<h2>Must know</h2>
<ul>
<li><strong>AI agents get a brain boost in VS Code and Claude</strong>: VS Code is rolling out ‘Agent Skills,’ an experimental feature allowing AI agents like GitHub Copilot to dynamically load specialized instructions and resources for specific tasks, enhancing automation and context-aware workflows. This open standard aims for portability across compatible AI agents. Meanwhile, Anthropic’s Claude Code 2.1 introduces asynchronous sub-agents for parallel task execution, session teleportation, and persistent memory, significantly upgrading developer productivity, as reported by Visual Studio Code and other tech outlets. <a href="https://app.daily.dev/feed-by-ids?id=xn90KJxf1&amp;id=WChBmOscl&amp;id=nDBQDaZMg&amp;id=7Q1krnrII" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Zero Trust security is a must for AI agents and the Model Context Protocol</strong>: The new Model Context Protocol (MCP), now supported by Laravel Boost v1.8.9, enables AI agents to access tools and data more efficiently. However, security experts warn that this widens context windows and increases risk. They emphasize that Zero Trust principles must treat AI agents as first-class identities with unique credentials and fine-grained authorization, rather than relying on shared API keys, to prevent legacy security gaps from being exploited by agents’ broad operating scope, according to security analyses. <a href="https://app.daily.dev/feed-by-ids?id=8827svTqM&amp;id=lqqfnNpcw&amp;id=K8nRVnuYq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated code quality varies wildly, and developers aren’t always checking it</strong>: SonarSource’s analysis reveals a huge disparity in Java code quality from different LLMs, with Opus 4.5 Thinking achieving an 83.62% pass rate while GPT-5.2 High generates three times more code with significant security vulnerabilities like SQL injection risks. Despite 96% of developers distrusting AI-generated code, only 48% consistently verify it before committing. This highlights a critical gap between AI’s rapid code generation and human review capacity, as detailed by <a href="http://Foojay.io" target="_blank" rel="noopener nofollow">Foojay.io</a>. <a href="https://app.daily.dev/feed-by-ids?id=E02dc9OJZ&amp;id=q2sjuwRsf" 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 by integrating visibility across data pipelines, AI models, and infrastructure health, aiming to make observability more cost-effective than traditional tools. <a href="https://app.daily.dev/feed-by-ids?id=XlAd95QPv&amp;id=dkp1P3j5W" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cursor AI editor challenges VS Code, Stack Overflow usage plummets</strong>: Cursor, an AI-native code editor, has gained 7 million users by deeply integrating AI, leading to competitive friction with Microsoft and a dramatic drop in Stack Overflow’s question volume to 2008 levels, partly due to AI tools. <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>North Korean hackers use QR codes to target US institutions</strong>: The FBI warns that North Korean state-sponsored hackers are using malicious QR codes in spear-phishing campaigns to target U.S. think tanks, academic institutions, and government organizations, exploiting personal mobile devices to bypass traditional security. <a href="https://app.daily.dev/posts/XmPhFlPf5" 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>Russian APT28 conducts credential harvesting campaigns</strong>: Russian state-sponsored APT28 conducted credential theft operations targeting organizations in the Balkans, Middle East, and Europe using localized phishing emails and fake login pages, relying on off-the-shelf infrastructure to reduce detection. <a href="https://app.daily.dev/posts/bNDY7v5ay" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI adoption risks: knowledge erosion and hidden system failures</strong>: Integrating AI tools into software development can speed up ideation but risks ‘knowledge erosion’ among developers due to overreliance on AI output, while operational systems face ‘partial failures’ that go undetected by current monitoring, leading to prolonged harm. <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>OpenEverest and Amazon Lightsail expand database offerings</strong>: Percona is donating its database management platform to CNCF, rebranding it as OpenEverest for Kubernetes, while Amazon Lightsail introduces larger managed database bundles for production workloads, both supporting PostgreSQL and MySQL. <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>Hooded Horse bans generative AI, calls it ‘cancerous’</strong>: Indie publisher Hooded Horse, known for Manor Lords, has taken a bold stance, contractually banning generative AI from all its partnered game development studios. CEO Tim Bender dramatically labels generative AI ‘cancerous,’ citing copyright concerns and the risk of AI-created assets ‘slipping’ into final releases. Apparently, the industry is so divided that some are still fighting the last war, while others are busy figuring out how to actually use the tech, as reported by various gaming news outlets. <a href="https://app.daily.dev/feed-by-ids?id=ufoxpOMSz&amp;id=cBuEU5Zrv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents need ‘durable, stateful Linux environments’ now?</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 apparently, AI agents can’t function in ephemeral sandboxes. They claim agents need ‘actual computers’ to maintain state, eliminating the need to rebuild environments. Because, you know, AI agents are just like us, they hate losing their work when the container dies. It’s almost like they’re trying to sell us on a new kind of persistent infrastructure for something that’s supposed to be stateless and scalable. <a href="https://app.daily.dev/posts/HAEIV3Yfw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Senior engineers: the only ones smart enough to use AI?</strong>: Matthew Rocklin, a Hacker News contributor, suggests that only senior engineers can truly benefit from AI-assisted development because they’re supposedly adept at avoiding low-quality output. The rest of us plebs, I guess, are doomed to drown in AI-generated garbage. He frames it as ‘climbing the abstraction ladder,’ which sounds suspiciously like ‘let AI do the grunt work while I take all the credit.’ It’s a productivity gain, sure, but let’s not pretend it’s exclusive to the enlightened few. <a href="https://app.daily.dev/posts/srWtRgwUR" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
