<h2>Must know</h2>
<ul>
<li><strong>European Parliament bans cloud AI on devices</strong>: The European Parliament has disabled AI features on lawmakers’ corporate devices due to significant data security and privacy concerns. Their IT department can’t guarantee data security when AI assistants use cloud services, fearing confidential information could be processed in unknown locations, especially given many AI company servers are US-based. This move highlights a pragmatic approach to AI adoption, prioritizing security and control over convenience, despite the EU’s own AI Act. <a href="https://app.daily.dev/posts/NW6nIRqB9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI security threats escalate amidst ROI pressure</strong>: Check Point Research demonstrated how AI assistants like Grok and Microsoft Copilot can be exploited as covert Command-and-Control proxies, enabling attackers to tunnel malware communications without API keys. This comes as CIOs face immense pressure to show measurable ROI from AI investments by mid-2026, with 71% expecting budget cuts if targets aren’t met, according to a Dataiku survey. The survey also found 82% of CIOs struggle to govern employee-built AI tools, highlighting a growing security and governance gap. <a href="https://app.daily.dev/feed-by-ids?id=8kxTLeIWY&amp;id=DCxMGa8ig&amp;id=H92K14NPx&amp;id=JY689C6zi&amp;id=K2suxKLLr&amp;id=T8PHWMs0E" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents disrupt SaaS and drive protocol standardization</strong>: AI developments are threatening the traditional SaaS per-seat pricing model, potentially making it obsolete as AI intelligence becomes abundant and cheap, with major software companies reportedly losing $1 trillion in market cap. OpenAI published the architecture of its Codex App Server, a bidirectional protocol unifying all Codex coding agent experiences via a single API, while the Agent Client Protocol (ACP) aims to standardize agent-editor communication, paralleling the Language Server Protocol (LSP). This indicates a significant shift towards platform-native integration and standardized protocols for AI agents. <a href="https://app.daily.dev/feed-by-ids?id=BEYbuELEG&amp;id=LytVQeSZ8&amp;id=V16f6xKqK&amp;id=v0Pl5Xxg1&amp;id=wj4NMewlB" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Gentoo Linux migrates from GitHub due to AI concerns</strong>: Gentoo Linux is migrating its repository mirrors from GitHub to Codeberg, citing concerns over GitHub Copilot and AI-generated contributions, maintaining a strict policy against AI-assisted code due to copyright, quality, and ethical issues. <a href="https://app.daily.dev/posts/Ed9ofBB82" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Alibaba releases Qwen3.5 multimodal AI model</strong>: Alibaba launched Qwen3.5, a new multimodal AI model designed for enterprise agent workflows, claiming benchmark improvements over GPT-5.2, Claude 4.5 Opus, and Gemini 3 Pro, with expanded language support and context windows up to 1 million tokens. <a href="https://app.daily.dev/posts/UOocLbfK4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Figma integrates Claude Code for bidirectional design-code workflow</strong>: Figma now supports converting live code interfaces from Claude Code into editable Figma designs, enabling a bidirectional workflow from code-first exploration to collaborative canvas-based design iteration. <a href="https://app.daily.dev/feed-by-ids?id=iBUqWA8OS&amp;id=jwuzwzVmy" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Python Security Response Team seeks new members</strong>: The Python Security Response Team (PSRT) is actively seeking new members with security expertise to help coordinate vulnerability reports and remediations for CPython and pip, following the establishment of formal governance via PEP 811. <a href="https://app.daily.dev/posts/k8sHlR5Xv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 optimizes close_range system call</strong>: Linux 7.0 introduces a performance optimization for the <code>close_range</code> system call, reducing its complexity from O(range size) to O(active file descriptors) for significantly faster operations on sparse file descriptor tables. <a href="https://app.daily.dev/posts/2IrhD5aAX" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>MongoDB Java driver sees 20-90% throughput improvements</strong>: MongoDB’s Java team achieved significant throughput improvements in their Java driver (20-90%) by profiling and optimizing hot paths, including caching BSON array indexes and leveraging JVM intrinsics. <a href="https://app.daily.dev/feed-by-ids?id=0JRJsAJbi&amp;id=6dgAlz6XB&amp;id=C9og01r53&amp;id=Ii1csxjgn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GlassFish 8.0 released with Jakarta EE 11 compatibility</strong>: GlassFish 8.0 has been released as a compatible implementation of Jakarta EE 11, requiring JDK 21+ and featuring virtual thread support, Jakarta Data 1.0 integration, and enhanced security. <a href="https://app.daily.dev/posts/c8IbRHUTq" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Quest Software launches AI-powered data management platform</strong>: Quest Software launched its ‘Trusted Data Management Platform’ with an ‘Automated Data Product Factory’ that uses generative AI to create data products from natural language prompts. They claim it delivers up to 54% faster time to data product delivery and 40% cost savings. Because, you know, just <em>trust</em> them on those numbers; no independent verification needed when AI is involved, right? <a href="https://app.daily.dev/posts/wlE5eAAsI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Palliser Capital claims Toto undervalued as AI play</strong>: Activist investor Palliser Capital is trying to convince everyone that Japanese toilet manufacturer Toto is an ‘AI play’ because its advanced ceramics division makes electrostatic chucks for 3D NAND memory chips. Apparently, a toilet maker is an ‘overlooked AI play’ and has a ‘$3.6 billion valuation gap.’ It’s almost like they’re trying to pump up a stock by slapping ‘AI’ on anything that vaguely touches tech. <a href="https://app.daily.dev/posts/08TkDSLKo" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Moonshot AI launches Kimi K2.5 multimodal LLM</strong>: Moonshot AI launched Kimi K2.5, an open-weight multimodal LLM, claiming its coding performance is ‘comparable to GPT-5 and Gemini’ and that it ‘outperformed frontier models’ on research benchmarks. These are, of course, self-reported performance claims from Moonshot AI, so you can take them with a grain of salt. Because when has a company ever exaggerated its own AI’s capabilities? <a href="https://app.daily.dev/posts/tfdJsRaBj" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
