<h2>Must know</h2>
<ul>
<li><strong>AI coding tools: productivity gains vs. quality decline</strong>: While AI coding tools like Cursor and Claude are showing increased pull request throughput and adoption, with GitHub Copilot reaching 9.76% daily adoption, a Carnegie Mellon University study reveals a darker side. After an initial 30-40% productivity boost, code quality declines, with static analysis warnings increasing by 30% and complexity rising over 40%. This suggests that while AI can accelerate coding, human oversight remains critical to prevent a long-term degradation of codebase health, as highlighted by CMU and GitClear. <a href="https://app.daily.dev/feed-by-ids?id=EZUXhYwRZ&amp;id=XkFn8TQqF&amp;id=7KqSlBkW9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI datacenter expansion faces local pushback</strong>: OpenAI’s ambitious $500 billion ‘Stargate’ datacenter project is facing significant community resistance, with 20 US datacenter projects blocked or delayed in Q2 2025. OpenAI is pledging self-funded energy infrastructure and closed-loop cooling to address concerns over electricity and water consumption. However, local communities remain skeptical, especially given the limited job creation (around 100 full-time operational jobs per campus), raising questions about resource prioritization and the true environmental impact of these massive AI infrastructure investments. <a href="https://app.daily.dev/posts/O0VoWxWDb" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical vulnerabilities in AI chatbot framework Chainlit</strong>: Two critical vulnerabilities (CVE-2026-22218, CVE-2026-22219) have been discovered in Chainlit, a popular Python-based AI chatbot development framework with over 5 million downloads. These flaws allow unauthenticated attackers to read any file on the server and perform CSRF/SSRF attacks, potentially leading to the exfiltration of API keys, sensitive credentials, and database contents. Users are urged to upgrade to Chainlit version 2.9.4 immediately to patch these severe security risks. <a href="https://app.daily.dev/posts/I9vckVFsM" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Amazon launches Bedrock AgentCore for AI agent deployment</strong>: Amazon has released Bedrock AgentCore, a serverless platform for deploying and managing AI agents at scale, offering modular services for runtime, memory, gateway, code interpretation, and observability, along with a CLI toolkit for rapid development and integration with frameworks like LangGraph and CrewAI. <a href="https://app.daily.dev/posts/PbCI8QH1q" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LLVM bans unchecked AI-assisted code contributions</strong>: The LLVM compiler project has implemented a ‘human in the loop’ policy, banning AI-generated code contributions without thorough human review, citing a surge in low-quality submissions and aiming to preserve code quality and beginner learning opportunities. <a href="https://app.daily.dev/posts/DXhDzQ15v" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>VS Code exploited by ‘Contagious Interview’ campaign</strong>: A new social engineering campaign, ‘Contagious Interview,’ is exploiting VS Code by embedding malicious code in <code>tasks.json</code> files within fake Git repositories, leading to persistent backdoors on developers’ machines once the repository is marked as trusted. <a href="https://app.daily.dev/posts/PT6kAI0i3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Netflix pivots to cloud gaming and app redesign</strong>: Netflix is making cloud-based gaming a top strategic priority for 2026, focusing on casual, party-oriented games, while also redesigning its mobile app to include vertical video feeds and video podcasts to boost engagement against competitors like TikTok. <a href="https://app.daily.dev/feed-by-ids?id=SBSWaa9fL&amp;id=VVEnUNMkF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Global mobile app spending surpasses gaming, driven by AI</strong>: For the first time in 2025, global mobile app spending, reaching over $85 billion, surpassed gaming revenue, primarily fueled by generative AI apps like ChatGPT, which alone generated $3.4 billion in in-app purchases. <a href="https://app.daily.dev/posts/NzKmno9qu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New AI-powered UI builder for frontend development</strong>: Shadcn Studio MCP is an AI-powered tool that integrates with IDEs to generate production-ready UI components, blocks, and pages for React, Next.js, Vite, and Laravel, offering lifetime access and unlimited project usage. <a href="https://app.daily.dev/feed-by-ids?id=84wxnOe0I&amp;id=ncQILOQ9m" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-driven geothermal exploration challenges DOE estimates</strong>: Zanskar, a geothermal exploration startup, raised $115 million to use AI for identifying conventional geothermal resources, claiming a potential of 1 terawatt, significantly exceeding the U.S. Department of Energy’s 60 gigawatt estimate by 2050. <a href="https://app.daily.dev/posts/5U9R3y3j3" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Palantir CEO: AI will solve immigration, ignore the skeptics</strong>: Palantir CEO Alex Karp, speaking at the World Economic Forum, boldly declared that AI will eliminate the need for large-scale immigration in Western economies, except for specialized skills. This pronouncement comes despite widespread industry skepticism, with Gartner reporting AI in the ‘trough of disillusionment’ and PwC finding most CEOs see zero payoff from their AI investments. Apparently, Palantir’s AI is so magical it can solve complex societal issues while the rest of the industry is still trying to figure out how to make it profitable. <a href="https://app.daily.dev/posts/3s1vrBGbY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI world models: the next big thing, or just more learned hallucinations?</strong>: The tech world is buzzing about ‘AI world models’ and their potential convergence with game engines, promising new forms of interactive content. However, the article conveniently glosses over the fact that these AI models are prone to ‘learned hallucinations,’ meaning they might just make things up. So, while game engines prioritize control and efficiency, these new AI models prioritize speed, even if it means generating content that’s not entirely deterministic or, you know, real. Sounds like a recipe for some truly ‘innovative’ bugs. <a href="https://app.daily.dev/posts/OIj9HQhHA" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ChatGPT UI replication: looks good, breaks everything</strong>: Apparently, ChatGPT can replicate Apple demo UIs visually, which sounds impressive until you realize it completely misses the underlying system constraints. The generated implementations ‘break under performance load’ because ChatGPT lacks a ‘deep framework comprehension.’ So, it’s great at making things look pretty, but utterly useless for actual development. Maybe we should teach it to understand how things <em>actually</em> work before we let it design our next critical application. <a href="https://app.daily.dev/posts/LQOYGhWvV" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
