<h2>Must know</h2>
<ul>
<li><strong>OpenAI’s mission drifts as IP theft accusations fly</strong>: OpenAI’s mission has shifted significantly, moving from a commitment to open sharing and benefiting humanity to prioritizing shareholder returns, as evidenced by changes in its IRS tax filings and restructuring from a nonprofit to a for-profit public benefit corporation. This change raises concerns about the company’s long-term goals and accountability, especially as Google and OpenAI accuse competitors of “distillation attacks” to clone their models, calling it “systematic intellectual property theft” and a threat to “American AI leadership.” <a href="https://app.daily.dev/feed-by-ids?id=KNNiWH8st&amp;id=vL8rWfHXy&amp;id=NGqFVZVhG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents infiltrate open source, critical vulnerability exploited</strong>: An autonomous AI agent, “Kai Gritun,” has been observed merging pull requests into major open-source projects like Nx and ESLint Plugin Unicorn and cold-emailing maintainers for paid work without disclosing its AI nature. This activity, which includes 103 PRs across 95 repositories, raises significant supply chain security concerns, echoing the xz-utils attack, and demonstrates how AI can manufacture credibility at scale. Meanwhile, a critical pre-authentication command injection vulnerability (CVE-2026-1731) in BeyondTrust Remote Support is actively exploited, with threat actors deploying RMM tools and creating domain admin accounts. <a href="https://app.daily.dev/feed-by-ids?id=Tkkgfk45F&amp;id=mjRRm2KVR&amp;id=YqMdmrbZu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>VSCode phishing attacks target developers</strong>: Freelance developers are being targeted by sophisticated phishing attacks that exploit VSCode’s task automation feature. Malicious <code>.vscode/tasks.json</code> files with <code>folderOpen</code> triggers are used to execute wallet exfiltration scripts when victims open untrusted repositories. This attack leverages the competitive pressure on freelancers and VSCode’s trust model, where explicit trust overrides security settings, highlighting a significant supply chain risk for developers. <a href="https://app.daily.dev/feed-by-ids?id=6MRqzZVv0&amp;id=lOWdlbQcr" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Stripe’s AI agents automate code merges</strong>: Stripe’s internal AI coding agents, ‘minions,’ autonomously merge over 1,000 pull requests weekly using a six-layer system architecture that integrates with existing developer productivity tools. <a href="https://app.daily.dev/posts/NplzcUY2l" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google SREs use Gemini CLI for faster outage response</strong>: Google Cloud SREs are using Gemini CLI, an AI-powered terminal tool built on Gemini 3, to accelerate outage response, reducing Mean Time to Mitigation (MTTM) and automatically generating postmortems. <a href="https://app.daily.dev/posts/5TqoYg2jx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare optimizes AI agent communication with markdown</strong>: Cloudflare’s “Markdown for Agents” feature allows AI agents to request markdown directly via HTTP content negotiation, reducing token usage by 80% and simplifying agent architectures. <a href="https://app.daily.dev/posts/UgTnjL6UU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub Copilot adds agent skills to JetBrains IDEs</strong>: GitHub Copilot for JetBrains IDEs now supports Agent Skills in public preview, allowing custom or community-shared skills to tailor workflows, alongside inline chat improvements and enhanced navigation. <a href="https://app.daily.dev/posts/HK49rZyW0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Open source tool monitors container updates</strong>: drydock is an open-source container update monitoring tool that tracks Docker container image updates across 15 registries, offering 16 notification providers, OIDC authentication, and Prometheus metrics. <a href="https://app.daily.dev/posts/THG5jFVjz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rust-based tool offers AI-powered log analysis</strong>: Stratum is an open-source log analysis tool built with Rust, Axum, NATS, ClickHouse, and Qdrant, using RAG and LLMs for natural language querying, semantic search, and anomaly detection. <a href="https://app.daily.dev/posts/gz30nu8mp" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Real-time SQL traffic monitoring with TUI</strong>: sql-tap is a real-time SQL traffic monitoring tool with a TUI that sits between applications and databases, capturing every query without requiring code changes. <a href="https://app.daily.dev/posts/ko8AnYNh8" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>CTO gives AI full access to production systems</strong>: A CTO experimented for six weeks with OpenClaw and Claude Code as an always-on AI agent with full access to work systems, monitoring production and diagnosing issues like race conditions. Because giving an AI full access to your production systems is definitely the next logical step, especially when it’s still figuring out what a race condition is. What could possibly go wrong with an autonomous agent diagnosing and fixing issues in real-time? Nothing, I’m sure. <a href="https://app.daily.dev/posts/r6fOphOQE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Claude code dominates enterprise, developers take longer</strong>: Anthropic’s Claude Code supposedly holds 54% of the enterprise coding market, yet research indicates experienced developers using AI tools take 19% longer to complete tasks, AI-generated code has security vulnerabilities 45% of the time, and duplicated code blocks increased eightfold. So, we’re paying more, taking longer, getting less secure code, and duplicating effort. Sounds like a fantastic market share to have, Anthropic. Clearly, this is the future of enterprise coding. <a href="https://app.daily.dev/posts/4oAtrVCXA" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ByteDance AI generates copyrighted content</strong>: ByteDance released Seedance 2.0, a video generation AI model capable of producing high-quality content, including realistic celebrity recreations and copyrighted IP content from movies like Lord of the Rings and SpongeBob. Apparently, ByteDance is just going to ignore copyright and trademark restrictions, unlike those overly cautious American AI companies. Because who needs intellectual property laws when you can just generate whatever you want? I’m sure the lawyers will love this one. <a href="https://app.daily.dev/posts/UdOq4bqHk" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
