<h2>Must know</h2>
<ul>
<li><strong>AI is eating open source revenue models</strong>: Tailwind CSS, a widely used framework, laid off 75% of its engineering staff due to an 80% revenue decline. The culprit? AI tools generating code and answering questions, drastically reducing traffic to their documentation—which was key for monetizing paid products. This highlights a growing problem: AI leverages free community content without compensation, threatening the sustainability of critical open-source projects, as reported by multiple tech outlets and Tailwind’s founder. Google AI Studio has stepped in with sponsorship, but the long-term solution for open-source funding in an AI-driven world remains unclear. <a href="https://app.daily.dev/feed-by-ids?id=3KrG8AfuZ&amp;id=d57ILPjBN&amp;id=hsznY6oi4&amp;id=mpCeVdjI4&amp;id=re8TItYuD&amp;id=u4S1Vnazf&amp;id=yhQvgBQbs&amp;id=WwALdiGFY&amp;id=tFhv6QoOp" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New AI models push multimodal and edge capabilities</strong>: NVIDIA launched TensorRT Edge-LLM, an open-source C++ framework for high-speed LLM and VLM inference on embedded automotive and robotics systems, optimizing for low-latency edge use cases. Meanwhile, Google introduced Gemini 3, boasting massive multimodal input handling (over 1 million tokens) and advanced reasoning with its Deep Think mode. This aims to unify fragmented AI systems into a single API for enterprise, as highlighted by Google’s promotional materials. Both releases, occurring this week, signify major advancements in deploying AI across diverse environments, from autonomous vehicles to complex enterprise workflows. <a href="https://app.daily.dev/feed-by-ids?id=1Lrxf6i15&amp;id=IYZKv2e0G&amp;id=NYhUyiiE6&amp;id=Ss6K85gTn&amp;id=vZbhC083l" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical vulnerabilities in HPE OneView and Coolify demand immediate attention</strong>: CISA has flagged a critical remote code execution vulnerability (CVE-2025-37164) in HPE OneView, rated CVSS 10.0, which is actively being exploited. This flaw grants attackers full system control, emphasizing the urgency for immediate patching. Separately, Coolify, an open-source self-hosting platform, disclosed 11 critical vulnerabilities (also CVSS 10.0) allowing root command execution and full server compromise on its 52,000+ exposed instances. While no active exploitation is reported for Coolify, immediate upgrades are advised, as detailed by The Hacker News. <a href="https://app.daily.dev/feed-by-ids?id=VmTOY3pRn&amp;id=RXFyRlNEd&amp;id=ZRs47yc16" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI coding agents are coming, bringing new security risks</strong>: Simon Willison predicts LLMs producing high-quality code will be mainstream by 2026, but also foresees a major security incident involving coding agents, highlighting the need for secure sandboxing solutions. <a href="https://app.daily.dev/feed-by-ids?id=QaeYUARam&amp;id=cLDAM9Yxb&amp;id=tmcYDkbnT" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux kernel rejects AI-specific policies</strong>: Linus Torvalds has dismissed proposals to add AI-specific disclosure requirements to the Linux kernel documentation, arguing that existing code review processes are sufficient to handle low-quality submissions, regardless of whether they’re AI-generated. <a href="https://app.daily.dev/posts/4cToJt8jS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI models leak copyrighted works and private data</strong>: Researchers found that major AI models like Anthropic’s Claude 3.7 Sonnet and Google’s Gemini 2.5 Pro can reproduce large verbatim sections of copyrighted books, raising significant copyright and data privacy concerns. <a href="https://app.daily.dev/feed-by-ids?id=RZ3NYDnWq&amp;id=pnbMxpKrW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>CrowdStrike acquires SGNL for identity security</strong>: CrowdStrike acquired SGNL for $740 million to bolster its Falcon platform’s identity security, focusing on continuous, context-aware authorization for both human and AI agents amid a 32% rise in identity-based cyberattacks. <a href="https://app.daily.dev/feed-by-ids?id=kHk9YWnOk&amp;id=kl0LpaKDU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks introduces Instructed Retriever for enterprise AI accuracy</strong>: Databricks unveiled Instructed Retriever, an AI architecture that integrates database queries with RAG similarity search, parsing user instructions <em>before</em> retrieval to deliver more accurate AI-generated answers in enterprise settings. <a href="https://app.daily.dev/posts/YccZnqmwj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ruby 4.0 released with major compiler and library updates</strong>: Ruby 4.0 is out, featuring ZJIT as a new optional JIT compiler, Prism as the default parser, and updated ecosystem tools like RBS and Pry, all supporting the new version. <a href="https://app.daily.dev/posts/LoVcIyHKl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google expands AI features in Workspace products</strong>: Google Meet now offers an automatic AI note-taking toggle for hosts, and Gemini AI features like AI Overviews and Help Me Write are rolling out to all US English Gmail users, with privacy assurances that content won’t train models. <a href="https://app.daily.dev/feed-by-ids?id=13zTeB5TT&amp;id=G027yRnAU" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI-driven ‘vibe coding’ is the future, apparently</strong>: Norwegian tech outlets are hyping ‘vibe coding’ where AI generates entire programs, promising massive productivity gains. Never mind the security vulnerabilities, like missing authentication and prompt injection attacks, that security experts keep finding in this AI-generated code. Management is pushing for this ‘efficiency,’ but it seems they’re conveniently overlooking the ‘minor’ detail of actually delivering secure, functional software, as noted by various industry commentators. <a href="https://app.daily.dev/feed-by-ids?id=CSSbkecce&amp;id=mpCeVdjI4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents will automate everything, eventually</strong>: Dynatrace is touting ‘agentic workflows’ with Davis CoPilot to automate monitoring and response, claiming a shift from deterministic automation to human-like judgment. While they’re great for simple tasks like cloud cost management, experts are quick to point out that full autonomy for mission-critical infrastructure is still 5-10 years away. So, for now, your job is safe, unless you’re just managing simple, repetitive tasks, in which case, good luck. <a href="https://app.daily.dev/feed-by-ids?id=LH3z3eMJi&amp;id=gQn7Cgin4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Stack Overflow is dead, long live AI</strong>: Stack Overflow’s monthly questions have plummeted, with AI tools like ChatGPT and GitHub Copilot taking the blame. Apparently, developers prefer instant, judgment-free answers from AI over the ‘hostile moderation and gatekeeping culture’ of human communities. The irony, of course, is that these AI models were trained on the very human knowledge they’re now replacing, creating a delightful feedback loop where less human interaction means less new data for the AI. What could possibly go wrong? <a href="https://app.daily.dev/posts/XI68jmY7m" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
