<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 for enterprises using third-party LLMs and agents, with affected platforms including 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. <a href="https://app.daily.dev/feed-by-ids?id=0BdgmYNfg&amp;id=0e28QTlYb&amp;id=2vNCw5dB4&amp;id=dL9ys7HAk&amp;id=the2EenMO&amp;id=RMzBzQnM3&amp;id=YLMZjOxxP&amp;id=Kcn4LiMDb&amp;id=Le8LdoPAn&amp;id=HQBjOB0ei" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic and OpenAI escalate 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 ad principles and labeling Anthropic as ‘authoritarian.’ This public spat highlights differing business models and intense rivalry, with Anthropic relying on enterprise contracts while OpenAI reportedly burns $9 billion annually. <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 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, N-day exploits, and a new Linux kernel rootkit, ShadowGuard, to exfiltrate sensitive data. 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>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Anthropic releases Claude Opus 4.6 with agent teams</strong>: Anthropic launched Claude Opus 4.6, featuring ‘agent teams’ for parallel task execution, a 1 million token context window, and new PowerPoint integrations, now available on Amazon Bedrock, Vercel’s AI Gateway, and Google’s Vertex AI. <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>OpenAI launches enterprise AI agent platform ‘Frontier’</strong>: OpenAI introduced ‘Frontier,’ an end-to-end 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>Microsoft appoints engineering quality czar</strong>: Microsoft CEO Satya Nadella appointed Charlie Bell as engineering quality czar to improve product quality across all Microsoft offerings, addressing issues like Azure outages and Windows patch problems. <a href="https://app.daily.dev/posts/NGytU6zYE" 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 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>Substack user data breach disclosed</strong>: Substack disclosed a data breach from October 2025 that exposed user email addresses, phone numbers, and internal metadata, though passwords and financial information were not compromised. <a href="https://app.daily.dev/posts/myrJIoXYs" 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 for optimized real-time AI applications. <a href="https://app.daily.dev/posts/pKNE7iWlw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Laravel launches AI SDK for application development</strong>: Laravel released an official AI SDK in beta, providing a unified API for integrating AI capabilities from multiple providers like Anthropic, Gemini, and OpenAI into Laravel applications, supporting agents, memory, and structured outputs. <a href="https://app.daily.dev/posts/ND9RAWZ9f" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI bot traffic surges, threatening publishers</strong>: AI bots now account for one visit for every 31 human visits, primarily driven by RAG bots, causing clickthrough rates to publisher sites to plummet to 0.27%. The internet is ‘reshaping’ as AI users initiate searches directly within AI platforms, threatening publishers’ financial viability. Apparently, nobody saw this coming when we started feeding all the internet’s content into black boxes. <a href="https://app.daily.dev/posts/5O2tzW5oh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Elixir is the ‘best language for AI’</strong>: An article claims Elixir is the ‘best language for AI’ based on an unverified ‘Tencent study’ that supposedly found a 97.5% LLM completion rate. It touts Elixir’s immutability and stability as key benefits. Because, you know, a single, unsourced study is all you need to declare a language superior for an entire field, especially when it’s your own product. <a href="https://app.daily.dev/posts/1D7tGyRwk" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Wirth’s law and LLM efficiency concerns</strong>: The article ‘Wirth’s Revenge’ suggests LLMs are creating a ‘bad’ Wirth tradeoff, where software gets slower faster than hardware gets faster, leading to ‘astronomical computational costs for simple tasks.’ It also warns LLMs may erode fundamental programming skills. Who knew repeatedly querying an expensive model for deterministic solutions might be inefficient? Shocking. <a href="https://app.daily.dev/posts/SELxucgnM" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
