<h2>Must know</h2>
<ul>
<li><strong>Meta goes all in on AMD for AI chips</strong>: Meta has inked a multiyear deal with AMD, reportedly worth up to $100 billion, to power its global data centers with AMD’s MI540 GPUs and CPUs. This massive investment aims to add 6GW of compute capacity, accelerating Meta’s AI model development and diversifying its AI compute stack beyond Nvidia. Mark Zuckerberg’s vision of ‘personal superintelligence’ is apparently going to be built on a lot of AMD silicon. <a href="https://app.daily.dev/posts/vYZNsZv8f" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Hetzner price hikes due to AI demand</strong>: Hetzner, a major European cloud provider, is increasing prices by up to 50% starting April 1, 2026, citing dramatic increases in hardware costs for DRAM, NAND flash, and server CPUs. The culprit? The insatiable demand from the AI industry, which is straining global supply chains. This highlights how smaller providers, unlike hyperscalers with long-term contracts, are directly passing these costs to users. <a href="https://app.daily.dev/posts/van1jyyDC" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Pentagon and Anthropic clash over AI guardrails</strong>: The Pentagon has issued an ultimatum to Anthropic, demanding the removal of AI guardrails on Claude for unrestricted military use, including mass surveillance and autonomous weapons. Anthropic is refusing, leading to threats from the Pentagon to label them a ‘supply chain risk’ or invoke the Defense Production Act. This standoff reveals the DOD’s single-vendor reliance on Anthropic for classified frontier AI access and the growing tension between ethical AI development and military applications. <a href="https://app.daily.dev/posts/vbdfUwGoN" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Apple introduces on-device UI AI model</strong>: Apple researchers unveiled Ferret-UI Lite, a lightweight 3B-parameter multimodal AI model designed for on-device understanding and interaction with user interfaces, prioritizing low latency, privacy, and offline capability. <a href="https://app.daily.dev/posts/qj9GoBvQx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>npm supply chain attack targets dev tools and AI</strong>: Socket researchers uncovered a new npm supply chain attack, SANDWORM_MODE, using typosquatted packages to steal npm/GitHub tokens and cloud credentials, even deploying malicious servers into AI coding assistant configurations for prompt-injection attacks. <a href="https://app.daily.dev/posts/q6c2RLQKH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Quesma releases OTelBench for OpenTelemetry and AI evaluation</strong>: Quesma launched OTelBench, an open-source benchmarking suite to evaluate OpenTelemetry pipeline performance and assess 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>Node.js 25.7.0 released with HTTP/2 and SEA improvements</strong>: Node.js 25.7.0 is out, featuring HTTP/2 <code>http1Options</code>, ESM entry point support for Single Executable Applications, a release candidate for the built-in SQLite module, and various fixes and upgrades. <a href="https://app.daily.dev/posts/K4geCuziz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Platform Engineering Labs expands formae with multi-cloud support</strong>: Platform Engineering Labs updated formae, its open-source IaC platform, with beta support for GCP, Azure, OCI, and OVHcloud, alongside a redesigned Plugin SDK and AI-assisted development tools, aiming to simplify infrastructure codification without fragile state files. <a href="https://app.daily.dev/posts/oAyNaPkCA" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>DMARCbis update discussion released</strong>: Sendmarc published a fireside chat detailing DMARCbis (draft-ietf-dmarc-dmarcbis-41), an upcoming evolution of the DMARC email authentication standard with clearer reporting expectations and standardized DNS tree walks for receiver-side policy discovery. <a href="https://app.daily.dev/posts/yEFoBF8zN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>JFrog offers 99.99% uptime SLA for software supply chain</strong>: JFrog introduced Premium Availability, a new tier providing a contractual 99.99% in-region uptime SLA for its software supply chain platform, reducing allowable annual downtime to just 52 minutes for critical AI workloads and high-velocity CI/CD pipelines. <a href="https://app.daily.dev/posts/6JJpZNvHL" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Google’s opal workflow builder: more ‘vibe-coding’ than breakthrough</strong>: Google upgraded its Opal Workflow builder with new ‘agentic’ capabilities, powered by Gemini 3 Flash, allowing users to create automated workflows with text prompts. They’re calling it a ‘vibe-coding app’ and claiming it lets non-technical users build complex mini-apps. Because apparently, we needed another tool to generate code that still needs a developer to fix it, but now with more ‘vibe.’ <a href="https://app.daily.dev/posts/ZbbWKhN8A" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent browser sandbox project announced</strong>: A new ‘browser sandbox’ project was announced by <code>@nickscamara_</code>, promising a ‘secure browser environment for AI agents to interact with the web’ with ‘zero config.’ The announcement is from 2026-02-24, which is in the future, and the details are conveniently ‘truncated.’ So, a future project with no details, from the person promoting it. Sounds like a solid plan for world domination, or at least a good way to get some early buzz for something that might never materialize. <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: the big reveal that wasn’t</strong>: Microsoft Research and Salesforce AI supposedly analyzed over 200,000 AI conversations and ‘confirmed industry suspicions.’ The catch? They didn’t actually tell us what those suspicions were, or what they found. This vague, social media-sourced ‘announcement’ is less about sharing insights and more about generating clickbait. Because nothing says ‘groundbreaking research’ like a retweet that tells you absolutely nothing. <a href="https://app.daily.dev/posts/FAzFsVs8B" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
