<h2>Must know</h2>
<ul>
<li><strong>PackageGate zero-days expose npm and yarn to worm attacks</strong>: Six zero-day vulnerabilities, dubbed PackageGate, are affecting JavaScript package managers like npm and yarn, allowing attackers to bypass security controls designed to prevent worm attacks. While pnpm, vlt, and Bun have already patched these flaws, npm claims the behavior is ‘by design,’ and yarn remains unpatched. Developers are strongly advised by researcher Oren Yomtov to migrate to patched package managers to avoid potential malicious payload delivery. <a href="https://app.daily.dev/posts/klUkuFZvK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft Office zero-day forces emergency patch</strong>: Microsoft issued an emergency out-of-band patch for a critical zero-day vulnerability (CVE-2026-21509) in Office, actively exploited in the wild. The flaw, a security feature bypass, allows attackers to circumvent protections against legacy COM and OLE components using malicious Office files. While newer versions are patched, Office 2016 and 2019 users must rely on manual registry workarounds, and CISA has mandated federal agencies patch by February 16, 2026. <a href="https://app.daily.dev/posts/qC937DXHe" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents reshape developer roles and enterprise operations</strong>: The rise of AI agents is shifting developer roles from coding to product architecture and outcome engineering, as AI automates much of the execution. This creates new infrastructure needs for orchestration and fault tolerance, with companies like Temporal positioning themselves as critical providers. Enterprises are embedding AI agents as integrated features, moving towards semi- or fully-autonomous ‘coworkers,’ but face challenges in governance and avoiding ‘agent sprawl,’ according to analyses in “What will AI do to your career?” and “How Agents are quietly transforming the Enterprise.” <a href="https://app.daily.dev/feed-by-ids?id=b2fEANN64&amp;id=yBxtGCVqr" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI coding agents’ web browser claims disputed</strong>: Cursor’s claims that its AI agents independently created a functional web browser with a custom Rust rendering engine are being heavily disputed by developers, who describe the code as ‘spaghetti,’ failing to compile, and suffering from poor performance, suggesting the project relies heavily on external dependencies rather than being built from scratch. <a href="https://app.daily.dev/posts/gcgtwbk5H" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Open-source SERA coding agents released by Ai2</strong>: The Allen Institute for AI (Ai2) has released SERA, an open-source family of coding agent models designed for customization and fine-tuning on private codebases, with claimed cost reductions of up to 11 times cheaper than comparable approaches and the largest model solving 55% of SWE-Bench Verified problems. <a href="https://app.daily.dev/posts/J0iLEN79d" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Teleport unveils agentic identity framework for AI security</strong>: Teleport launched a new Agentic Identity Framework that applies cryptographic authentication to AI agents, moving away from traditional static credentials to reduce credential theft risks and enforce least-privileged access, while also providing real-time visibility and prompt tracking to combat injection attacks. <a href="https://app.daily.dev/posts/CI1C8FH63" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding agents impact open source engagement and revenue</strong>: A research paper suggests AI coding tools are disrupting open-source ecosystems by reducing direct developer and maintainer engagement, citing a 40% drop in Tailwind CSS documentation traffic that reportedly contributed to team layoffs, and proposing revenue-sharing models from AI companies to mitigate the impact. <a href="https://app.daily.dev/posts/aDodkq3hd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vinted automates Vespa search maintenance using Temporal</strong>: Vinted’s platform engineering team replaced Bash scripts with Temporal for managing its Vespa search engine operations, enabling fault-tolerant orchestration and zero-impact upgrades for its thousands of nodes, significantly reducing manual toil and allowing engineers to focus on higher-value development. <a href="https://app.daily.dev/posts/psgDmdeNF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AWS Network Firewall adds GenAI traffic controls</strong>: AWS Network Firewall now supports web category-based filtering for generative AI application traffic, allowing security teams to block or restrict access to GenAI services at the network level and enforce compliance policies, especially when combined with TLS inspection for deeper traffic analysis. <a href="https://app.daily.dev/posts/AdbvgaiZY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks SQL 2025 performance enhancements</strong>: Databricks announced significant performance improvements for Databricks SQL in 2025, including up to 40% faster production workloads, a 10x reduction in Unity Catalog latency, and 30% faster Delta Sharing queries, all automatically applied in DBSQL Serverless without user configuration. <a href="https://app.daily.dev/posts/rT9ZMYOFd" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Ricursive Intelligence: another AI unicorn born in two months</strong>: Ricursive Intelligence, founded by ex-Google researchers, just bagged $300 million at a $4 billion valuation, two months after launch. Their grand plan? An AI system to design and improve AI chips. Because apparently, the only thing faster than AI is venture capital FOMO. No products, no independent verification of their ‘AlphaChip’ claims, just a whole lot of money for a concept. What could possibly go wrong? <a href="https://app.daily.dev/posts/pqGutcjN8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding agents: still need a human in the loop, shocker</strong>: OpenAI’s latest technical breakdown of Codex CLI admits that AI coding agents, while good for boilerplate, are ‘brittle’ and prone to errors, especially outside their training data. They still require human oversight for production. So, after all the hype, it turns out we still need actual engineers to clean up the AI’s mess. Who would’ve thought that machines aren’t perfect? <a href="https://app.daily.dev/posts/qZX3wHxMK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Context engineering: new name, same RAG, more hype</strong>: RAG (Retrieval-Augmented Generation) is now ‘context engineering,’ a ‘broader field’ that includes MCP, because apparently, we can’t just stick with a perfectly descriptive acronym. This rebranding is supposedly driven by ‘agentic AI systems’ that manage ‘entire context states.’ It sounds like a fancy way to say ‘we’re still trying to make AI less dumb,’ but now with more buzzwords and a new product called ‘Agent Composer’ to sell to enterprises. Don’t worry, it’s ‘hot’ now. <a href="https://app.daily.dev/posts/tHZI0osnL" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
