<h2>Must know</h2>
<ul>
<li><strong>AI agents are reputation farming open source projects</strong>: Security experts are warning about ‘reputation farming’ by AI agents like ‘Kai Gritun,’ which submitted over 100 pull requests to nearly 100 open-source repositories in days. This activity, which doesn’t disclose the AI’s nature, aims to build trust for potential future supply chain attacks, shifting the attack surface from code to governance processes. This echoes the XZ-utils backdoor incident, but with AI, trust can be accumulated rapidly, as reported by Socket. <a href="https://app.daily.dev/feed-by-ids?id=D9JdnEIyW&amp;id=QHtbtBXdQ&amp;id=UNO27M8rA&amp;id=VGEsVlZIo" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Chrome extensions caught exfiltrating browsing data</strong>: A security researcher uncovered 287 Chrome extensions, with a combined 37 million installs, secretly transmitting users’ browsing histories. The data exfiltration was hidden using encryption and obfuscation, affecting VPNs, productivity tools, and shopping add-ons, sending data to brokers and unknown operators. This highlights a significant privacy risk in widely used browser tools. <a href="https://app.daily.dev/posts/G9AfdVV0Z" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated malware exploits Docker honeypot</strong>: Attackers successfully exploited a React2Shell vulnerability in a Docker honeypot using LLM-generated malware. This incident demonstrates how AI tools can lower the barrier to entry for cyberattacks, enabling less skilled operators to build exploitation frameworks. Security experts warn this signifies a shift in the threat landscape, emphasizing the need for hardening exposed services and runtime behavior monitoring. <a href="https://app.daily.dev/posts/hzvvYMQze" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI’s impact on software development and developer roles</strong>: AI coding tools are fundamentally changing software development, shifting developer roles from pure coders to orchestrators and validators, with poor documentation posing a significant risk as AI relies on it. <a href="https://app.daily.dev/feed-by-ids?id=2XGWtlWnL&amp;id=JV1i29nbu&amp;id=WYcCi2eVO&amp;id=Wvk48DEZ7&amp;id=dVvuHhcvo&amp;id=oZ8mOxinB&amp;id=qtVi65Njq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA claims massive performance gains for agentic AI</strong>: NVIDIA’s Blackwell Ultra platform (GB300 NVL72) reportedly achieves up to 50x higher throughput and 35x lower cost per token for agentic AI workloads compared to Hopper, driven by hardware and software optimizations. <a href="https://app.daily.dev/posts/XLkxKsrjj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents drive terminal tool renaissance and cloud automation</strong>: Terminal tools are seeing a resurgence with Rust-based CLIs and GPU-accelerated emulators, while major cloud providers offer official MCP servers for AI agents to automate cloud operations using natural language commands. <a href="https://app.daily.dev/feed-by-ids?id=DEJVonMxG&amp;id=Nvd4waKyi" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI security risks and governance challenges escalate</strong>: Prompt injection attacks are evolving into serious DevOps threats, with AI coding assistants vulnerable to malicious instructions, and the Model Context Protocol (MCP) facing significant security and privacy challenges. <a href="https://app.daily.dev/feed-by-ids?id=42OhpINI3&amp;id=BiRxQ3qBx&amp;id=XBLlvLDf1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Java null safety and performance upgrade coming</strong>: Java is introducing null safety through Project Valhalla, planned for 2026, aiming for C-like performance with null-restricted value types and memory-efficient data structures while maintaining backward compatibility. <a href="https://app.daily.dev/posts/YYmRfNhUM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cisco developing VMware alternative for UC applications</strong>: Cisco is developing NFVIS-for-UC, a lightweight hypervisor specifically for its Unified Communications applications, to provide an alternative to VMware’s increasingly costly Cloud Foundation bundle under Broadcom. <a href="https://app.daily.dev/posts/IW24Gs1vm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PostgreSQL diagnostic toolkit released</strong>: postgres_dba 7.0, a psql-based diagnostic toolkit for PostgreSQL, was released with 34 diagnostic reports for bloat analysis, vacuum monitoring, index health, and more, requiring no extensions and compatible with versions 13-18. <a href="https://app.daily.dev/feed-by-ids?id=A4pYssSEb&amp;id=sS4pUElCH" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI will write ‘essentially all code’ in 12 months, says Anthropic CEO</strong>: Anthropic’s CEO predicted a year ago that AI would write ‘essentially all code’ within 12 months. With one month left, some developers claim they’re already there, while others point to experiments where multi-agent AI systems building a SQLite-like database still had 54.5% coordination overhead. So, ‘essentially all code’ might mean ‘a lot of code, plus a lot of AI babysitting.’ <a href="https://app.daily.dev/feed-by-ids?id=pbRQZ4xYf&amp;id=wVbWM6nJH&amp;id=xfda4p1KO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated C compiler fails basic tests, but hey, it’s ‘progress’</strong>: Claude’s AI-generated C compiler, developed over two weeks, can’t even compile ‘hello world’ on most Linux distros, produces assembly syntax errors for the Linux kernel, and runs SQLite queries 158,000x slower than GCC. Yet, some ‘experts’ predict these tools will become widespread, leading to buggier, less secure software. Because who needs functional code when you have AI-generated code? <a href="https://app.daily.dev/feed-by-ids?id=QCdIy73we&amp;id=ijXBeCZ92&amp;id=oHgSGtpkl&amp;id=qpa1K5xV7&amp;id=wQAJlLNnu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI education tools: more dashboards, less learning</strong>: AI education tools and learning dashboards are being criticized for over-optimizing measurable outcomes at the expense of human connection, creativity, and deep engagement. Apparently, quantifying education through metrics leads to declining student interest and teacher burnout. Because nothing says ‘meaningful learning’ like a dashboard telling you how much ‘engagement’ your child isn’t having. <a href="https://app.daily.dev/feed-by-ids?id=2eDP2Fyot&amp;id=zJAFJHd8H" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
