<h2>Must know</h2>
<ul>
<li><strong>Musk’s companies eye a trillion-dollar merger</strong>: Elon Musk’s SpaceX, xAI, and potentially Tesla are in early talks to merge, aiming to create a super-company valued at over a trillion dollars. This consolidation could combine rockets, satellites, AI, and electric vehicles, with a key driver being the plan for space-based AI data centers to cut training costs. Corporate filings in Nevada suggest these discussions are serious, as reported by multiple sources. <a href="https://app.daily.dev/posts/wMYF638eZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google DeepMind launches Project Genie AI world generator</strong>: Google DeepMind has unveiled Project Genie, an experimental web prototype powered by the Genie 3 world model, capable of generating interactive environments in real-time. Available to Google AI Ultra subscribers, it allows users to create and explore dynamic worlds using text and image prompts. While impressive for whimsical styles, it struggles with photorealism and object interaction, and the gaming industry is already concerned about AI’s negative impact, according to #5WiIjyYNP and #cXEPGpaCA. <a href="https://app.daily.dev/posts/iit38VC43" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubernetes RCE flaw identified in common permissions</strong>: Edera Advisory has uncovered a critical design flaw in Kubernetes that allows remote code execution (RCE) in any container on a node through commonly used <code>nodes/proxy GET</code> permissions. This poses a significant risk, especially if monitoring tools or network overlays are compromised. While a fix is slated for Kubernetes 1.36, many production clusters will take time to update, necessitating regular RCE scans and fine-grained authorization controls. <a href="https://app.daily.dev/posts/phMyeDnVg" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Meta doubles down on AI infrastructure, Reality Labs pivots to wearables</strong>: Meta plans to invest up to $135 billion in AI infrastructure in 2026, nearly doubling last year’s spend, to support its Superintelligence Labs and integrate LLMs with its core platforms, while its Reality Labs division, despite $19.2 billion in losses, is pivoting towards AI-integrated glasses and wearables. <a href="https://app.daily.dev/feed-by-ids?id=JHGPGk0MW&amp;id=ALCffHVkc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft’s hybrid AI chip strategy and Xbox struggles</strong>: Microsoft has deployed its in-house Maia 200 AI inference chip in data centers for frontier AI models, including OpenAI’s, but will continue purchasing chips from Nvidia and AMD, while its Xbox gaming division reported significant declines in Q2 FY26, with console sales down 32% and gaming revenue dropping 9%. <a href="https://app.daily.dev/feed-by-ids?id=kN85HgrLU&amp;id=uKbAAElcM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google Cloud brings conversational analytics to BigQuery</strong>: Google Cloud has launched Conversational Analytics in BigQuery (preview), an AI-powered agent integrated into BigQuery Studio that uses Gemini models to enable natural language data analysis, generating and visualizing SQL queries without requiring users to write code. <a href="https://app.daily.dev/posts/2YKMeABdR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Apiiro launches Guardian Agent for AI code security</strong>: Apiiro introduced Guardian Agent, an AI security tool designed to prevent AI coding agents from generating vulnerable code by rewriting developer prompts into secure ones in real-time within IDEs and CLIs. <a href="https://app.daily.dev/posts/5AaMXyHJd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA GeForce NOW now supports native Linux</strong>: NVIDIA’s GeForce NOW cloud gaming service has launched native Linux PC support in beta, offering up to 5K resolution at 120fps or 1080p at 360fps on Ubuntu 24.04 LTS and later, distributed as a Flatpak. <a href="https://app.daily.dev/posts/RB8XUehQ7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Apple acquires audio AI startup <a href="http://Q.ai" target="_blank" rel="noopener nofollow">Q.ai</a></strong>: Apple has acquired Israeli audio AI startup <a href="http://Q.ai" target="_blank" rel="noopener nofollow">Q.ai</a> for an estimated $1.6 billion to $2 billion, strengthening its AI hardware capabilities with technology specializing in whispered speech recognition and audio enhancement, expected to benefit AirPods and Vision Pro. <a href="https://app.daily.dev/posts/PknWg8Xks" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>formae IaC platform adds multi-cloud support and Plugin SDK</strong>: Platform Engineering Labs announced beta multi-cloud support for its open-source IaC platform, formae, now including Google Cloud, Azure, Oracle Cloud, and OVHcloud, alongside a new Plugin SDK for custom extensions. <a href="https://app.daily.dev/posts/bqzmDPKsL" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>GDC report says AI harms industry, but adoption grows anyway</strong>: The 2026 GDC State of the Industry Report reveals that 52% of game developers believe generative AI harms the industry, up from 30% last year. Yet, generative AI adoption still grew to 36% among professionals. It’s almost as if developers are being forced to use tools they don’t trust, probably by management who think ‘AI’ is a magic word for ‘more profit, less people.’ <a href="https://app.daily.dev/posts/REyf1LA3H" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI’s Prism: more ‘AI slop’ for scientific research</strong>: OpenAI launched Prism, a free AI workspace integrating GPT-5.2 with LaTeX for scientific writing, which has researchers worried about a flood of low-quality, AI-generated papers. Because what the world really needs is more polished-looking but scientifically weak manuscripts overwhelming peer review, right? It’s like giving everyone a fancy pen but forgetting to teach them how to write anything meaningful, as Ars Technica points out. <a href="https://app.daily.dev/posts/Rz7pKpx22" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Human-in-the-loop: the new bottleneck, apparently</strong>: Turns out, having humans review every single AI decision might actually slow things down. Who knew? A new ‘tiered architecture’ promises to fix this by only bothering humans with the ‘uncertain or high-risk’ cases, claiming a 73% reduction in review time. So, we’re just admitting our AI isn’t good enough to trust, but we’ll still let it handle 85% of decisions unsupervised. What could possibly go wrong? <a href="https://app.daily.dev/posts/1FaR1iNmw" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
