<h2>Must know</h2>
<ul>
<li><strong>Cloud hardware costs skyrocket due to AI demand</strong>: Hetzner, a major European cloud provider, is hiking prices by up to 50% starting April 1, 2026, across all regions. This is due to a dramatic increase 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 costs to users, with RAM prices reportedly up 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 deal includes a performance-based warrant for AMD to issue Meta up to 160 million shares if AMD’s stock hits $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</strong>: Apple researchers introduced Ferret-UI Lite, a lightweight, 3B-parameter multimodal AI model designed for on-device UI understanding and interaction. This model aims for efficient inference without cloud processing, enabling visual perception and control of UIs directly on mobile, web, and desktop platforms. It prioritizes low latency, privacy, and offline capability, potentially reducing 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>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 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 dev tools and AI</strong>: A new npm supply chain attack, dubbed SANDWORM_MODE, is targeting developer utilities and AI coding tools like Claude Code, harvesting tokens and credentials to inject malicious code into CI pipelines. <a href="https://app.daily.dev/posts/q6c2RLQKH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OTelBench released 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. <a href="https://app.daily.dev/posts/OB1MlP8PG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Node.js 25.7.0 released with HTTP/2 and ESM updates</strong>: Node.js 25.7.0 introduces HTTP/2 <code>http1Options</code>, ESM support for Single Executable Applications, and a release candidate for the built-in SQLite module. <a href="https://app.daily.dev/posts/K4geCuziz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Elasticsearch v9 boosts vector search performance</strong>: Trendyol achieved a 4x performance boost in vector search by upgrading to Elasticsearch v9.2.1, showing significantly higher throughput and lower latency compared to v8.18.3 for hybrid search queries. <a href="https://app.daily.dev/posts/Iw8wLbMFe" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitLab extends GPG key validity</strong>: GitLab extended its GPG key expiration to February 2028 to maintain repository metadata 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>AI agent browser sandbox project announced</strong>: A new browser sandbox project aims to create a secure environment for AI agents to interact with the web, featuring “zero config” setup. <a href="https://app.daily.dev/posts/AY0JUdVv6" 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.” Because nothing says “innovation” like threatening a company for not wanting its AI to build autonomous weapons. Apparently, the DOD is so dependent on Anthropic that they forgot about that whole “avoid single-vendor reliance” directive. <a href="https://app.daily.dev/posts/vbdfUwGoN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI models in Cursor are now “faster”</strong>: OpenAI models integrated into Cursor are now “up to 30% faster” thanks to WebSockets with the Responses API, according to a promotional tweet. Because a 30% speed boost, especially when it’s “up to,” is definitely a game-changer for your coding workflow, not just a marketing number. <a href="https://app.daily.dev/posts/CNQCNZ2N0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft and Salesforce confirm “AI suspicions” (details withheld)</strong>: Microsoft Research and Salesforce AI analyzed over 200,000 AI conversations, reportedly confirming “industry suspicions.” The social media post, however, provides zero details on what those suspicions are or what was actually found. So, they confirmed something, but they’re not telling us what. Riveting. <a href="https://app.daily.dev/posts/FAzFsVs8B" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
