<h2>Must know</h2>
<ul>
<li><strong>European Parliament bans cloud AI on devices</strong>: The European Parliament has banned cloud-based AI features on lawmakers’ corporate devices due to data security and privacy concerns, specifically citing the inability to guarantee data security when AI assistants use cloud services. This move reflects broader European unease about reliance on U.S. tech companies and the handling of sensitive legislative data, highlighting a pragmatic approach to AI adoption that prioritizes security over convenience. On-device AI processing is seen as a potential solution. <a href="https://app.daily.dev/posts/NW6nIRqB9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Code signing policies get stricter</strong>: Major tech companies are hardening code signing policies, enforcing a new 460-day maximum validity period for certificates to reduce supply chain attack windows. This change, driven by the CA/Browser Forum, requires organizations to transition from file-based keys to FIPS-compliant hardware and automate signing workflows to avoid distribution disruptions, as reported by Latest Hacking News. <a href="https://app.daily.dev/posts/VDBUnXRHh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OWASP highlights smart contract governance risks</strong>: OWASP released its Smart Contract Top 10 2026, identifying governance failures and access control issues as dominant security risks in blockchain systems, based on hundreds of millions in losses from 2025 exploits. The framework emphasizes that passing audits alone is insufficient, advocating for upstream security integration with role-based validation and invariant-driven design, according to Latest Hacking News. <a href="https://app.daily.dev/feed-by-ids?id=U7Z7c8wKd&amp;id=WFh6lNRMq" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AMD optimizes Linux kernel for SEV-SNP servers</strong>: AMD is submitting Linux kernel patches for RMPOPT, a new instruction to reduce overhead on SEV-SNP servers by optimizing RMP checks for 1GB memory regions, as reported by Phoronix. <a href="https://app.daily.dev/posts/xpFUQM5D2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>.NET on Linux gets io_uring for massive performance gains</strong>: A pull request for .NET Runtime on Linux proposes using io_uring for socket operations, showing significant performance improvements (15-50% CPU reduction, 30-50% memory reduction) across various workloads, according to Phoronix. <a href="https://app.daily.dev/posts/XXNGjMHHi" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 optimizes <code>close_range</code> 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) by using <code>find_next_bit()</code> on the <code>open_fds</code> bitmap. <a href="https://app.daily.dev/posts/2IrhD5aAX" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>JDK 27 gets post-quantum TLS 1.3</strong>: JEP 527 introduces post-quantum hybrid key exchange to TLS 1.3 in JDK 27, combining quantum-resistant ML-KEM with traditional elliptic-curve algorithms to protect against “harvest now, decrypt later” threats. <a href="https://app.daily.dev/posts/duMdZgpLR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Go 1.26 <code>go fix</code> command gets a rewrite</strong>: Go 1.26 introduces a rewritten <code>go fix</code> command that automatically modernizes Go codebases by identifying opportunities to use newer language features and library functions, including a self-service paradigm for custom modernizers, as detailed by Go. <a href="https://app.daily.dev/posts/mInrXYefK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Temporal raises $300M for AI workflow orchestration</strong>: Temporal secured $300M Series D funding at a $5B valuation, driven by demand for Durable Execution in AI and long-running workflows, with companies like OpenAI and ADP using it for AI agent orchestration. <a href="https://app.daily.dev/posts/HqHJFpyjl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Alibaba releases Qwen3.5 multimodal AI model</strong>: Alibaba released 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, and expanding language support to 201 languages. <a href="https://app.daily.dev/posts/UOocLbfK4" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Quest Software’s AI-powered data trust factory</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. Because, you know, trust is built on AI-generated data products, and who needs actual data governance when you have a “trust scoring system” and marketing claims of 54% faster delivery and 40% cost savings? Sounds totally legit. <a href="https://app.daily.dev/posts/wlE5eAAsI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Toilet maker Toto is an ‘AI play’</strong>: Palliser Capital, an activist investor, claims Japanese toilet manufacturer Toto is an “AI play” due to its advanced ceramics division making electrostatic chucks for 3D NAND memory chips. Because nothing screams “AI innovation” like toilet parts, right? This is clearly a brilliant strategy to unlock a “$3.6 billion valuation gap” and a 55%+ upside for shareholders, by rebranding plumbing as high-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 with 100-agent swarm mode</strong>: Moonshot AI launched Kimi K2.5, an open-weight multimodal LLM with an “Agent Swarm mode” capable of orchestrating up to 100 sub-agents using “Parallel Agent Reinforcement Learning.” They claim it outperforms “frontier models” on benchmarks and achieves “significant speed improvements.” Because, obviously, if you can’t beat them with one agent, just throw 100 at it, and call it “parallel reinforcement learning.” What could possibly go wrong? <a href="https://app.daily.dev/posts/tfdJsRaBj" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
