<h2>Must know</h2>
<ul>
<li><strong>Anthropic’s Claude Cowork and Microsoft Copilot hit with critical vulnerabilities</strong>: Security researchers have uncovered critical vulnerabilities in both Microsoft Copilot and Anthropic’s Claude Cowork. Copilot was susceptible to data exfiltration via a single malicious URL click, allowing attackers to extract user data and chat histories, a flaw Microsoft has since patched. Claude Cowork, however, remains unpatched against file exfiltration through indirect prompt injection, where hidden prompts in uploaded files could command Cowork to upload user data to attacker-controlled servers, as reported by multiple sources. Anthropic’s response places responsibility on users, which is a bit rich given Cowork’s marketing to non-technical users. <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>
<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. NIST has released initial post-quantum cryptography standards, recommending hybrid methods as interim solutions. This multi-year transition emphasizes immediate protection of current infrastructure and proactive safeguarding of quantum research, supply chains, and workforce, as reported by various government sources. <a href="https://app.daily.dev/posts/lRZsyJgWP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>G7 urges quantum-resilient finance, IBM launches sovereign core platform</strong>: The G7 Cyber Expert Group has issued guidance for financial institutions to transition to quantum-resilient cryptography, addressing the risk quantum computers pose to current encryption protocols. This guidance, co-chaired by the U.S. Treasury and Bank of England, offers a flexible approach rather than mandatory requirements. Concurrently, IBM announced its Sovereign Core platform, built on Red Hat OpenShift, offering rapid deployment of isolated workloads to meet data sovereignty and compliance needs, as detailed in official announcements. <a href="https://app.daily.dev/feed-by-ids?id=GCSg0YbNZ&amp;id=eWD0cjmPm" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>MemOS: An open-source memory OS for LLMs and AI agents</strong>: MemOS is an open-source framework providing unified long-term memory management for LLMs and AI agents, featuring a graph-structured memory store, multi-modal data support, and claims of significant accuracy improvements and token savings over OpenAI Memory. <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>CrowdStrike acquires Seraphic Security for browser threat protection</strong>: CrowdStrike is acquiring Seraphic Security for approximately $400 million to integrate its enterprise browser security technology, which isolates JavaScript engines and randomizes memory addresses, into the Falcon security platform. <a href="https://app.daily.dev/posts/AKJAUyPZD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI integration in mathematical and scientific problem-solving</strong>: OpenAI’s GPT 5.2 is solving a significant number of Erdős problems, and tools like Harmonic’s Aristotle are formalizing mathematics, with experts like Terence Tao noting AI’s strength in addressing ‘long tail’ problems, 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>CNCF graduates Dragonfly for cloud-native image and file distribution</strong>: Dragonfly, a cloud-native peer-to-peer platform for distributing container images and AI models, has graduated from CNCF incubation, boasting widespread adoption and claims of up to 90% storage bandwidth 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>Pulumi IAM now available for self-hosted Pulumi Cloud</strong>: Self-hosted Pulumi Cloud instances now support Pulumi IAM with custom roles and scoped access tokens, enabling fine-grained permissions and Zero Trust security for CI/CD automation. <a href="https://app.daily.dev/posts/wd03nDp6A" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Intel Compute Runtime adds initial Crescent Island and Nova Lake S support</strong>: Intel’s Compute Runtime version 26.01.36711.4 introduces initial support for Crescent Island and Nova Lake S graphics hardware, maintaining Level Zero and OpenCL support for new and past Intel GPUs. <a href="https://app.daily.dev/posts/54Uqeadiv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>BeaglePlay board integrates 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>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI-powered email assistant: Because passive-aggressive bots are the future</strong>: Someone built an AI email assistant named ‘Billie’ that uses Claude AI to classify emails and draft ‘technically professional but menacing’ responses. Apparently, we’ve reached the pinnacle of human-AI collaboration: automating our disdain for spam. Because nothing says ‘efficiency’ like an AI that can perfectly mimic your passive-aggressive tendencies, as reported by The Last Week in AWS. <a href="https://app.daily.dev/posts/d94QppmRP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated code: Functionally correct, architecturally a dumpster fire</strong>: A senior Android developer reviewed AI-generated Kotlin code and found it functionally correct but architecturally problematic, riddled with issues like improper dispatcher injection and poor state management. So, while AI can churn out code, it still lacks the ‘deep understanding’ of Android specifics. Who knew that context and architectural foresight were still human domains? Philipp Lackner’s review confirms that AI is great at writing code you’ll have to rewrite. <a href="https://app.daily.dev/posts/O3swuQWEt" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Quantum computing claims challenged by a dog, literally</strong>: Researchers supposedly replicated IBM’s quantum factorization results using a dog, suggesting overhyped claims in quantum computing. While this sounds like a headline from The Onion, it highlights the ongoing skepticism around quantum breakthroughs. Meanwhile, Rice University is doing actual science with trapped-ion quantum simulators, proving molecular vibrations can steer energy transfer. So, one group is using dogs, the other is using physics. You decide who’s making real progress. <a href="https://app.daily.dev/feed-by-ids?id=7Bd1YG4JM&amp;id=FBqZ5fVz5" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
