<h2>Must know</h2>
<ul>
<li><strong>AI agents and zero trust security: a new frontier</strong>: The Model Context Protocol (MCP) is gaining traction, enabling AI agents like Claude Code to access tools and data more efficiently. However, this also widens context windows and increases security risks. Experts are pushing for Zero Trust principles, advocating for unique credentials and fine-grained authorization for each AI agent, rather than shared API keys, to prevent uncontrolled access. This is critical as AI agents, lacking human judgment, might attempt all accessible actions, as highlighted by security experts. <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>North Korean hackers weaponize QR codes for credential theft</strong>: The FBI is warning about North Korean state-sponsored hackers (Kimsuky/APT43) using malicious QR codes in spear-phishing campaigns. These ‘quishing’ attacks, observed in May-June 2025, target U.S. institutions by embedding QR codes in emails that redirect users to credential harvesting sites. This technique bypasses traditional email security and endpoint detection, especially when users scan codes on unmanaged mobile devices, as reported by the FBI. <a href="https://app.daily.dev/posts/XmPhFlPf5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Russian APT28 continues credential harvesting against strategic targets</strong>: Russian state-sponsored APT28 (Fancy Bear/BlueDelta) conducted credential theft operations from February to September 2025, targeting organizations in the Balkans, Middle East, and Europe. They used localized phishing emails with legitimate-looking PDF decoys and fake login pages for services like Microsoft OWA and Google. These campaigns, which relied on commercial VPNs and free hosting, served as initial access points for intelligence gathering, demonstrating a focus on operational efficiency over technical sophistication, according to multiple reports. <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>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 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 has launched ‘Agent Skills,’ an experimental feature allowing AI agents like GitHub Copilot to dynamically load specialized instructions and resources 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>Cursor AI code editor gains traction, Stack Overflow declines</strong>: Cursor, an AI-native code editor with integrated AI architecture, has reached 7 million users, leading to competitive friction with Microsoft and a significant drop in Stack Overflow’s question volume to 2008 levels. <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. <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>New developments in Kubernetes database management and cloud database bundles</strong>: Percona is donating its database management platform, now OpenEverest, to the CNCF for Kubernetes, 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>Google announces 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>React and Python ecosystems see major early 2026 updates</strong>: The React core team reported on React 19.2 and the React Compiler’s progress, while the Python ecosystem saw the public beta of Astral’s ty type checker, Django 6’s release, and a fix for Python 3.14’s tail-calling interpreter optimization. <a href="https://app.daily.dev/feed-by-ids?id=D2DOQ44q3&amp;id=Golang+Weekly+Issue+584%3A+January+9%2C+2026&amp;id=caNcWqICw" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI-generated code quality: still a dumpster fire, apparently</strong>: SonarSource, in its infinite wisdom, has analyzed LLM-generated Java code and found ‘major differences’ in quality. Apparently, Opus 4.5 Thinking is the least terrible, while GPT-5.2 High generates three times more code and introduces security vulnerabilities like SQL injection. And get this: 96% of developers ‘distrust’ AI-generated code, yet 48% don’t even bother to verify it before committing. 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, but dumber?</strong>: A software consultancy, after integrating LLMs, found that AI speeds up ideation but creates ‘anchoring effects’ and ‘review bottlenecks.’ They’re worried about ‘knowledge erosion’ among developers, meaning we’ll all forget how to code because AI does it for us. So, we’re trading speed for a potential brain drain and hidden system failures. Sounds like a solid plan. <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>Hooded Horse bans generative AI: because it’s ‘cancerous’</strong>: Indie publisher Hooded Horse, known for Manor Lords, has instituted a ‘strict contractual ban’ on generative AI, with CEO Tim Bender calling it ‘cancerous.’ They’re worried about AI-created assets ‘slipping’ into final games and copyright concerns. Meanwhile, major gaming companies are embracing it. It’s almost like there’s a nuanced discussion to be had, but why bother when you can just call it ‘cancerous’? <a href="https://app.daily.dev/posts/ufoxpOMSz" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
