<h2>Must know</h2>
<ul>
<li><strong>Kubernetes is now the standard for enterprise AI</strong>: The 2025 CNCF Annual Cloud Native Survey reveals Kubernetes production usage hit 82%, solidifying its role as the infrastructure standard for enterprise workloads, especially AI. For AI workloads, 66% of adopters use Kubernetes for inference. Interestingly, cultural resistance, not technical complexity, is now the primary barrier to adoption, as reported by the CNCF survey. This shift highlights the maturity of Kubernetes and the need for organizational adaptation. <a href="https://app.daily.dev/posts/Lwsgn3lZ5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding agents show velocity gains but need guardrails</strong>: AI coding agents like Claude Code can deliver significant velocity gains (10–100x) for well-defined tasks, as seen in a 150K-line codebase. However, they struggle with ambiguous requirements, often leading to code repetition and omitted tests, according to developer accounts. Research also indicates that while tools like GitHub Copilot boost individual speed by up to 55%, they correlate with 322% more code vulnerabilities and lower delivery stability. This suggests that robust guardrails like automated quality checks, mandatory peer reviews, and strong test coverage are crucial for safe deployment, as highlighted by multiple expert narratives. <a href="https://app.daily.dev/feed-by-ids?id=EqMZfdLoq&amp;id=GPLZOgOgP&amp;id=J24eplQdr&amp;id=cSbwL1BS5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI accelerates exploit and malware development</strong>: Large language models (LLMs) like Opus 4.5 and GPT-5.2 are now generating sophisticated exploits for zero-day vulnerabilities, bypassing modern security mitigations such as ASLR and CFI. One experiment showed GPT-5.2 solving a complex scenario in 3 hours for about $50 in compute. Check Point Research also identified VoidLink, the first known AI-generated sophisticated malware suite, developed by a single person using AI agents in under a week—a task estimated to take 30 weeks for multiple human teams. These developments, reported on January 20, 2026, indicate a significant shift towards industrialized offensive cybersecurity, where AI is not just assisting but directly enabling threat actors to scale complex operations. <a href="https://app.daily.dev/feed-by-ids?id=WkT9zuzY1&amp;id=nLTC7GTyu" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>New Go-based SQL client with AI features</strong>: SQDesk CLI, a terminal-based SQL client built with Go 1.21+, now offers AI-powered query generation and refactoring for PostgreSQL, MySQL, and SQLite, alongside syntax highlighting and auto-completion. <a href="https://app.daily.dev/posts/da02fw80D" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>WebAssembly advances with new features and adoption</strong>: WebAssembly saw significant progress in 2025 with Safari closing feature gaps, new proposals reaching standardization, and increased adoption on 5.5% of Chrome-visited websites, with WASI 0.3 expected in February 2026. <a href="https://app.daily.dev/posts/jtf8mrjCJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Next.js 16 boosts performance with Turbopack and caching</strong>: Next.js 16 stabilizes Turbopack as the default bundler for 2-5x faster production builds and 10x faster hot reloading, and introduces explicit ‘use cache’ directive for more predictable caching behavior. <a href="https://app.daily.dev/posts/91xiEmcVe" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>FalkorDB claims massive speedup over Neo4j for graph queries</strong>: FalkorDB, a graph database using linear algebra and sparse matrices, claims to be 10-496x faster than Neo4j for multi-hop queries, targeting real-time AI agents and RAG systems. <a href="https://app.daily.dev/posts/GE8qOCt7y" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft updates React Native for Windows</strong>: Microsoft released React Native Windows version 0.81, adding Hermes debugger support and Fabric architecture, but still lags mobile versions and has limited third-party package support. <a href="https://app.daily.dev/posts/7jYV4dEG1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>EU launches decentralized vulnerability database</strong>: GCVE has launched <a href="http://db.gcve.eu" target="_blank" rel="noopener nofollow">db.gcve.eu</a>, a public, decentralized cybersecurity vulnerability database integrating data from over 25 sources to reduce European reliance on US-based systems like CVE. <a href="https://app.daily.dev/posts/i2uCJ3RGE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>North Korean hackers target fintech developers via VS Code</strong>: North Korean threat actors are exploiting VS Code projects by tricking developers into cloning malicious repositories, leading to automatic payload execution and deployment of backdoors and cryptocurrency miners on compromised machines. <a href="https://app.daily.dev/posts/1wKLyk8gG" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI-powered ‘vibecoding’ startup raises $70M</strong>: Emergent, an Indian AI-powered ‘vibecoding’ startup, just bagged $70 million in Series B funding, claiming $50 million ARR and 5 million users after only seven months. Apparently, this ‘vibecoding’ lets executives build full-stack apps without needing actual coding expertise. Because nothing says ‘robust software’ like a senior leader ‘vibecoding’ a hackathon platform. What could possibly go wrong? <a href="https://app.daily.dev/feed-by-ids?id=AllrxKqHW&amp;id=McILqauKu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated SwiftUI code: the gift that keeps on leaking</strong>: Turns out, AI-generated SwiftUI code is a master of memory leaks, creating retain cycles by strongly capturing <code>self</code> in deeply nested closures. Engineers had to manually refactor with weak references and explicit capture lists to stabilize the app. So much for AI being the infallible coding guru; seems it still needs a human to clean up its mess and teach it basic memory management. <a href="https://app.daily.dev/feed-by-ids?id=0Dxh8ahUn&amp;id=NOlmy6yd2&amp;id=yTmR3XbQ9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI debugging: because your code wasn’t complex enough</strong>: Next.js 16 is rolling out ‘AI-assisted debugging’ via Model Context Protocol (MCP) integration, promising AI agents will access unified logs and suggest fixes. Because what every developer needs is another layer of abstraction and an AI telling them what’s wrong, instead of, you know, writing clear code in the first place. I’m sure this won’t just add more noise to the already chaotic debugging process. <a href="https://app.daily.dev/posts/91xiEmcVe" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
