetes, and Envoy, allowing teams to deploy individualized services efficiently. <a href="https://app.daily.dev/posts/v99Yv1oDt" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Node.js releases security fixes for critical vulnerabilities</strong>: Node.js versions 20.20.0 and 25.3.0 were released with security fixes addressing multiple CVEs, including a denial-of-service vulnerability from stack space exhaustion impacting async_hooks and related frameworks, and stricter permission checks for symlink APIs. <a href="https://app.daily.dev/feed-by-ids?id=DwsoMgMKH&amp;id=uw83EGG5r" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents for enterprise databases with LEDGE MCP Server</strong>: LangGrant launched the LEDGE MCP Server, enabling LLMs to perform analytics on enterprise databases like Postgres without direct data access, using schema and metadata to reduce token costs and ensure security compliance. <a href="https://app.daily.dev/posts/v8MGOR2uR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rust in production for embedded smart door locks</strong>: Achilles has been using Rust for 4-5 years in production for smart door lock firmware, leveraging its safety features and efficient code size (300-420KB) with the Embassy async runtime for efficient, interrupt-driven programming. <a href="https://app.daily.dev/feed-by-ids?id=SVl4BYYw0&amp;id=U09VghD1S" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Bifrost offers 40x faster LLM gateway in Go</strong>: Bifrost, an open-source LLM gateway written in Go, provides unified access to 15+ AI providers with a single OpenAI-compatible API, demonstrating 40x lower gateway overhead and 68% lower memory usage compared to LiteLLM. <a href="https://app.daily.dev/posts/P8RYoHcNL" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Meta’s AI leadership shakeup and alleged Llama 4 manipulation</strong>: Yann LeCun, Meta’s former chief AI scientist, reportedly left after Alexander Wang, CEO of Scale AI, was appointed head of superintelligence. LeCun publicly criticized Wang as ‘inexperienced’ and alleged Meta’s AI team manipulated Llama 4 test results. So, the internal drama at Meta AI is apparently more compelling than their actual AI products. Who needs a reality show when you have corporate infighting? <a href="https://app.daily.dev/posts/ILAkYeoup" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated book spam: the new frontier of ‘content’</strong>: HiTeX Press has published over 800 AI-generated technical books in one year, without human review, selling them for 8€ each on platforms like Amazon. A Starlark programming language book, for example, was found to contain fabricated code and features. Because nothing says ‘expert knowledge’ like a book written by an algorithm that hallucinates facts and sells it for the price of a fancy coffee. <a href="https://app.daily.dev/posts/PmK6l9z8J" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Pentagon integrates Grok AI despite ‘inappropriate images’ controversy</strong>: US Defense Secretary Pete Hegseth announced plans to integrate Elon Musk’s Grok AI into Pentagon networks, despite recent controversies over Grok generating ‘inappropriate images.’ Apparently, the Pentagon’s AI acceleration strategy means ignoring minor details like ‘generating inappropriate content’ in favor of ‘eliminating bureaucratic barriers.’ What could possibly go wrong when your classified networks are powered by an AI that can’t tell a joke from a national security threat? <a href="https://app.daily.dev/feed-by-ids?id=KoB4l9TcK&amp;id=jRGkCZr4W" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
