<h2>Must know</h2>
<ul>
<li><strong>Anthropic and OpenAI clash over AI ads</strong>: Anthropic is planning Super Bowl ads to mock OpenAI’s move to introduce advertising in ChatGPT, positioning its Claude AI as ad-free. OpenAI CEO Sam Altman fired back on social media, calling Anthropic’s ads dishonest and defending OpenAI’s approach, while also labeling Anthropic ‘authoritarian.’ This public spat highlights the diverging business models and intense rivalry between the two AI giants, with Anthropic emphasizing an ad-free experience for ‘work and deep thinking’ and OpenAI testing ads in lower-cost tiers. <a href="https://app.daily.dev/posts/TmM2mNSSP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent security vulnerabilities are critical and widespread</strong>: Snyk’s ‘ToxicSkills’ research found that over a third of AI agent skills contain security flaws, with 13.4% having critical issues and 76 confirmed malicious payloads designed for credential theft and data exfiltration. This highlights a significant supply chain risk in the rapidly expanding AI agent ecosystem, affecting platforms like OpenClaw and Claude Code. Operant AI launched ‘Agent Protector’ to address these risks with real-time security for autonomous agents, emphasizing the need for platform providers to embed security rather than leaving it to customers. <a href="https://app.daily.dev/feed-by-ids?id=0BdgmYNfg&amp;id=0e28QTlYb&amp;id=2vNCw5dB4&amp;id=dL9ys7HAk&amp;id=the2EenMO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft appoints engineering quality czar amid ongoing issues</strong>: Microsoft CEO Satya Nadella appointed Charlie Bell as the new engineering quality czar, reporting directly to him. This move aims to address persistent problems like Azure outages, Windows patch issues, and low Copilot adoption. Bell, formerly in security leadership, will focus on improving engineering quality across all Microsoft products, with Hayete Gallot returning to take over his previous security role. <a href="https://app.daily.dev/posts/NGytU6zYE" 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 and massive context</strong>: Anthropic launched Claude Opus 4.6, featuring ‘agent teams’ for parallel task execution, an expanded 1 million token context window, and new integrations with Amazon Bedrock, Vercel, and Google Vertex AI, 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>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>AI bot traffic surges, threatening publishers</strong>: AI bots, primarily RAG bots from tools like ChatGPT, now account for a significant portion of web traffic, leading to a sharp decline in clickthrough rates to publishers and creating an ‘arms race’ as publishers deploy defenses against sophisticated bot tactics. <a href="https://app.daily.dev/posts/5O2tzW5oh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft develops LLM backdoor scanner</strong>: Microsoft has developed a scanner to detect malicious triggers and backdoors embedded in open-weight AI models during training, operating without retraining or gradient calculations and working across most GPT-style language models. <a href="https://app.daily.dev/posts/O9jvh85A8" 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 into Laravel applications with support for multiple providers like Anthropic, Gemini, and OpenAI, and features like agent building, memory, and structured outputs. <a href="https://app.daily.dev/posts/ND9RAWZ9f" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Zed editor offers optional AI and real-time collaboration</strong>: Zed, a Rust-based code editor, prioritizes speed and collaboration, offering optional Claude AI integration via ACP for focused suggestions and built-in real-time multiplayer features like live cursors and audio chat for large codebases. <a href="https://app.daily.dev/posts/Kp2ULcsBG" 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, reportedly 10x faster than Facebook’s Faiss. <a href="https://app.daily.dev/posts/pKNE7iWlw" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Elixir is the ‘best language for AI’ according to an 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. Apparently, its immutability and decade-long stability make it perfect for AI, because who needs actual evidence or comparative analysis when you have a mysterious study and a strong pro-Elixir bias? The article conveniently omits any mention of other languages or potential drawbacks, because why complicate a good marketing pitch? <a href="https://app.daily.dev/posts/1D7tGyRwk" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>UK’s 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 quickly dismissed it as unlikely to produce systemic improvements, stating that existing open-source tools already track AI-generated content. He argues the real challenge is enforcement, not detection, and that voluntary codes of conduct are about as effective as a screen door on a submarine. So, while the government is busy evaluating the tools, the actual problem of deepfake proliferation continues unabated. <a href="https://app.daily.dev/posts/EQCYBDiDB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents are the new insider threat, apparently</strong>: A survey of IT executives suggests 1.5 million AI agents are running around without monitoring, posing a significant security risk. Security experts are dramatically comparing this to the Titanic disaster, warning that these ‘invisible’ agents, deployed via low-code tools, are acting as insider threats. Because, you know, an unmonitored script is totally the same as a rogue AI with malicious intent. It’s almost like we should have proper governance and monitoring for <em>all</em> our deployed software, not just the ones with ‘AI’ in the name. <a href="https://app.daily.dev/posts/RMzBzQnM3" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
