<h2>Must know</h2>
<ul>
<li><strong>Cloud provider Hetzner hikes prices up to 50% due to AI-driven hardware costs</strong>: Hetzner, a major European cloud provider, is increasing prices by up to 50% starting April 1, 2026, across all regions. This is primarily due to a dramatic surge in hardware costs, particularly for DRAM, NAND flash, and server CPUs, driven by the AI industry’s insatiable demand. Unlike hyperscalers with long-term vendor contracts, smaller providers like Hetzner are directly passing these increased costs to users, with RAM prices alone rising 80-400%. <a href="https://app.daily.dev/posts/van1jyyDC" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Meta inks massive AI chip deal with AMD</strong>: Meta has signed a multiyear deal with AMD, reportedly worth up to $100 billion, to acquire MI540 GPUs and CPUs for its data centers. This massive investment aims to add 6GW of compute capacity to accelerate AI model development, diversifying Meta’s AI compute stack beyond Nvidia. The agreement includes a performance-based warrant for AMD to issue Meta up to 160 million shares, contingent on AMD’s stock reaching $600. <a href="https://app.daily.dev/posts/vYZNsZv8f" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Apple unveils on-device UI AI model Ferret-UI Lite</strong>: Apple researchers introduced Ferret-UI Lite, a lightweight, 3B-parameter multimodal AI model designed for on-device UI understanding and interaction. This model enables visual perception and control of user interfaces directly on mobile, web, and desktop platforms without cloud processing, prioritizing low latency, privacy, and offline capability. It achieved 91.6% on the ScreenSpot-V2 GUI grounding benchmark, competing with much larger models, and could reduce Siri’s reliance on Google Cloud. <a href="https://app.daily.dev/posts/qj9GoBvQx" 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 observability scenarios. <a href="https://app.daily.dev/posts/OB1MlP8PG" 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, allowing users to create automated workflows using text prompts and enabling an “Agent step” that analyzes goals and calls tools automatically, powered by Gemini 3 Flash. <a href="https://app.daily.dev/posts/ZbbWKhN8A" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NPM supply chain attack targets developer tools and AI</strong>: A new npm supply chain attack, SANDWORM_MODE, targets developer utilities and AI coding tools like Claude Code, harvesting npm/GitHub tokens and cloud credentials to inject malicious code into CI pipelines and enable prompt-injection attacks. <a href="https://app.daily.dev/posts/q6c2RLQKH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>DMARCbis update discussion released</strong>: DMARCbis (draft-ietf-dmarc-dmarcbis-41), an upcoming update to the DMARC email authentication standard, introduces clearer reporting expectations and standardized DNS tree walks for receiver-side policy discovery, as discussed in a Sendmarc fireside chat. <a href="https://app.daily.dev/posts/yEFoBF8zN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Node.js 25.7.0 released</strong>: Node.js 25.7.0 (Current) has been released, featuring HTTP/2 <code>http1Options</code>, ESM entry point support for Single Executable Applications, and an upgraded npm to 11.10.1. <a href="https://app.daily.dev/posts/K4geCuziz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cilium 1.19 enhances Kubernetes networking</strong>: Cilium 1.19 introduces BPF Host Routing with IPsec, Gateway API v1.4 support with native GRPCRoute, stable Multi-Pool IPAM, and initial Ztunnel integration for ambient mesh-style transparent encryption. <a href="https://app.daily.dev/posts/zYLvcHpYh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Debezium joins Google Summer of Code 2026</strong>: Debezium is participating in Google Summer of Code 2026, offering project ideas including source connectors for Milvus and SQLite, a Debezium CLI tool, and PyDebeziumAI for AI integration via Python. <a href="https://app.daily.dev/posts/XH8DOZHBb" 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 an ultimatum to Anthropic: remove AI guardrails on Claude for unrestricted military use or face designation as a ‘supply chain risk’ or Defense Production Act invocation. Because apparently, the DOD’s single-vendor reliance on Anthropic for classified frontier AI access means they get to dictate ethical boundaries. Who needs principles when you have a deadline? <a href="https://app.daily.dev/posts/vbdfUwGoN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent browser sandbox project announced</strong>: A new project called “browser sandbox” was announced by <code>@nickscamara_</code> for 2026-02-24, promising a “secure browser environment for AI agents to interact with the web” with “zero config.” It’s a pre-announcement, so details are scarce, but the hype is already building for this future tech that will surely revolutionize… something. Probably. <a href="https://app.daily.dev/posts/AY0JUdVv6" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft Research and Salesforce AI conversation analysis announced</strong>: Microsoft Research and Salesforce AI supposedly analyzed over 200,000 AI conversations, confirming “industry suspicions” about AI. The tweet from @rryssf_ on 2026-02-24, however, provides zero details on what those suspicions are or what was actually found. So, we’re left with a vague, clickbait-y announcement that confirms nothing but the industry’s love for generating curiosity without delivering information. <a href="https://app.daily.dev/posts/FAzFsVs8B" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
