<h2>Must know</h2>
<ul>
<li><strong>Meta goes all in on AMD for AI</strong>: Meta just dropped a cool $100 billion on AMD chips, including MI540 GPUs, to supercharge its AI infrastructure. This massive multiyear deal aims to add 6GW of compute capacity, diversifying Meta’s AI stack beyond Nvidia and pushing towards ‘personal superintelligence.’ It’s a clear signal that the AI arms race is heating up, with Meta committing at least $600 billion to US AI infrastructure. <a href="https://app.daily.dev/posts/vYZNsZv8f" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Apple introduces on-device UI AI model</strong>: Apple researchers unveiled Ferret-UI Lite, a compact 3B-parameter multimodal AI model designed for on-device UI understanding and interaction. This lightweight model promises low latency, privacy, and offline capabilities across mobile, web, and desktop, aiming to reduce reliance on cloud processing for GUI agents. It’s a significant step towards more efficient and private AI-driven user interfaces, potentially lessening Siri’s dependence on Google Cloud. <a href="https://app.daily.dev/posts/qj9GoBvQx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google enhances Opal with agentic AI workflows</strong>: Google upgraded its Opal Workflow builder with new agentic capabilities, including an ‘Agent step’ that analyzes goals and calls tools automatically, powered by Gemini 3 Flash. This allows non-technical users to create complex automated workflows using text prompts, with memory support via Google Sheets. It’s a move towards making sophisticated AI agentic engineering accessible to a broader audience, competing with other natural-language app builders. <a href="https://app.daily.dev/posts/ZbbWKhN8A" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Quesma releases OTelBench for OpenTelemetry and AI performance</strong>: Quesma released OTelBench, an open-source benchmarking suite to evaluate OpenTelemetry pipeline performance and AI agent effectiveness in SRE tasks, revealing LLMs often achieve success rates below 30% in production-grade instrumentation scenarios. <a href="https://app.daily.dev/posts/OB1MlP8PG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Hetzner announces significant price hikes due to AI demand</strong>: Hetzner, a major European cloud provider, is increasing prices by up to 50% due to dramatic hardware cost increases, particularly for DRAM, NAND flash, and server CPUs, driven by the AI industry’s demand. <a href="https://app.daily.dev/posts/van1jyyDC" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitLab extends GPG key for repository metadata</strong>: GitLab extended its GPG key for repository metadata from February 27, 2026, to February 6, 2028, to ensure integrity and minimize disruption for users. <a href="https://app.daily.dev/posts/JTFdrzyr9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI model performance upgrade in Cursor</strong>: OpenAI models integrated into Cursor were upgraded using WebSockets with the Responses API, reportedly resulting in up to 30% faster performance for all Cursor users. <a href="https://app.daily.dev/posts/CNQCNZ2N0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>DMARCbis update discussion released</strong>: DMARCbis, an upcoming update to the DMARC email authentication standard, introduces clearer reporting expectations and standardized DNS tree walks to reduce ambiguity. <a href="https://app.daily.dev/posts/yEFoBF8zN" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Pentagon-Anthropic AI standoff over military use</strong>: The Pentagon issued a Friday deadline to Anthropic to remove AI guardrails on Claude for unrestricted military use, threatening to label them a ‘supply chain risk’ if they refuse. Because apparently, the DOD’s single-vendor reliance on Anthropic for classified frontier AI access is Anthropic’s problem, not the Pentagon’s brilliant procurement strategy. <a href="https://app.daily.dev/posts/vbdfUwGoN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Npm supply chain attack targets developer tools and AI</strong>: Socket researchers uncovered a new npm supply chain attack, SANDWORM_MODE, using 19 typosquatted packages to impersonate developer utilities and AI coding tools like Claude Code. The malware harvests credentials and injects malicious code, deploying an MCP server for prompt-injection attacks. So, while you’re busy trying to get your AI agent to write code, someone else’s AI agent is busy trying to steal your SSH keys. Progress! <a href="https://app.daily.dev/posts/q6c2RLQKH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft and Salesforce AI conversation analysis announced</strong>: Microsoft Research and Salesforce AI analyzed over 200,000 AI conversations, reportedly confirming ‘industry suspicions’ regarding AI. However, they provided absolutely no details about these suspicions or findings. It’s like saying, ‘We found something interesting, but we’re not telling you what it is.’ Pure clickbait, generating curiosity without delivering any actual intelligence. Thanks for the vague retweet, @rryssf_! <a href="https://app.daily.dev/posts/FAzFsVs8B" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
