<h2>Must know</h2>
<ul>
<li><strong>AI agents are changing how we build and secure software</strong>: VS Code is rolling out ‘Agent Skills’ to let AI agents like GitHub Copilot dynamically load specialized instructions for developer tasks, enhancing automation and context-aware workflows. This comes as security experts warn that the Model Context Protocol (MCP), which enables AI agents to access tools and data, significantly increases security risks by widening context windows. They advocate for Zero Trust principles, treating AI agents as first-class identities with unique credentials and fine-grained authorization, as highlighted by sources like The Register and SD Times. The rapid evolution of AI agents demands immediate upgrades and robust security practices to prevent uncontrolled access and maintain system integrity. <a href="https://app.daily.dev/feed-by-ids?id=xn90KJxf1&amp;id=8827svTqM&amp;id=lqqfnNpcw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is eating open-source revenue, but not replacing developers</strong>: Tailwind CSS reported a 75% staff layoff and 80% revenue drop, attributing it to AI models directly answering developer questions and generating code, bypassing their documentation-reliant monetization. This highlights a critical challenge for open-source projects. However, despite AI’s growing role (72% of developers use it daily), a Sonar survey found 96% distrust AI-generated code, and 48% don’t always verify it. Stefan Mischook argues AI lacks true logical reasoning, making human developers essential for problem-solving and architectural decisions, suggesting AI is a productivity tool, not a replacement. <a href="https://app.daily.dev/feed-by-ids?id=RuMhQtj6W&amp;id=f3fYMXuKZ&amp;id=q2sjuwRsf&amp;id=4FklVf4AA" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>North Korean hackers weaponize QR codes for credential theft</strong>: The FBI warns that North Korean state-sponsored hackers (Kimsuky/APT43) are using malicious QR codes (‘quishing’) in spear-phishing campaigns targeting U.S. institutions. These attacks, documented in May-June 2025, embed QR codes in emails that redirect users to credential harvesting sites mimicking Microsoft 365 and Okta. This technique exploits the lack of enterprise security on personal mobile devices, bypassing traditional endpoint detection and email security tools, as reported by multiple security outlets. Organizations are urged to control QR-based links and manage all mobile devices as endpoints. <a href="https://app.daily.dev/posts/XmPhFlPf5" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<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 and AI model behavior, potentially making observability more cost-effective than current solutions like Datadog and Splunk. <a href="https://app.daily.dev/posts/XlAd95QPv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google’s AI-focused Gmail redesign</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>Cursor AI code editor challenges VSCode, Stack Overflow declines</strong>: Cursor, an AI-native code editor, has gained 7 million users by deeply integrating AI, contrasting with VSCode’s extension model, while Stack Overflow’s question volume has plummeted to 2008 levels, partly due to AI tools and its own ‘toxic moderation’ issues. <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>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>Percona donates database platform to CNCF, Amazon expands Lightsail</strong>: Percona is donating its database management platform, now OpenEverest, to the CNCF for vendor-neutral Kubernetes database management, 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>
<li><strong>GPU-accelerated simulations in upcoming games and tech demos</strong>: Independent developer Zolden is creating ‘Simulario,’ a physics sandbox game using a custom GPU-based engine for fluid and particle simulations, while Unity’s ‘Time Ghost’ tech demo showcases new real-time graphics features like advanced hair systems and large-scale environments. <a href="https://app.daily.dev/feed-by-ids?id=Ie1ynbbMv&amp;id=qHlkEVUn8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New open-source React scheduler component released</strong>: CalendarKit, a new open-source React calendar and scheduler component library, offers month, week, day, agenda, and resource views with features like drag-and-drop, dark mode, and TypeScript support, available in both Basic and Pro versions under an MIT license. <a href="https://app.daily.dev/feed-by-ids?id=MBDqbsDnk&amp;id=RcOUV5N3R&amp;id=mf2DGEww0" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI-generated code: the gift that keeps on giving… security vulnerabilities</strong>: SonarSource’s latest leaderboard shows a wide range in LLM Java code quality, with some models generating three times more code and introducing security vulnerabilities like SQL injection. Meanwhile, 96% of developers don’t trust AI-generated code, yet only 48% bother to verify it. So, we’re getting more code, less trust, and a whole lot of new security holes. 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 in software development: faster iteration, slower understanding</strong>: A software consultancy found that while AI tools speed up early ideation and simple code changes, they also create ‘anchoring effects’ that limit innovation and ‘review bottlenecks’ because AI generates changes faster than humans can understand them. This leads to ‘knowledge erosion’ among developers, making software harder to maintain. So, we’re trading deep understanding for superficial speed. Sounds like a solid long-term strategy. <a href="https://app.daily.dev/feed-by-ids?id=DRTczTJyG&amp;id=K8Q8uNU3E&amp;id=Kf8FXuyo" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Game publishers: AI is ‘cancerous,’ unless we’re using it ourselves</strong>: Hooded Horse, an indie publisher, has banned generative AI from its partnered game development, with its CEO calling it ‘cancerous’ due to copyright concerns. Yet, major gaming companies are embracing it, and Larian Studios, after public disapproval, will only use AI trained on proprietary data for internal processes. So, AI is a threat when others use it, but a ‘development process improvement’ when we do. Got it. <a href="https://app.daily.dev/feed-by-ids?id=ufoxpOMSz&amp;id=cBuEU5Zrv&amp;id=cDdqzyRVb&amp;id=o9kOEf5qk" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
