<h2>Must know</h2>
<ul>
<li><strong>Kubernetes is now the standard for enterprise AI, but culture is the new bottleneck</strong>: The 2025 CNCF Annual Cloud Native Survey reveals Kubernetes production usage hit 82%, with 66% of AI workloads running on it for inference. This solidifies Kubernetes as the go-to infrastructure for modern enterprise and AI. However, the survey, widely cited by sources like the CNCF, highlights that cultural resistance, not technical hurdles, is now the primary barrier to adoption. Mature organizations are leveraging GitOps and platform engineering to scale, but the human element remains the biggest challenge. <a href="https://app.daily.dev/posts/Lwsgn3lZ5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding agents are fast, but still need human guardrails</strong>: AI coding agents like Claude Code are delivering significant velocity gains (10–100x) for well-defined tasks, as seen in a 150K-line project. Cursor even built a web browser in Rust with AI agents in about a week, according to their blog. However, these agents struggle with ambiguous requirements, often repeating code or omitting tests. Research cited in J24eplQdr indicates GitHub Copilot users see 322% more code vulnerabilities. The consensus is that strong human oversight, peer reviews, and robust testing are essential to prevent technical debt and quality regressions. <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>North Korean hackers target fintech developers via VS Code projects</strong>: North Korean threat actors are exploiting VS Code projects to target fintech, crypto, and blockchain developers. They distribute malicious job assessment projects that, when opened in VS Code, automatically execute payloads, deploying backdoors and malware. This campaign leverages developers’ trust in open-source workflows, enabling remote code execution and cryptocurrency mining on compromised machines, as reported by technical security firms. This highlights a critical evolving threat to high-value sectors. <a href="https://app.daily.dev/feed-by-ids?id=1wKLyk8gG&amp;id=Q9M3WuQfH" 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 is a new terminal-based SQL client built with Go 1.21+ that supports PostgreSQL, MySQL, and SQLite, featuring AI-powered query generation, syntax highlighting, and visual connection management. <a href="https://app.daily.dev/posts/da02fw80D" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Next.js 16 boosts performance with Turbopack and AI debugging</strong>: Next.js 16 stabilizes Turbopack as the default bundler for 2-5x faster builds, introduces explicit Cache Components, and integrates AI-assisted debugging via Model Context Protocol (MCP) in DevTools. <a href="https://app.daily.dev/posts/91xiEmcVe" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenCode offers flexible, open-source AI agent for terminal coding</strong>: OpenCode is an open-source AI agent that runs in the terminal, providing flexible coding automation without vendor lock-in to specific AI models or restrictive licensing terms. <a href="https://app.daily.dev/posts/NR4LrQXl3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Lighthouse for PHP simplifies web development with minimal framework</strong>: Lighthouse is a new minimal PHP micro-framework designed for rapid development of MVPs and internal tools, featuring zero-config SQLite, built-in security, HTMX for frontend, and Pico.css for styling. <a href="https://app.daily.dev/posts/bI3vfttQJ" 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, claims to be 10-496x faster than Neo4j for multi-hop queries by implementing graphs as sparse matrices and leveraging linear algebra for parallel traversal. <a href="https://app.daily.dev/posts/GE8qOCt7y" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>EU launches decentralized vulnerability database</strong>: The EU’s 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 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>Microsoft mandates Intune security updates for mobile apps</strong>: Microsoft is enforcing mandatory Intune Mobile Application Management security updates from January 19, 2026, requiring iOS and Android apps to update to the latest SDK versions or face being blocked from corporate resources. <a href="https://app.daily.dev/posts/PNEoUht10" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI investment: all hype, no ROI for most CEOs</strong>: PwC surveyed 4,454 CEOs, and guess what? 56% saw no revenue increase or cost decrease from their AI spending. Only a measly 12% saw both. So much for the AI revolution, it seems most companies are just throwing money at the problem, hoping for magic, while the ‘agentic AI’ hype cycle continues to spin, as noted in various industry reports. <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-generated exploits: because humans weren’t fast enough at breaking things</strong>: Apparently, humans are too slow at creating exploits. Now, LLMs like Opus 4.5 and GPT-5.2 are generating 40 distinct zero-day exploits, bypassing modern security mitigations, and GPT-5.2 even solved a ‘challenging protected scenario’ in 3 hours for $50. Check Point Research also found the first AI-generated sophisticated malware suite, VoidLink, developed in under a week. So, AI isn’t just writing bad code, it’s actively trying to break everything, and it’s getting industrial-scale efficient at it. <a href="https://app.daily.dev/feed-by-ids?id=WkT9zuzY1&amp;id=nLTC7GTyu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google Gemini’s prompt injection: because who needs traditional exploits when you have calendar invites?</strong>: Security researchers found a vulnerability in Google Gemini where malicious calendar invitations could be used for prompt injection, allowing attackers to manipulate the model and exfiltrate private meeting data. It seems Google’s AI is so smart it can be tricked by a simple calendar invite. This highlights the ‘semantic security challenges’ of AI, where intent manipulation is the new code exploit, and traditional defenses are, apparently, useless. <a href="https://app.daily.dev/feed-by-ids?id=dHjNBMkFO&amp;id=5ZpMYMEZm&amp;id=dRAo2Mdf1&amp;id=j6bBmLvCW" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
