<h2>Must know</h2>
<ul>
<li><strong>AI coding agents are fast, but still need human guardrails</strong>: AI coding agents like Claude Code and GitHub Copilot can significantly accelerate development, with some projects seeing 10-100x velocity gains for well-defined tasks. Cursor’s AI agents even built a web browser in Rust with over a million lines of code in about a week. However, these tools struggle with ambiguous requirements, can introduce more vulnerabilities, and decrease code refactoring. Experts, including those cited in developer discussions, emphasize the need for strong human oversight, automated quality checks, and peer reviews to prevent technical debt and ensure quality. LLVM and Oh My Zsh maintainers are now formalizing policies requiring human review and accountability for AI-assisted contributions, highlighting that the human in the loop is still critical. <a href="https://app.daily.dev/feed-by-ids?id=EqMZfdLoq&amp;id=GPLZOgOgP&amp;id=J24eplQdr&amp;id=cSbwL1BS5&amp;id=EdB2gt1Vu&amp;id=tr2l7aEc5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubernetes is now the standard for enterprise AI, but culture is the new bottleneck</strong>: Kubernetes production usage has soared to 82% among surveyed organizations, establishing it as the go-to infrastructure for enterprise AI workloads, with 66% using it for inference. The 2025 CNCF Annual Cloud Native Survey, widely cited, reveals that cultural resistance, not technical complexity, is now the primary barrier to adoption. Mature organizations are leveraging GitOps, platform engineering, and observability tools like OpenTelemetry to scale their Kubernetes deployments, especially for AI. This shift indicates that while the technology is mature, organizational dynamics are now the critical factor for successful AI integration. <a href="https://app.daily.dev/posts/Lwsgn3lZ5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-driven exploit and malware development is accelerating rapidly</strong>: Large language models are no longer just assisting but actively generating sophisticated exploits and malware. Experiments show LLMs like GPT-5.2 can create over 40 distinct exploits for zero-day vulnerabilities, bypassing modern security mitigations in hours for minimal cost. 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 human teams. These developments, reported on January 20, 2026, indicate a significant shift towards industrialized cyber offense, making the limiting factor LLM token throughput rather than human expertise. <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>AI-powered observability for OpenShift AI environments</strong>: Red Hat has released an AI-powered observability tool that translates complex infrastructure metrics into conversational insights for OpenShift AI, allowing natural language queries like “Why is my model slow?” to diagnose issues across GPU performance, model throughput, and cluster health. <a href="https://app.daily.dev/posts/jpKDLQQHM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Python package manager ‘uv’ challenges pip</strong>: The new Python package manager ‘uv’ offers faster installs, reproducible environments, and better uninstall mechanisms compared to the standard ‘pip’, prompting developers to evaluate based on project needs and workflow efficiency. <a href="https://app.daily.dev/posts/Y2d0smXfY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>EU launches decentralized vulnerability database</strong>: The 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, aiming to reduce European reliance on US-based systems and provide autonomous vulnerability identifiers. <a href="https://app.daily.dev/posts/i2uCJ3RGE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft mandates Intune security updates for mobile apps</strong>: From January 19, 2026, Microsoft will enforce mandatory Intune Mobile Application Management security updates, requiring iOS and Android apps to use the latest Intune SDK versions or risk being blocked from accessing corporate resources. <a href="https://app.daily.dev/posts/PNEoUht10" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Taiwan and US sign AI and semiconductor supply chain pact</strong>: Taiwan and the United States have finalized a strategic trade agreement focused on semiconductors and AI supply chains, with Taiwan committing $250 billion in investments and credit guarantees for US-based production, and Taiwanese chip makers receiving lowered tariffs. <a href="https://app.daily.dev/posts/suaVSewS2" 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 that implements graphs as sparse matrices, claims to be 10-496x faster than Neo4j for multi-hop queries by leveraging linear algebra for parallel traversal, 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>North Korean hackers target fintech developers via VS Code projects</strong>: North Korean threat actors are exploiting VS Code project configuration files to deploy backdoors and malware on developer machines, targeting those in cryptocurrency, blockchain, and fintech sectors through fake job assessment projects. <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 coding tools: the ultimate productivity hack, or just more tech debt?</strong>: Some developers are claiming AI coding tools are transforming workflows, allowing them to offload routine tasks and focus on ‘complex problem-solving.’ Yet, others, like the Ruby on Rails creator, call AI tools ‘flickering light bulbs’ that generate inconsistent, hard-to-maintain code. It seems we’re still debating whether these tools are actually making us more productive or just creating more work for the humans who have to clean up the AI’s mess. <a href="https://app.daily.dev/posts/zWcZCkr3l" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI investment: CEOs still waiting for the magic money tree</strong>: PwC’s latest survey of 4,454 CEOs reveals a harsh truth: 56% saw no revenue increase or cost decrease from their AI spending. Only a measly 12% got both. Meanwhile, ‘agentic AI’ is being flagged as the next big hype cycle, promising 25% productivity gains and 45% processing time reduction, but without any specified sources. It seems many are still waiting for AI to deliver on its promises, while others are already calling out the next wave of over-promising. <a href="https://app.daily.dev/feed-by-ids?id=Kov24DV15&amp;id=ogb2Wd6xs&amp;id=rEdiBiFbV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents: the new ‘psychosis’ in your production environment</strong>: Changelog News is apparently worried about ‘AI agent psychosis,’ a fancy term for when AI agents act erratically in production. Because, you know, software has never been erratic before AI. Meanwhile, <a href="http://Monday.com" target="_blank" rel="noopener nofollow">Monday.com</a> is actually solving real problems with LLM hallucinations by pre-computing data, proving that sometimes, good old engineering beats sensational metaphors. <a href="https://app.daily.dev/feed-by-ids?id=5ZpMYMEZm&amp;id=dRAo2Mdf1&amp;id=j6bBmLvCW" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
