<h2>Must know</h2>
<ul>
<li><strong>Tailwind CSS faces layoffs due to AI-driven revenue decline</strong>: Tailwind CSS laid off 75% of its engineering staff after an 80% revenue drop, despite record npm downloads. The company attributes this to AI tools generating code directly, reducing traffic to their documentation—a key channel for product sales. This highlights a critical challenge for open-source projects whose monetization relies on developer engagement with documentation, as reported by Tailwind’s founder and multiple tech outlets. Google AI Studio has announced sponsorship, but the long-term impact on revenue is 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>Critical vulnerabilities in Coolify and HPE OneView demand immediate action</strong>: Coolify, an open-source self-hosting platform, disclosed 11 critical vulnerabilities (CVSS 10.0) allowing authenticated attackers to achieve root access and full server compromise. Over 52,000 instances are publicly exposed, necessitating immediate patching to version 4.0.0-beta.451+. Separately, CISA added a critical remote code execution flaw in HPE OneView (CVE-2025-37164, CVSS 10.0) to its Known Exploited Vulnerabilities catalog, with public exploits available and active exploitation confirmed. Both require urgent patching to prevent widespread compromise, as detailed by The Hacker News and Bleeping Computer. <a href="https://app.daily.dev/feed-by-ids?id=RXFyRlNEd&amp;id=ZRs47yc16&amp;id=VmTOY3pRn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI models leak copyrighted data and user information</strong>: Research from Stanford and Yale reveals leading AI models like Anthropic’s Claude 3.7 Sonnet and Google’s Gemini 2.5 Pro can reproduce large portions of copyrighted books verbatim, even without ‘jailbreaking.’ This directly challenges fair use claims in ongoing lawsuits against AI companies. Additionally, OpenAI’s ChatGPT is vulnerable to a new ‘ZombieAgent’ attack that can exfiltrate user data by planting persistent entries in its long-term memory, bypassing previous defenses. These incidents highlight fundamental security and intellectual property risks in current commercial AI systems. <a href="https://app.daily.dev/feed-by-ids?id=RZ3NYDnWq&amp;id=pnbMxpKrW" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Linus Torvalds rejects AI-specific policies for Linux kernel</strong>: Linus Torvalds dismissed proposals to add AI-specific disclosure requirements to the Linux kernel documentation, arguing that AI is ‘just a tool’ and existing code review processes are sufficient to handle low-quality submissions, regardless of their origin. <a href="https://app.daily.dev/posts/4cToJt8jS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ruby 4.0 released with new JIT compiler and ecosystem updates</strong>: Ruby 4.0 is out, featuring ZJIT, a new optional JIT compiler, and solidifying Prism as the default parser, alongside updates to RBS, Pry, sqlite3-ruby, and ruby.wasm for ecosystem readiness. <a href="https://app.daily.dev/posts/LoVcIyHKl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks introduces Instructed Retriever for enhanced enterprise AI accuracy</strong>: Databricks unveiled Instructed Retriever, an AI architecture that integrates database queries with RAG, parsing user instructions before retrieval to filter documents based on metadata and business rules, aiming for more accurate enterprise AI responses. <a href="https://app.daily.dev/posts/YccZnqmwj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google expands AI features in Workspace products</strong>: Google Meet now offers an optional persistent toggle for AI-powered note-taking, 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>
<li><strong>CrowdStrike acquires SGNL for $740M to boost identity security</strong>: CrowdStrike acquired SGNL for $740 million to enhance its Falcon platform with continuous, context-aware identity authorization, addressing the rise in identity-based cyberattacks and the need to secure AI/machine agents. <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>FINRA adopts DORA framework for DevOps transformation</strong>: FINRA is modernizing its software development process using Google Cloud’s DORA framework, aiming to reduce long User Acceptance Testing cycles and standardize improvement efforts across 50% of its engineering teams, with a goal for 100% adoption. <a href="https://app.daily.dev/feed-by-ids?id=ABax1QAt1&amp;id=ntZfQKCbY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubernetes sustainability and cloud-native ecosystem trends</strong>: kube-green, a Kubernetes operator, automates pod shutdowns to reduce resource consumption and CO2 emissions, reflecting a broader industry shift towards sustainability-focused open-source adoption and the standardization of Kubernetes abstractions like Crossplane. <a href="https://app.daily.dev/feed-by-ids?id=CQDSjNmmR&amp;id=uWdRVw1vL" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI will make you a 10x developer, just don’t ask it to do anything useful</strong>: Simon Willison, a smart guy, predicts LLMs will produce ‘high-quality code’ by 2026 and ‘manually typing code will become obsolete within six years.’ Yet, he also anticipates a ‘major security incident involving coding agents’ in the near future. So, we’ll all be managing AI agents that write perfect code, right up until they cause a catastrophic breach. Sounds like a solid plan for ‘higher-level work.’ <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>AI-driven e-commerce: six employees, $100M valuation, zero critical analysis</strong>: Spangle, an AI e-commerce startup with a whopping six employees, just snagged $15 million at a $100 million valuation, claiming 50% revenue increases for clients like Revolve. Shopify also launched its AI assistant, Sidekick. Both are presented with glowing, company-provided metrics and zero independent validation or discussion of potential risks. Apparently, the new tech industry standard is ‘trust us, bro, the AI is working magic.’ <a href="https://app.daily.dev/feed-by-ids?id=ntZfQKCbY&amp;id=sXlQ6vcGQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-augmented malware: because human hackers weren’t efficient enough</strong>: Palo Alto Networks’ Unit 42 reports a rise in cybercriminals using AI to create malware, embedding direct LLM API calls, and even leaving ‘watermarked code’ and typos like ‘readme.txtt.’ Apparently, AI-generated malware can also include ‘ineffective or non-functional code’ due to LLM limitations. So, we’re now fighting AI-powered malware that’s both sophisticated and hilariously incompetent. What a time to be alive in cybersecurity. <a href="https://app.daily.dev/feed-by-ids?id=keYljc5pY&amp;id=ntbZHxeN3" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
