<h2>Must know</h2>
<ul>
<li><strong>Critical ServiceNow AI agent vulnerability disclosed</strong>: AppOmni discovered ‘BodySnatcher’ (CVE-2025-12420), a critical flaw in ServiceNow’s AI agent integration that allows 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. ServiceNow has released a patch, but the incident highlights the growing attack surface with enterprise AI adoption, as reported by AppOmni. <a href="https://app.daily.dev/posts/wzAqd15ET" target="_blank" rel="noopener nofollow">Read more</a></li>
<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, officials detailed new cryptographic migration strategies to counter ‘harvest now, decrypt later’ attacks. NIST has released initial post-quantum cryptography standards, recommending hybrid methods as interim solutions, emphasizing immediate action for comprehensive security, as reported by various news outlets. <a href="https://app.daily.dev/posts/lRZsyJgWP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic’s Claude Code gets new features, restricts API access</strong>: Anthropic released ‘Cowork’ for Claude Code, enabling automation of enterprise tasks beyond software development, such as summarizing calendars and organizing data. However, Anthropic has also blocked API access for third-party applications across all subscription plans, forcing users to their official Claude Code tool. This move has caused significant developer backlash, while Anthropic simultaneously promotes open-source tooling like Claude-on-incus for secure, isolated execution of Claude Code, as detailed in their documentation and developer forums. <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>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI integration in mathematical problem-solving advances</strong>: OpenAI’s GPT 5.2 is credited with solving 11 of 15 newly resolved Erdős problems since December, and ChatGPT is increasingly used as a reasoning assistant in mathematical research, though human oversight remains critical for verification. <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>CrowdStrike acquires Seraphic Security for browser protection</strong>: CrowdStrike is acquiring Seraphic Security for approximately $400 million to integrate browser security capabilities into its Falcon platform, aiming to protect against threats like session hijacking and data exfiltration through generative AI prompts. <a href="https://app.daily.dev/posts/AKJAUyPZD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>CNCF graduates Dragonfly, MemOS launches for LLM memory</strong>: The CNCF has graduated Dragonfly, a peer-to-peer image and file distribution platform, while MemOS launched as an open-source memory operating system for LLMs, claiming significant accuracy improvements and token savings. <a href="https://app.daily.dev/feed-by-ids?id=0rVfXzPjQ&amp;id=bjqQdB8fJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>G7 urges quantum-resilient finance, IBM launches Sovereign Core</strong>: The G7 Cyber Expert Group issued guidance for financial institutions to transition to quantum-resilient cryptography, while IBM announced its Sovereign Core platform, built on Red Hat OpenShift, for rapid, compliant deployment of isolated workloads. <a href="https://app.daily.dev/feed-by-ids?id=GCSg0YbNZ&amp;id=eWD0cjmPm" 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 centralized Apache Spark-based system called FINUDP to eliminate metric inconsistencies, processing millions of transactions daily with machine learning for anomaly detection. <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>BeaglePlay board gets 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>Firefox 147 and Thunderbird 147 released with improvements</strong>: Firefox 147 introduces automatic Picture-in-Picture, enhanced WebGPU support for Apple Silicon, and stricter Enhanced Tracking Protection, while Thunderbird 147 adds folder-name localization and a ‘Show Full Path’ option. <a href="https://app.daily.dev/posts/f7pzHnDhN" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI ‘influentists’ are still peddling hype</strong>: Apparently, we’re still dealing with ‘influentists’ making viral, dramatic claims about AI without bothering with pesky things like ‘reproducible evidence’ or ‘full context.’ This, according to Hacker News, creates unrealistic expectations and shifts focus from actual technical rigor to social media virality. Who knew that people on the internet might exaggerate for clicks? Shocking. <a href="https://app.daily.dev/posts/xNvqrJb4c" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Generative AI isn’t going all that well, but don’t tell the VCs</strong>: A critical take on Hacker News suggests that generative AI is still struggling with fundamental limitations like reliability, heavy reliance on memorization, and minimal quantifiable economic value. Apparently, AI can only perform about 2.5% of jobs, and scaling is hitting diminishing returns. But hey, let’s base all our economic and geopolitical policy on this, because what could possibly go wrong? <a href="https://app.daily.dev/posts/NNn4Em1hp" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The unhinged world of tech in 2026: more of the same, but with robots</strong>: Fireship offers a satirical glimpse into 2026, where AI has plateaued but the hype cycle continues, humanoid robots are a thing, and quantum computing breakthroughs are still just around the corner. The job market is a mixed bag, with AI creating ‘code janitor’ roles while threatening mid-level positions. So, basically, it’s 2024, but with more expensive toys and the same old BS. <a href="https://app.daily.dev/posts/K65F2Co4Y" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
