<h2>Must know</h2>
<ul>
<li><strong>AI agent security vulnerabilities are critical</strong>: Snyk’s ToxicSkills research found that 13.4% of AI agent skills have critical security issues, with 76 confirmed malicious payloads designed for credential theft and data exfiltration. This highlights a significant supply chain risk in AI agent development, affecting platforms like OpenClaw and Claude Code. Operant AI launched Agent Protector to provide real-time security for autonomous AI agents, focusing on shadow agent discovery and inline protection against privilege escalation and data exfiltration. Meanwhile, a survey of IT executives revealed 1.5 million AI agents lack monitoring, leading to widespread security incidents. <a href="https://app.daily.dev/feed-by-ids?id=0BdgmYNfg&amp;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>Anthropic and OpenAI escalate AI chatbot ad war</strong>: Anthropic plans Super Bowl ads mocking OpenAI’s move to introduce advertising in ChatGPT, positioning its Claude AI as ad-free. OpenAI CEO Sam Altman called Anthropic’s ads dishonest, defending their approach while Anthropic highlights its enterprise-focused, ad-free model. This public spat underscores differing business models and competitive tension, with Anthropic emphasizing Claude’s role as a helpful assistant for work and deep thinking, free from commercial interruptions. <a href="https://app.daily.dev/posts/TmM2mNSSP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Global cyberespionage campaign targets critical infrastructure</strong>: A sophisticated nation-state 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, N-day exploits, and a new Linux kernel rootkit, ShadowGuard, to exfiltrate sensitive data. This ongoing threat highlights the need for robust cybersecurity measures and vigilance against advanced persistent threats. <a href="https://app.daily.dev/posts/KNu0fQxJH" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>OpenAI launches enterprise AI agent platform ‘Frontier’</strong>: OpenAI introduced ‘Frontier,’ a platform for enterprises to build, deploy, and manage AI agents, allowing them to connect to external data and applications with controlled access and permissions. <a href="https://app.daily.dev/posts/TGZyay3jK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic releases Claude Opus 4.6 with agent teams and expanded context</strong>: Anthropic launched Opus 4.6, featuring ‘agent teams’ for parallel task execution, a 1 million token context window, and new PowerPoint integrations, aiming to broaden its appeal beyond developers to knowledge workers. <a href="https://app.daily.dev/feed-by-ids?id=G7Sre82A1&amp;id=Ijw4mCG7B" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft appoints engineering quality czar</strong>: Microsoft CEO Satya Nadella appointed Charlie Bell as engineering quality czar to improve product quality across the board, addressing issues like Azure outages and low Copilot adoption. <a href="https://app.daily.dev/posts/NGytU6zYE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI bot traffic surges, impacting web publishers</strong>: AI bots now account for a significant portion of web traffic, primarily for retrieval augmented generation (RAG), leading to a sharp decline in clickthrough rates to publishers and reshaping internet usage. <a href="https://app.daily.dev/posts/5O2tzW5oh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AWS launches new EC2 instances</strong>: AWS released new EC2 High Memory U7i-6TB instances with Intel Xeon Sapphire Rapids 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>Databricks introduces MemAlign for LLM evaluation</strong>: Databricks launched MemAlign, a new MLflow framework that uses a dual-memory system to reduce the cost and latency of training LLM-based judges, enabling rapid adaptation to new domains. <a href="https://app.daily.dev/posts/eCdeDEih6" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft OneDrive Copilot agents generally available</strong>: Microsoft made OneDrive agents generally available, allowing users to query up to 20 documents simultaneously via Copilot, with agents saved as ‘.agent’ files for sharing, though privacy concerns regarding data handling remain unaddressed. <a href="https://app.daily.dev/posts/0UIAIJz1z" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Elixir is the ‘best language for AI’ says unverified study</strong>: An article boldly claims Elixir is the ‘best language for AI,’ citing an unnamed ‘Tencent study’ that supposedly found Elixir achieved a 97.5% LLM completion rate. Because nothing says ‘scientific rigor’ like an anonymous study from a major tech company that just happens to perfectly align with your product’s marketing. Sure, Elixir’s great, but let’s not pretend a secret Tencent report is the new peer review. <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 insufficient</strong>: The UK government announced plans to develop a framework for evaluating deepfake detection technologies, partnering with Microsoft and academics. However, cybersecurity expert Dr. Ilia Kolochenko dismisses this as unlikely to produce systemic improvements, stating the real challenge is enforcement, not detection. Because clearly, the solution to a global problem is another committee and some academic papers, rather than, you know, actual legislation or enforcement. <a href="https://app.daily.dev/posts/EQCYBDiDB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI will make software slower, not faster</strong>: A new article, ‘Wirth’s Revenge,’ suggests LLMs are leading us into a problematic era where software gets slower faster than hardware improves, eroding fundamental programming skills. Apparently, repeatedly querying expensive models for simple tasks is the pinnacle of efficiency. Who needs deterministic solutions when you can just throw more compute at it and call it ‘AI-powered’? <a href="https://app.daily.dev/posts/SELxucgnM" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
