<h2>Must know</h2>
<ul>
<li><strong>US agencies push post-quantum security transition</strong>: U.S. federal agencies, including the FBI, NIST, and CISA, are moving beyond awareness to active implementation of post-quantum security measures. At the YQS2026 conference, new cryptographic migration strategies were detailed to counter ‘harvest now, decrypt later’ attacks, as reported by various news outlets. NIST has released initial post-quantum cryptography standards, recommending hybrid methods as interim solutions and emphasizing immediate action for cryptographic inventory and integration. <a href="https://app.daily.dev/posts/lRZsyJgWP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical ServiceNow AI agent vulnerability disclosed</strong>: AppOmni disclosed ‘BodySnatcher’ (CVE-2025-12420), a critical ServiceNow security flaw allowing unauthenticated attackers to impersonate users with just an email address, bypassing MFA and SSO. This could lead to rogue AI agents with escalated privileges accessing external environments via ServiceNow’s Virtual Agent API. ServiceNow has confirmed the vulnerability and released an immediate patch, as detailed by AppOmni’s report. <a href="https://app.daily.dev/posts/wzAqd15ET" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>CrowdStrike acquires Seraphic Security for browser protection</strong>: CrowdStrike is acquiring Seraphic Security for approximately $400 million, marking its fourth acquisition since August. Seraphic specializes in enterprise browser security, using an abstraction layer to isolate JavaScript engines and randomize memory addresses against exploits. CrowdStrike plans to integrate Seraphic into its Falcon platform, enhancing real-time correlation of endpoint and browser activity, as reported by official announcements. <a href="https://app.daily.dev/posts/AKJAUyPZD" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>React frameworks converge, AI integration on the horizon</strong>: At React Conf, framework creators noted a convergence in features, making developer choice less risky, with all major frameworks exploring AI integration via Model Context Protocol (MCP) endpoints. <a href="https://app.daily.dev/feed-by-ids?id=Ii8I7s3X0&amp;id=oWHYd6BqQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Agoda centralizes financial data with Apache Spark</strong>: Agoda consolidated its financial data pipelines into a unified Apache Spark-based platform, FINUDP, to eliminate metric inconsistencies and process millions of daily transactions with hourly updates. <a href="https://app.daily.dev/feed-by-ids?id=SIYRILab7&amp;id=WuRILab7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Firefox 147 enhances privacy and performance</strong>: Firefox 147 introduces automatic Picture-in-Picture, reorganized tab settings, enhanced WebGPU support for Apple Silicon, zero-copy hardware video playback for Windows AMD users, and stricter Enhanced Tracking Protection. <a href="https://app.daily.dev/posts/f7pzHnDhN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Livewire 4 introduces single-file components and performance boosts</strong>: Livewire 4 is out with single-file components, an ‘islands’ feature for isolated rendering, native drag-and-drop, and View Transitions API integration, all aimed at improving performance and developer flexibility. <a href="https://app.daily.dev/posts/OxhgZzaoQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>BeaglePlay board gets fully open-source PowerVR graphics</strong>: The BeaglePlay board now supports fully open-source, upstream PowerVR graphics drivers, addressing long-standing demands from the Linux community for transparency and improved compatibility. <a href="https://app.daily.dev/feed-by-ids?id=HwIiwOf9Q&amp;id=NjVG9GX8g" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>CNCF graduates Dragonfly for cloud-native distribution</strong>: Dragonfly, a peer-to-peer image and file distribution platform, has graduated from CNCF, signifying its production readiness and widespread adoption for container images, OCI artifacts, and AI models. <a href="https://app.daily.dev/posts/0rVfXzPjQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>MemOS launches for memory-native AI agents</strong>: MemOS, an open-source memory operating system for LLMs and AI agents, provides unified long-term memory management for multi-modal data, claiming significant accuracy improvements and token savings. <a href="https://app.daily.dev/posts/bjqQdB8fJ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI solves math problems, but humans still needed for the ‘hard’ stuff</strong>: OpenAI’s GPT 5.2 is apparently solving a bunch of Erdős problems, and mathematicians like Terence Tao are saying AI is great for the ‘long tail’ of obscure, straightforward math. So, AI can do the grunt work, but don’t worry, humans are still essential for ‘verifying proofs’ and ‘complex cases.’ It’s almost like they’re saying AI is a glorified calculator, but with more hype. <a href="https://app.daily.dev/feed-by-ids?id=8o6H9wDXV&amp;id=mhyD4ij4A" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic’s Cowork: more features, less developer freedom</strong>: Anthropic released Cowork, a new feature for Claude Code to automate enterprise tasks, but simultaneously blocked API access for third-party applications from all subscription plans. So, they give you a shiny new tool while taking away your ability to integrate Claude with your existing developer tools. It’s a classic move: ‘Here’s a new toy, but you can only play with it our way.’ <a href="https://app.daily.dev/feed-by-ids?id=GMZHoSF60&amp;id=QvZHdydfi&amp;id=U2HzInRUP&amp;id=ejPwSyHLL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI assistant vulnerabilities: because who needs security when you have ‘AI’?</strong>: Apparently, Microsoft Copilot had a vulnerability that let attackers exfiltrate user data with a single malicious URL click, and Claude Cowork is still vulnerable to indirect prompt injection for file exfiltration. Microsoft patched theirs, but Anthropic’s response is basically ‘users should monitor for suspicious activity.’ It’s reassuring to know that these ‘advanced’ AI assistants are as secure as a wet paper bag, and the vendors are just shrugging. <a href="https://app.daily.dev/feed-by-ids?id=G7pZqTe0S&amp;id=jaKSkIh92&amp;id=rAHVvRPEj" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
