<h2>Must know</h2>
<ul>
<li><strong>AI coding tools are changing the game, for better or worse</strong>: AI coding assistants are rapidly evolving, with tools like Claude Code enabling a developer to port 100,000 lines of JavaScript to Rust in a month, as reported by the project author. However, this speed comes with caveats: AI-generated code often lacks architectural coherence and requires significant human oversight for quality and maintainability, according to a developer’s two-year experience. The industry is seeing a proliferation of AI code review tools, with some, like Greptile, advocating for independent validation agents to ensure unbiased approvals. Meanwhile, Cursor shipped its Composer coding agent to production, using a mixture-of-experts model with sandboxed execution and iterative loops for precise code edits. This shift means developers are becoming orchestrators of AI agents, focusing more on ‘context engineering’ than raw coding, as highlighted by a virtual developer panel. <a href="https://app.daily.dev/feed-by-ids?id=a8LWkElwv&amp;id=FG4y5lmMt&amp;id=MVxk3C8r1&amp;id=NAHzSpYqv&amp;id=W6qwFXBzp" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft and Google push new AI chips and platforms</strong>: Microsoft launched the Maia 200 chip for AI inference, claiming it outperforms Amazon’s Trainium3 and Google’s seventh-generation TPU in FP4 and FP8 performance, respectively. This move aligns with the trend of tech giants developing custom silicon to reduce reliance on NVIDIA, as reported by Microsoft. Concurrently, Google released FunctionGemma, a 270M parameter AI model optimized for on-device function calling on mobile and edge devices, aiming to translate natural language into structured API calls. Google also launched Gemini for Government, an AI platform authorized at FedRAMP High, already adopted by the U.S. Department of Defense and FDA, according to Gartner. <a href="https://app.daily.dev/feed-by-ids?id=fDSp3OUaH&amp;id=DFyxBQQFb&amp;id=Fm4JFG1tC" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI and competitors want a piece of the drug discovery pie</strong>: OpenAI is looking to get a share of drug discoveries made using its AI models through licensing and outcome-based pricing, as stated by its CFO. This marks a shift towards claiming intellectual property from AI-assisted biological inventions. Competitors like Anthropic, Google DeepMind, and Isomorphic Labs are pursuing similar strategies, seeking data licensing and partnerships with biotech firms. This new model could raise legal and ethical questions about ownership and value distribution for AI-driven discoveries. <a href="https://app.daily.dev/posts/FoOQYVciF" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Docker sandboxes for AI coding agents now on macOS</strong>: Docker Sandboxes now offers microVM-based isolation for autonomous AI coding agents like Claude Code and Gemini CLI, currently available on macOS with Windows and Linux support planned. <a href="https://app.daily.dev/posts/6YdOupFK7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ChatGPT code execution environment gets a major upgrade</strong>: OpenAI’s ChatGPT code execution environment now supports Bash commands, multiple programming languages (JavaScript, Ruby, Perl, PHP, Go, Java, Swift, Kotlin, C, C++), and package installations via pip and npm, all while maintaining network isolation through internal proxies. <a href="https://app.daily.dev/posts/7V1v7MpYn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Figma and Claude AI team up for diagram generation</strong>: Figma has integrated Claude AI into FigJam, allowing users to automatically generate editable diagrams like user journeys and system architectures from text prompts, PDFs, or images, accessible via the Figma MCP app. <a href="https://app.daily.dev/posts/Yk7n7ifdw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mozilla launches paid program for AI and web innovation</strong>: Mozilla has launched ‘Mozilla Pioneers,’ a paid program for experienced technologists and designers to develop foundational ideas for AI and web technologies, offering flexible, part-time engagement and potential full-time opportunities. <a href="https://app.daily.dev/posts/57mQTlLV0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI infrastructure relies on standard tech, not exotic solutions</strong>: Most AI infrastructure uses traditional, mature tech stacks like API gateways, microservices, Kubernetes, and observability tools, with service meshes and exotic solutions being uncommon, according to unnamed data sources. <a href="https://app.daily.dev/feed-by-ids?id=RHjYjNoa6&amp;id=nBkzOJhUQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Node.js 25.5.0 enhances security and single executable apps</strong>: Node.js 25.5.0 introduces updated root certificates, support for Single Executable Application (SEA) generation with a new <code>--build-sea</code> flag, an <code>ignore</code> option for <code>fs.watch</code>, and activates SQLite defensive mode by default. <a href="https://app.daily.dev/posts/BrQRGrBKE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>KDE Plasma desktop now compatible with Apple M3 via Fedora Asahi Remix</strong>: The KDE Plasma desktop now runs on Apple M3 chips, a significant milestone for Linux on Apple Silicon, achieved through the Fedora Asahi Remix project. <a href="https://app.daily.dev/posts/cxEiHEruW" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI will take your job, eventually, maybe, probably not today</strong>: Tech executives, including those from Anthropic and NVIDIA, keep predicting the imminent obsolescence of software engineers due to AI, often repeating the same 6-12 month timelines year after year at events like Davos. Yet, despite all the hype, there’s still no concrete evidence of widespread layoffs or role eliminations directly attributable to AI automation. It seems the ‘AI will replace developers’ narrative is just another installment in a long-running series of overhyped tech predictions, designed more for clickbait and investor excitement than actual reality, as noted by Import AI and other analysts. <a href="https://app.daily.dev/feed-by-ids?id=QxbydGQJy&amp;id=1iSdibthb&amp;id=7WOFbf92G&amp;id=9yXNAQlaN&amp;id=YDOofaap7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated code: the fast fashion of software</strong>: Apparently, AI is now mass-producing ‘mediocre, narrow, and rote’ software, turning development into ‘fast fashion’ where speed trumps quality. This ‘vibe coding’ leads to inconsistent, inefficient, and ultimately disposable code, undermining software as a durable asset, according to various industry commentators. So, while AI might churn out lines of code faster than ever, don’t expect it to be a masterpiece; it’s more like a cheap, ill-fitting garment destined for the landfill of technical debt. <a href="https://app.daily.dev/feed-by-ids?id=dABjwqV7N&amp;id=vz4g5mSDD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google’s AI Overviews prefer YouTube over actual medical experts</strong>: A German study found that Google’s AI Overviews feature cites YouTube more often than established medical websites for health-related queries, raising concerns that Google prioritizes content visibility over medical authority. So, next time you ask Google’s AI about your symptoms, you might get advice from a cat video influencer instead of a doctor. Google, of course, insists its AI prioritizes ‘high-quality content’—even if that content happens to be on a platform where anyone can upload anything. <a href="https://app.daily.dev/posts/ZEb80nG1f" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
