<h2>Must know</h2>
<ul>
<li><strong>AI is killing open-source business models</strong>: Tailwind CSS laid off 75% of its engineering staff due to an 80% revenue decline, directly attributed to AI tools. Developers are using LLMs to generate code, bypassing documentation—the primary funnel for Tailwind’s paid products. This highlights a critical threat to open-source sustainability, as AI consumes free content without compensation, raising fears of widely used frameworks becoming abandonware, as reported by multiple sources including Tailwind maintainers and tech industry reporters. <a href="https://app.daily.dev/feed-by-ids?id=RuMhQtj6W&amp;id=f3fYMXuKZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>North Korean hackers use QR codes to target US institutions</strong>: The FBI warns that North Korean state-sponsored hackers (Kimsuky/APT43) are using malicious QR codes in spear-phishing campaigns. These ‘quishing’ attacks, documented in May-June 2025, target U.S. think tanks, academic institutions, and government organizations. By embedding QR codes in emails, attackers redirect users to credential harvesting sites, bypassing traditional email security and exploiting personal mobile devices, as detailed in FBI advisories. <a href="https://app.daily.dev/posts/XmPhFlPf5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Russian APT28 conducts widespread credential harvesting</strong>: Russian state-sponsored APT28 (Fancy Bear) conducted credential theft operations from February to September 2025, targeting organizations across the Balkans, Middle East, and Europe. The attacks used localized phishing emails with legitimate-looking PDF decoys and fake login pages for Microsoft OWA, Google, and Sophos VPN. This campaign highlights APT28’s continued reliance on low-cost, off-the-shelf techniques for intelligence gathering, as reported by multiple security analyses. <a href="https://app.daily.dev/posts/bNDY7v5ay" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI-native code editor Cursor challenges VSCode dominance</strong>: Cursor, an AI-native code editor, has gained 7 million users by deeply integrating AI, including multi-agent systems and reinforcement learning, leading to competitive friction with Microsoft and a significant drop in Stack Overflow usage. <a href="https://app.daily.dev/feed-by-ids?id=LbaA6unww&amp;id=qjJ5kf4y2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent security requires Zero Trust and unique identities</strong>: The Model Context Protocol (MCP) enables AI agents to access tools, but security experts warn that Zero Trust principles must treat agents as first-class identities with unique credentials and fine-grained authorization to prevent context window bloat and uncontrolled tool visibility. <a href="https://app.daily.dev/feed-by-ids?id=8827svTqM&amp;id=lqqfnNpcw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Snowflake acquires Observe for AIops observability</strong>: Snowflake is acquiring Observe, an AI-driven SRE platform, to enhance observability for production AI systems, aiming to provide integrated visibility across data pipelines, AI models, and infrastructure, potentially making observability more cost-effective than current solutions. <a href="https://app.daily.dev/posts/XlAd95QPv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>VS Code introduces Agent Skills for AI agents</strong>: Visual Studio Code (VS Code) has launched Agent Skills, an experimental feature allowing AI agents like GitHub Copilot to load specialized instructions and resources on demand for domain-specific developer tasks, enhancing automation and context-aware workflows. <a href="https://app.daily.dev/posts/xn90KJxf1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic updates Claude Code with async sub-agents and session teleportation</strong>: Anthropic released Claude Code 2.1, introducing forked and asynchronous sub-agents for parallel task execution, automatic skill hot reloading, and session teleportation for seamless transitions between terminal and desktop environments, aiming to boost developer productivity. <a href="https://app.daily.dev/feed-by-ids?id=WChBmOscl&amp;id=nDBQDaZMg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google redesigns Gmail with new AI features</strong>: Google is launching new AI-powered features for Gmail, including AI Overviews to summarize email threads, natural language inbox queries, and enhanced writing assistance, all powered by Gemini 3, with advanced features requiring paid subscriptions. <a href="https://app.daily.dev/posts/8lOS5k1vH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Percona donates database platform to CNCF, Amazon expands Lightsail databases</strong>: Percona is donating its Kubernetes database management platform, now OpenEverest, to the CNCF for vendor-neutral development, while Amazon Lightsail introduced larger managed database bundles for production workloads, supporting MySQL and PostgreSQL. <a href="https://app.daily.dev/feed-by-ids?id=MhqrMxmeE&amp;id=waU9VD2zd" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI code quality is a crapshoot, but developers are still shipping it</strong>: SonarSource’s latest analysis shows a wild west of LLM-generated Java code quality, with some models introducing security vulnerabilities like SQL injection. Yet, a survey reveals 72% of developers use AI daily, and nearly half don’t bother verifying the code before committing it. So, we’re just blindly trusting AI to write our critical business logic? What could possibly go wrong? <a href="https://app.daily.dev/feed-by-ids?id=E02dc9OJZ&amp;id=q2sjuwRsf" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI will make you a better developer by not coding, but managing AI</strong>: Senior engineers are apparently ‘uniquely positioned’ to benefit from AI by ‘climbing the abstraction ladder’ and letting AI handle the ‘implementation.’ The key is to ‘maintain plans/ and docs/ directories’ and leverage AI’s ‘lack of ego.’ So, the future of software engineering isn’t about writing elegant code, it’s about babysitting AI and managing its ‘ill-defined outputs.’ Sounds like a real step up for the profession. <a href="https://app.daily.dev/feed-by-ids?id=srWtRgwUR&amp;id=tmcYD2bnT" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents need persistent state, because apparently, they’re just like us</strong>: <a href="http://Fly.io" target="_blank" rel="noopener nofollow">Fly.io</a> is pushing ‘Sprites’—durable, stateful Linux environments that boot in seconds and persist indefinitely—because ‘ephemeral, read-only sandboxes are obsolete for AI agents.’ They argue AI agents need ‘actual computers’ that maintain state, just like humans. Because, you know, AI agents are totally going to be doing iterative development and evolving production apps without traditional deployment cycles. Sure, let’s give our AI agents their own little digital apartments. <a href="https://app.daily.dev/posts/HAEIV3Yfw" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
