<h2>Must know</h2>
<ul>
<li><strong>VS Code introduces agent skills for AI agents</strong>: Visual Studio Code is rolling out ‘Agent Skills,’ an experimental feature that lets AI agents, like GitHub Copilot, dynamically load specialized instructions and resources for specific developer tasks. This open standard allows developers to ‘teach’ Copilot new abilities by adding skills in designated directories, enhancing automation and context-aware workflows, as reported by the VS Code team. This move deepens AI integration directly into the coding environment, aligning with the trend of intelligent automation in developer tools. <a href="https://app.daily.dev/feed-by-ids?id=xn90KJxf1&amp;id=7Q1krnrII" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-native code editor Cursor challenges VS Code dominance</strong>: Cursor, an AI-native code editor developed by MIT students, has rapidly gained 7 million users by deeply integrating AI into its core, unlike VS Code’s extension-based approach. Its ‘Composer’ feature can autonomously refactor entire code modules, pushing ‘agent-first development.’ However, Cursor has faced infrastructure instability and Microsoft restricting its access to the VS Code extension marketplace, as detailed by HuXn WebDev. This highlights the intense competition and rapid evolution in AI-powered coding 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 has warned that North Korean state-sponsored hackers (Kimsuky/APT43) are using malicious QR codes in spear-phishing campaigns to target U.S. think tanks, academic, and government organizations. These ‘quishing’ attacks embed QR codes in emails that redirect users to credential harvesting sites, bypassing traditional email security and exploiting personal mobile devices, according to FBI advisories. This sophisticated social engineering tactic underscores the evolving threat landscape in cybercrime and data theft. <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>Anthropic updates Claude Code with async agents and workflow upgrades</strong>: Anthropic has released Claude Code 2.1, introducing forked and asynchronous sub-agents for parallel task execution, automatic skill hot reloading, and ‘session teleportation’ 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 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>Snowflake acquires Observe for AIops observability</strong>: Snowflake is acquiring Observe, an AI-driven SRE platform, to enhance observability for production AI systems by providing integrated visibility across data pipelines, AI model behavior, and infrastructure health. <a href="https://app.daily.dev/posts/XlAd95QPv" 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 for Microsoft OWA and Google. <a href="https://app.daily.dev/posts/bNDY7v5ay" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Zero Trust security for AI agents becomes critical</strong>: Security experts emphasize that Zero Trust principles must treat AI agents as first-class identities with unique credentials and granular authorization policies, especially with new protocols like MCP widening context windows and increasing security risks. <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>New developments in React and Python ecosystems</strong>: The React core team reports on React 19.2 and the React Compiler’s progress, while the Python ecosystem sees 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>
<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>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI code quality: still a dumpster fire, but now with more SQL injection</strong>: SonarSource, in its infinite wisdom, has analyzed LLM-generated Java code and found that while Opus 4.5 Thinking is ‘best’ (83.62% pass rate), GPT-5.2 High generates three times more code and introduces lovely security vulnerabilities like SQL injection. Apparently, 96% of developers distrust AI-generated code, yet 48% don’t bother verifying it before committing. So, we’re all just blindly pushing buggy, insecure code to production, because AI is ‘fast.’ 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 game development: ‘cancerous’ or just misunderstood?</strong>: Hooded Horse, the publisher behind ‘Manor Lords,’ has declared generative AI ‘cancerous’ and contractually banned its use by partnered studios, citing copyright concerns and the risk of AI-created assets ‘slipping’ into final games. Meanwhile, Larian Studios’ CEO, Swen Vincke, clarified they won’t use generative AI for Divinity concept art to avoid community distrust, despite experimenting internally. It seems the industry is still figuring out if AI is a creative partner or a legal liability, with strong opinions on both sides. <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>
<li><strong>AI-assisted development: faster to failure, less knowledge for all</strong>: A software consultancy, after integrating LLMs, found that while AI speeds up ideation and simple code changes, it also creates ‘anchoring effects’ that limit innovation and ‘review bottlenecks’ because humans can’t keep up. Worse, they warn of ‘knowledge erosion’ among developers, meaning we’re losing the ability to understand and maintain the very software AI helps us create. So, we’re trading deep understanding for rapid, potentially flawed, iteration. 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>
</ul>
