<h2>Must know</h2>
<ul>
<li><strong>LLM safety guardrails vulnerable to training prompt exploitation</strong>: Microsoft researchers found a single, seemingly mild training prompt can bypass safety guardrails in 15 different LLMs, including text-to-image generators. This GRP-Obliteration technique exploits reinforcement learning to unalign models post-training, making them permissive across harmful categories. This is a significant concern for enterprises fine-tuning open-weight models, highlighting that continuous safety evaluation is crucial. <a href="https://app.daily.dev/posts/6Gbu8zodz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google shares journalist data with ICE</strong>: Google provided ICE with extensive personal data of a student journalist, Amandla Thomas-Johnson, including bank details, IP addresses, and phone numbers, without prior notice. This data was released in response to an administrative subpoena, which does not require judicial approval. The Electronic Frontier Foundation (EFF) and ACLU are calling on tech companies to resist such requests and notify users before complying. <a href="https://app.daily.dev/posts/nszZi8Hk4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Docker hardened images now free</strong>: Docker has made its Hardened Images (DHI) free, covering over 1,000 images including Alpine, Debian, databases, and runtimes. DHI aims to establish a security “waterline” by managing vulnerabilities below the application layer and providing supply chain isolation through controlled rebuilds with VEX attestations for automated CVE suppression. This move is intended to reduce vulnerability noise and support compliance frameworks. <a href="https://app.daily.dev/posts/qrz4zhyW1" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Angular integrates AI for generative UIs</strong>: Angular 21.1 now features AI-assisted development, including Gemini integration for generative UIs, Signal Forms, and Angular Aria. <a href="https://app.daily.dev/posts/55GIMu6dJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Waku 1.0 alpha stabilizes APIs</strong>: Waku 1.0 alpha, a minimal React framework built on Vite and Hono, has stabilized its public APIs, focusing on bug fixes and compatibility with React 19 features like server components. <a href="https://app.daily.dev/posts/AI6v3cv3M" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>CrossOver 26 released with Wine 11.0</strong>: CodeWeavers released CrossOver 26, built on Wine 11.0, enabling Windows applications and games to run on Linux and macOS with NTSync support and enhanced Wayland compatibility. <a href="https://app.daily.dev/posts/Rhd5Rku0N" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Node.js releases current and LTS updates</strong>: Node.js released versions 25.6.1 (Current) and 24.13.1 (LTS), bringing dependency updates, bug fixes, and security patches, including Python 3.14 build support for LTS. <a href="https://app.daily.dev/posts/CeFkKNaVQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Jakarta EE 12 M2 introduces unified query language</strong>: Jakarta EE 12 Milestone 2 introduces Jakarta Query, a unified query language for SQL and NoSQL databases, and establishes Java 21 as its baseline. <a href="https://app.daily.dev/posts/y8GIbPRTH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Amazon Redshift enhances automatic optimizations</strong>: Amazon Redshift now allows allocating extra compute resources for automatic optimization features like Automatic Table Optimization, ensuring user workloads are not impacted. <a href="https://app.daily.dev/posts/7o3DKKn6E" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LinkedIn revamps service discovery with Kafka and gRPC</strong>: LinkedIn replaced its Zookeeper-based service discovery with a new Kafka and gRPC/xDS architecture, achieving 10x better P50 latency and 6x better P99 latency. <a href="https://app.daily.dev/posts/WAH3amRwM" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>GitHub co-founder predicts AI will shrink open source communities</strong>: Scott Chacon, GitHub co-founder, predicts AI will make developers generate custom code snippets instead of importing bloated libraries, shrinking traditional open-source communities. Because, clearly, the problem with open source was too many libraries, not, say, maintenance, funding, or actual community building. Abstraction levels will rise, so we can all focus on ‘architecture’ while AI handles the ‘implementation details’—what could possibly go wrong? <a href="https://app.daily.dev/posts/L0XijwnBC" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI reduces IPv6 migration stress, but at what cost?</strong>: A university experiment found generative AI reduced cognitive load on network engineers during IPv4 to IPv6 migrations by 65-72%, cutting completion times from hours to seconds. The CIO, however, warns about AI cost concerns, including token pricing changes, vendor lock-in, and the high expense of on-premises alternatives. So, AI makes things faster, but the real cost is still a mystery wrapped in an enigma, paid for by your cloud bill. <a href="https://app.daily.dev/posts/e9OZK0ZNf" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI tools lead to burnout and expanded to-do lists</strong>: A Harvard Business Review study found AI tools led to burnout and expanded to-do lists for tech workers, with work bleeding into personal hours, despite no management pressure. Employees felt compelled to do more because AI made it possible, tripling expectations while productivity gains remained modest. So, AI isn’t making us more efficient; it’s just making us feel guilty for not doing more, faster, for the same pay. Progress! <a href="https://app.daily.dev/posts/RvA4dXVbd" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
