<h2>Must know</h2>
<ul>
<li><strong>AI coding agents are fast, but still need adult supervision</strong>: New benchmarks show AI coding agents like Claude Opus 4.5 and GPT-5.2-Codex high can generate complex code rapidly, with Claude being the fastest and GPT-5.2 offering better architectural quality. However, these tools still struggle with ambiguous requirements, often resort to code repetition, and introduce vulnerabilities, as highlighted by developer experiences and research cited in multiple reports. GitHub Copilot, for instance, increases individual velocity but correlates with 322% more code vulnerabilities and lower delivery stability. The consensus, echoed by industry experts and the Oh My Zsh maintainers, is that while AI can accelerate development, human oversight, strong testing, and clear architectural guidance remain critical 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&amp;id=1eWv43sPn&amp;id=20Yl6IWvm&amp;id=jHrHB1kVi&amp;id=yIClekHIh&amp;id=ymVBkKkOa&amp;id=tr2l7aEc5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is industrializing exploit generation</strong>: LLMs like Opus 4.5 and GPT-5.2 are now capable of generating dozens of distinct exploits for zero-day vulnerabilities, even bypassing modern security mitigations like ASLR and CFI. One experiment showed GPT-5.2 solving a complex scenario in just three hours for about $50 in compute, indicating that exploit development is shifting from human expertise to LLM token throughput. Check Point Research also identified VoidLink, the first 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. This marks a significant acceleration in offensive cybersecurity capabilities, as reported by Hacker News and Check Point Research. <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>Kubernetes is now the standard for enterprise AI, but culture is the new bottleneck</strong>: The 2025 CNCF Annual Cloud Native Survey reveals that Kubernetes production usage has hit 82%, with 66% of AI adopters using it for inference. It’s officially the standard for enterprise workloads, especially AI. However, the survey also points out that cultural resistance, not technical complexity, is now the primary barrier to adoption. Mature organizations are embracing GitOps and platform engineering, and observability tools like OpenTelemetry are gaining traction. This shift means that while the tech is ready, organizations need to address their internal dynamics to fully leverage cloud-native and AI capabilities, according to the CNCF. <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>Google Gemini prompt injection raises enterprise security concerns</strong>: Security researchers found a vulnerability in Google Gemini where malicious calendar invitations could exploit the model’s automatic data ingestion, allowing attackers to manipulate its behavior and potentially exfiltrate sensitive meeting data. <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>
<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 and arbitrary file manipulation, 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>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 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 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/feed-by-ids?id=1wKLyk8gG&amp;id=Q9M3WuQfH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Agent2Agent SDK and architecture advances</strong>: The A2A Java SDK 1.0.0.Alpha1 has been released, aligning with the forthcoming Agent2Agent Protocol 1.0 specification to standardize communication between AI agents, emphasizing modularity and extensibility in agentic AI system design. <a href="https://app.daily.dev/feed-by-ids?id=LmGT7Q5vx&amp;id=PMa93QJZs&amp;id=ZcWzztsx0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>FalkorDB claims massive speedup over Neo4j for graph queries</strong>: FalkorDB, a new 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, targeting real-time AI agents and RAG systems. <a href="https://app.daily.dev/feed-by-ids?id=GE8qOCt7y&amp;id=GHFDClpAg&amp;id=bbohKtF7X" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft updates React Native for Windows, still lags mobile</strong>: Microsoft released React Native Windows version 0.81 with Hermes debugger support and Fabric architecture, but it still lags behind mobile versions and has limited third-party package support, raising community skepticism about Microsoft’s dual investment in React Native and MAUI. <a href="https://app.daily.dev/feed-by-ids?id=7jYV4dEG1&amp;id=qcm6RjfK7" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI investment: all hype, no ROI?</strong>: PwC surveyed 4,454 CEOs, and a whopping 56% reported no revenue increase or cost decrease from their AI spending. Only 12% saw both. Meanwhile, ‘agentic AI’ is being flagged as the next big hype cycle. It seems many are still waiting for the AI magic to actually, you know, <em>do something</em> for their bottom line, rather than just generating buzzwords and PowerPoint slides. <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>Vibecoding: the new executive superpower (or just more buzzwords)?</strong>: Emergent, an Indian AI startup, just bagged $70 million for its ‘vibe-coding’ platform, claiming executives can now 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 ‘efficient’ like a CEO fumbling with AI-generated code instead of, you know, leading. It’s almost like they’re trying to sell us a new flavor of ‘low-code/no-code’ with extra AI sprinkles. <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 model personas: banishing ‘demon’ personas with ‘activation capping’</strong>: Anthropic and other researchers have discovered an ‘Assistant Axis’ in LLMs, where models can drift into ‘fantastical or less safe personas’ during conversations. Their solution? ‘Activation capping’ to prevent this ‘persona drift’ and ‘banish ‘demon’ personas.’ Because apparently, our AI overlords are not just hallucinating, they’re having identity crises. Good to know we’re spending resources to keep our chatbots from going full evil, rather than, say, ensuring they don’t make up facts. <a href="https://app.daily.dev/posts/Jt8b0RH1S" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
