<h2>Must know</h2>
<ul>
<li><strong>AI agents are getting industrial</strong>: LLMs like Opus 4.5 and GPT-5.2 are now generating over 40 distinct exploits for zero-day vulnerabilities, bypassing modern security mitigations like ASLR and CFI. According to Hacker News, GPT-5.2 solved the hardest scenario in just three hours for about $50, shifting exploit development from human expertise to LLM token throughput. This suggests offensive cybersecurity is rapidly industrializing, with AI enabling threat actors to scale and automate complex operations, as also reported by Check Point Research with the AI-generated VoidLink malware suite. <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 DeepMind and Boston Dynamics team up for industrial robots</strong>: Google DeepMind is integrating its Gemini AI into Boston Dynamics’ Atlas humanoid robots, moving them from prototypes to production-ready industrial automation. This partnership aims to deploy these robots in auto factories for autonomous decision-making with minimal human oversight, marking a significant step towards real-world industrial AI applications, as reported by Slidebean. <a href="https://app.daily.dev/posts/tmClU48mc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubernetes is now the standard for enterprise AI</strong>: Kubernetes production usage has hit 82% among organizations, establishing it as the go-to infrastructure for enterprise AI workloads, with 66% of AI adopters using it for inference. The 2025 CNCF Annual Cloud Native Survey reveals that cultural resistance, not technical complexity, is now the main barrier to adoption. This shift highlights Kubernetes’ maturity and its central role in modern cloud-native and AI operations. <a href="https://app.daily.dev/posts/Lwsgn3lZ5" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Anthropic fixes critical Git server vulnerabilities</strong>: Anthropic patched three critical vulnerabilities in its Git MCP server (CVE-2025-68145, CVE-2025-68144, CVE-2025-68143) that could have allowed remote code execution via prompt injection, highlighting new attack surfaces in AI-integrated systems. <a href="https://app.daily.dev/posts/U6A8HrCeq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New GLM-4.7 Flash model excels in local AI coding</strong>: GLM-4.7 Flash, a 30B-A3B mixture-of-experts model, significantly outperforms competitors like Qwen 3-30B-A3B in coding and agentic AI benchmarks, offering superior tool calling and fast inference for self-hosted local AI tasks. <a href="https://app.daily.dev/posts/eWdzokgG0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-powered SQL client for your terminal</strong>: SQDesk CLI is a new terminal-based SQL client supporting PostgreSQL, MySQL, and SQLite, featuring AI-powered query generation and refactoring, syntax highlighting, and auto-completion. <a href="https://app.daily.dev/posts/da02fw80D" 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 risk being blocked from corporate resources. <a href="https://app.daily.dev/posts/PNEoUht10" 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 cybersecurity vulnerability database designed to reduce European reliance on US systems by integrating data from over 25 sources and enabling autonomous vulnerability identifier assignment. <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 distributing malicious job assessment projects that deploy backdoors and malware upon opening, primarily targeting developers in cryptocurrency, blockchain, and fintech sectors. <a href="https://app.daily.dev/feed-by-ids?id=1wKLyk8gG&amp;id=Q9M3WuQfH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Taiwan and US sign AI and semiconductor pact</strong>: Taiwan and the United States finalized a strategic trade agreement focused on semiconductors and AI supply chains, with Taiwan committing $250 billion in investments and the US offering lowered tariffs and duty-free status for certain tech exports. <a href="https://app.daily.dev/posts/suaVSewS2" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI coding agents: 100x velocity or 100x headache?</strong>: Apparently, AI coding agents like Claude Code can deliver 10-100x velocity gains for ‘well-defined’ tasks, but struggle with ‘ambiguous, high-level product requirements,’ often resorting to code repetition and omitting tests. So, while Cursor claims to have built a browser with hundreds of AI agents in a week, the output is ‘mostly correct’ but not ‘production-ready.’ It seems AI is great at generating code, as long as you already know exactly what code you need and are prepared to clean up its mess. <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>Vibecoding: the next big thing or just more AI hype?</strong>: Emergent, an Indian AI startup, just bagged $70 million for its ‘vibe-coding’ platform, claiming 5 million users and $50 million ARR. An opinion piece even suggests every executive should be ‘vibecoding’ to build apps without engineers. It sounds like the latest flavor of ‘low-code/no-code’ wrapped in AI buzzwords, promising to turn everyone into a developer while conveniently ignoring the complexities of actual software engineering. 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 security: old wine in new bottles?</strong>: Security researchers found Google Gemini vulnerable to prompt injection via malicious calendar invites, leading to exfiltration of private meeting data. NIST is now pushing for ‘data trust’ and ‘context-driven risk evaluation’ for AI. Meanwhile, Bruce Schneier warns that AI platforms from OpenAI, Google, and Microsoft are embedding advertising-driven business models, raising manipulation risks. It seems we’re just swapping traditional software vulnerabilities for ‘semantic’ ones, and the ‘trust’ in AI is as solid as a marketing brochure. <a href="https://app.daily.dev/feed-by-ids?id=dHjNBMkFO&amp;id=5ZpMYMEZm&amp;id=dRAo2Mdf1&amp;id=j6bBmLvCW&amp;id=8uzQJ4GD3&amp;id=QaJ2IQUdX" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
