<h2>Must know</h2>
<ul>
<li><strong>AI agent security is a mess</strong>: New research from Snyk, dubbed ToxicSkills, found critical security vulnerabilities in 13.4% of AI agent skills scanned from public repositories, with 76 confirmed malicious payloads designed for credential theft and data exfiltration. This highlights a significant supply chain risk for enterprises adopting AI agents. Meanwhile, Operant AI launched Agent Protector, a real-time security solution for autonomous AI agents, aiming to prevent privilege escalation and data exfiltration with a zero-trust framework, suggesting the industry is finally acknowledging these glaring issues. <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</strong>: Microsoft CEO Satya Nadella has appointed Charlie Bell as the new engineering quality czar, reporting directly to him. This move aims to address persistent issues like Azure outages, Windows patch problems, and low Copilot adoption rates, signaling a serious effort to improve product reliability and user experience across the board. Hayete Gallot, with a strong background in Windows and Office, returns to Microsoft as Executive VP of Security, further bolstering their focus on core engineering excellence. <a href="https://app.daily.dev/posts/NGytU6zYE" 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 due to their passwordless authentication, the lack of transparency on the scope and delayed detection raises significant privacy concerns for its millions of users. Threat actors are already claiming to have nearly 700,000 records, underscoring the impact of such prolonged vulnerabilities. <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>Anthropic releases Claude Opus 4.6 with agent teams</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. <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, treating them like employees with onboarding and feedback loops, and integrating with major business applications. <a href="https://app.daily.dev/posts/TGZyay3jK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Laravel releases official AI SDK</strong>: Laravel launched 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>
<li><strong>AWS launches new EC2 instances</strong>: AWS rolled out 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>AI bot traffic surges, impacting 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 patterns. <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 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>ScyllaDB integrates USearch for vector search</strong>: ScyllaDB now includes vector search capabilities using the open-source USearch library, allowing storage of 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>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic’s super bowl ad stunt</strong>: Anthropic is reportedly planning Super Bowl ads to mock OpenAI’s move into advertising, depicting chatbots turning into annoying salespeople. OpenAI CEO Sam Altman, naturally, called it ‘dishonest.’ Because nothing says ‘we’re serious about AI’ like spending millions on a Super Bowl ad to complain about your competitor’s business model, especially when both offer similar free and paid tiers. The real innovation here is the marketing department. <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’</strong>: An article boldly claims Elixir is the ‘best language for AI,’ citing an unverified ‘Tencent study’ that supposedly found Elixir had a 97.5% LLM completion rate. Apparently, its immutability and ‘decade-long stability’ make it perfect for AI, unlike those other messy languages. Because what AI really needs is a language that’s been stable for a decade, not, you know, actual AI libraries and frameworks. <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</strong>: The UK government, in its infinite wisdom, announced plans to develop a framework for evaluating deepfake detection tech, partnering with Microsoft and academics. Cybersecurity expert Dr. Ilia Kolochenko, however, points out that existing open-source tools already do this, and the real problem is enforcement, not detection. So, another committee to study a problem that’s already been solved, while the actual issue of holding bad actors accountable remains untouched. Classic. <a href="https://app.daily.dev/posts/EQCYBDiDB" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
