<h2>Must know</h2>
<ul>
<li><strong>Unmonitored AI agents pose significant security risk</strong>: A survey of 750 IT executives reveals that 1.5 million out of 3 million deployed AI agents lack active monitoring or security oversight, leading to 88% of executives experiencing or suspecting AI agent-related security incidents. Security experts warn these unmonitored agents act as an insider threat due to delegated credentials and persistent access, bypassing traditional IT scrutiny. Operant AI launched Agent Protector to provide real-time security for autonomous AI agents, offering shadow agent discovery and inline protection against privilege escalation and data exfiltration. <a href="https://app.daily.dev/feed-by-ids?id=RMzBzQnM3&amp;id=0e28QTlYb&amp;id=2vNCw5dB4&amp;id=dL9ys7HAk&amp;id=the2EenMO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Global cyberespionage campaign targets critical infrastructure</strong>: A sophisticated nation-state threat group, TGR-STA-1030, has been conducting a global cyberespionage campaign since January 2024, compromising government and critical infrastructure in 37 countries. The group uses phishing and N-day exploits, deploying tools like Cobalt Strike and a new Linux kernel rootkit named ShadowGuard, which uses eBPF to hide activities. Unit 42 researchers uncovered the operations, and CISA is working to mitigate the threat. <a href="https://app.daily.dev/posts/KNu0fQxJH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Substack user data breach disclosed</strong>: Substack disclosed a data breach from October 2025 that exposed user email addresses, phone numbers, and internal metadata, remaining undetected for four to five months. While passwords and financial information were not compromised, a threat actor claims nearly 700,000 records were affected. Substack has fixed the vulnerability and is investigating, warning users to watch for phishing attempts. <a href="https://app.daily.dev/posts/myrJIoXYs" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Valve delays hardware due to component shortages</strong>: Valve has delayed pricing and shipping announcements for its Steam Machine and Steam Frame VR headset, attributing the delays to memory and storage shortages driven by AI data center demand. <a href="https://app.daily.dev/posts/iXYPYa4qJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AWS launches new EC2 instances</strong>: AWS launched new EC2 High Memory U7i-6TB instances with custom Intel Xeon Scalable Processors and new C8id, M8id, and R8id instances powered by custom Intel Xeon 6 processors, offering significant performance and memory bandwidth improvements. <a href="https://app.daily.dev/posts/JPt4mjvME" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ScyllaDB integrates USearch for vector search</strong>: ScyllaDB now includes vector search capabilities using the open-source USearch library, allowing storage of vector embeddings alongside structured data and leveraging ScyllaDB’s shard-per-core architecture for optimized performance. <a href="https://app.daily.dev/posts/pKNE7iWlw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Intel Arc B390 Linux performance analysis</strong>: Intel Arc B390 graphics with Xe3 architecture show competitive performance on Linux when properly configured, but out-of-the-box experience is hampered by lower default power limits compared to Windows. <a href="https://app.daily.dev/posts/GxsXcKIKI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ubuntu 24.04.4 LTS HWE update released</strong>: The Ubuntu 24.04.4 LTS hardware enablement update is now available, featuring Linux kernel 6.17 for improved power management and initial RDNA 4 support, along with Mesa 25.2.7 for full Vulkan 1.4 compliance. <a href="https://app.daily.dev/posts/7U8PoA0ga" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ardour 9.0 DAW released</strong>: Ardour 9.0, an open-source Digital Audio Workstation, was released with multi-touch GUI support and numerous new functionalities for musicians and audio professionals on Linux and other platforms. <a href="https://app.daily.dev/posts/AJXI8G1Z7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Intel ISPC 1.30 released with AMX support</strong>: Intel released ISPC (Implicit SPMD Program Compiler) version 1.30, which now includes AMX (Advanced Matrix Extensions) support in its standard library, simplifying parallel processing workloads on Intel CPUs and GPUs. <a href="https://app.daily.dev/posts/SaAG9DKe2" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic and OpenAI chatbot ad controversy escalates</strong>: Anthropic is planning Super Bowl ads to mock OpenAI’s decision to introduce advertising in ChatGPT, positioning its Claude AI as ad-free. OpenAI CEO Sam Altman, ever the diplomat, responded by calling Anthropic’s ads ‘dishonest’ and ‘authoritarian,’ claiming they only serve wealthy users. Because nothing says ‘helpful assistant for deep thinking’ like a multi-million dollar ad campaign about not having ads. <a href="https://app.daily.dev/posts/TmM2mNSSP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Elixir is the ‘best language for AI,’ apparently</strong>: An article claims Elixir is the ‘best language for AI,’ citing an unverified ‘Tencent study’ that supposedly found Elixir achieved a 97.5% LLM completion rate. Apparently, Elixir’s immutability and decade-long stability are revolutionary for AI, unlike all those other languages that just can’t keep up. Who needs actual benchmarks when you have a ‘Tencent study’ and a strong pro-Elixir bias? <a href="https://app.daily.dev/posts/1D7tGyRwk" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>UK deepfake detection plan criticized as ‘systemic improvements’ unlikely</strong>: The UK government announced plans to develop a framework for evaluating deepfake detection technologies, partnering with Microsoft and academics. Cybersecurity expert Dr. Ilia Kolochenko, however, states this initiative is unlikely to produce ‘systemic improvements,’ arguing that existing open-source tools already track AI-generated content. Because clearly, the problem isn’t detection, it’s the lack of global legislative amendments and the naive belief that clandestine media sources will cooperate with takedown requests. <a href="https://app.daily.dev/posts/EQCYBDiDB" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
