<h2>Must know</h2>
<ul>
<li><strong>AI coding agents are getting scary good (and fast)</strong>: New benchmarks show GLM-4.7 Flash, a 30B-A3B mixture-of-experts model, significantly outperforming competitors like Qwen 3-30B-A3B in coding and agentic AI tasks, especially for tool calling. Meanwhile, LLMs like Opus 4.5 and GPT-5.2 are now generating sophisticated exploits for zero-day vulnerabilities, bypassing modern security mitigations in hours for minimal cost, as reported by Hacker News. This suggests an ‘industrialized’ model for exploit development, shifting the bottleneck from human expertise to LLM token throughput. <a href="https://app.daily.dev/feed-by-ids?id=eWdzokgG0&amp;id=WkT9zuzY1&amp;id=nLTC7GTyu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Next.js 16 turbocharges builds and adds AI debugging</strong>: Next.js 16 stabilizes Turbopack as the default bundler, promising 2-5x faster production builds and 10x faster hot reloading. It also introduces AI-assisted debugging via Model Context Protocol (MCP) integration in DevTools, allowing AI agents to access unified logs and suggest fixes, according to Syncfusion. This release also brings explicit opt-in caching with ‘use cache’ and React Compiler support for automatic memoization. <a href="https://app.daily.dev/feed-by-ids?id=91xiEmcVe&amp;id=bbohKtF7X" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>WebAssembly continues its march to ubiquity</strong>: WebAssembly saw significant progress in 2025, with Safari closing key feature gaps and several proposals reaching standardization. Kotlin released a beta Wasm compiler, .NET 10 improved performance, and WASI advanced towards 0.3 with async support. Akamai even acquired Fermyon, signaling major movement in edge computing. Expect WASI 0.3 and continued Component Model advancements in 2026. <a href="https://app.daily.dev/posts/jtf8mrjCJ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI agent frameworks are standardizing</strong>: The A2A Java SDK 1.0.0.Alpha1 aligns with the forthcoming Agent2Agent Protocol 1.0, standardizing communication between AI agents, while Vercel released ‘skills,’ a CLI tool for managing skill packages for AI agents, supporting tools like Cursor, GitHub Copilot, and Claude Code. <a href="https://app.daily.dev/feed-by-ids?id=LmGT7Q5vx&amp;id=PMa93QJZs&amp;id=ZcWzztsx0&amp;id=JdhS5Ojd6" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubernetes is now the standard for enterprise AI</strong>: The 2025 CNCF survey shows Kubernetes production usage at 82%, with 66% of AI adopters using it for inference, establishing it as the standard infrastructure for enterprise AI workloads, though cultural resistance is now the main barrier. <a href="https://app.daily.dev/posts/Lwsgn3lZ5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Python package manager ‘uv’ challenges pip</strong>: While pip remains the standard, ‘uv’ is emerging as a faster alternative for Python package management, offering quicker installs, reproducible environments, and better uninstall mechanisms, prompting developers to evaluate based on project needs. <a href="https://app.daily.dev/posts/Y2d0smXfY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>TigerBeetle crushes PostgreSQL for financial ledgers</strong>: TigerBeetle, a specialized financial database, demonstrated 2.8x to 6.6x faster transaction processing than PostgreSQL for double-entry bookkeeping, achieving 42,000 TPS by using sparse matrices and linear algebra for parallel traversal. <a href="https://app.daily.dev/posts/DUFN7D8Ly" 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 SDK versions or face 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>North Korean hackers target fintech developers via VS Code</strong>: North Korean threat actors are exploiting VS Code projects by distributing malicious job assessment projects that deploy backdoors and malware upon opening, specifically targeting developers in cryptocurrency, blockchain, and fintech sectors. <a href="https://app.daily.dev/posts/1wKLyk8gG" 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 aiming to reduce European reliance on US-based systems like CVE, integrating data from over 25 sources with autonomous identifier assignment. <a href="https://app.daily.dev/posts/i2uCJ3RGE" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Vibecoding: the latest AI buzzword for executives who want to code</strong>: Emergent, an Indian AI startup, just bagged $70 million for its ‘vibe-coding’ platform, which supposedly lets executives build full-stack apps without ‘advanced coding expertise.’ An opinion piece even suggests every exec should be ‘vibecoding’ to become an individual contributor. Because nothing says ‘deep technical understanding’ like a C-suite exec ‘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 agent productivity: 100x velocity or just 100x the headaches?</strong>: Some claim AI coding agents deliver 10-100x velocity gains for well-defined tasks, with Cursor even building a 1M+ line web browser in a week. Yet, the same sources admit these agents struggle with ambiguous requirements, produce ‘mostly correct’ but not production-ready code, and increase vulnerabilities. It seems we’re trading human-made bugs for AI-generated ones, just faster, and with a side of ‘mostly correct’ glitches. <a href="https://app.daily.dev/feed-by-ids?id=EqMZfdLoq&amp;id=GPLZOgOgP&amp;id=J24eplQdr&amp;id=cSbwL1BS5&amp;id=zWcZCkr3l" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI security: because we needed more ways for AI to break things</strong>: Google Gemini was found vulnerable to prompt injection via calendar invites, allowing it to exfiltrate private meeting data. Meanwhile, Anthropic’s Git MCP server had three critical vulnerabilities enabling remote code execution via prompt injection. It’s almost like integrating AI into everything creates entirely new attack surfaces that traditional security measures can’t handle. Who knew that giving an AI access to your calendar could be a bad idea? <a href="https://app.daily.dev/feed-by-ids?id=dHjNBMkFO&amp;id=U6A8HrCeq&amp;id=5ZpMYMEZm&amp;id=dRAo2Mdf1&amp;id=j6bBmLjCW" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
