<h2>Must know</h2>
<ul>
<li><strong>Pentagon demands military access to AI models</strong>: The Pentagon is demanding AI companies, including Anthropic, allow military use of their technology for “all lawful purposes,” which includes autonomous weapons and mass surveillance. Anthropic is resisting these demands, particularly concerning autonomous weapons, risking the cancellation of its $200 million contract. This dispute highlights the growing tension between ethical AI development and national security interests, with other major AI players like OpenAI and Google also facing similar pressures. <a href="https://app.daily.dev/feed-by-ids?id=2BTVSGYkh&amp;id=VtyOOEppn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI doubles down on agentic AI with OpenClaw creator</strong>: OpenAI has hired Peter Steinberger, the creator of the open-source personal AI agent OpenClaw, to lead its next-generation agent development. OpenClaw, which automates tasks like email, will remain open source under a foundation supported by OpenAI. This move signals a significant push into agentic AI, where autonomous agents accomplish multi-step goals, with major tech companies like Microsoft also transforming operating systems into agentic platforms. <a href="https://app.daily.dev/posts/2BTVSGYkh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 enhances M.2 support and NFS performance</strong>: Linux 7.0 introduces a power sequencing driver for PCIe M.2 connectors, enabling Device Tree platforms to manage M.2 interfaces, a capability previously limited to ACPI firmware. This update, along with dynamic thread pool sizing for NFS servers, aims to optimize resource utilization and performance. Additionally, Ubuntu 26.04 LTS is splitting its <code>linux-firmware</code> package into 17 vendor-specific sub-packages to reduce update sizes and infrastructure strain. <a href="https://app.daily.dev/feed-by-ids?id=2zVNOvhqh&amp;id=8Lxd8Rnjy&amp;id=fP2szhpJX&amp;id=319goy50e" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Small team builds fast conversational AI model</strong>: Gradium’s Moshi model, developed by a small team, achieved 160ms latency for full-duplex conversational AI with only 7 billion parameters, outperforming larger labs in audio AI. <a href="https://app.daily.dev/posts/iRysiiuhz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Gwtar introduces new HTML archival format</strong>: A new HTML archival format, Gwtar, allows archiving gigabyte-sized web pages into a single, static, and efficient file that lazy-loads assets without special server software. <a href="https://app.daily.dev/posts/t2oeOs48x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Claude Code integrates with Google Search Console for SEO</strong>: A Python MCP server, <code>mcp-gsc</code>, integrates Google Search Console with Claude Code, enabling SEO analysis and fixes directly within a conversational interface. <a href="https://app.daily.dev/posts/rHujhRHLJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Palantir faces backlash over data contracts</strong>: Palantir is facing scrutiny over its data contracts with New York City hospitals and the UK’s NHS, with activists demanding termination due to privacy concerns and its work with ICE. <a href="https://app.daily.dev/feed-by-ids?id=i5hXbkiCe&amp;id=uL7dyR24I" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ex-L3Harris exec sold zero-days to Russia</strong>: A former L3Harris executive pleaded guilty to selling eight zero-day exploit kits to Russia, causing $35M in damages and national security harm. <a href="https://app.daily.dev/feed-by-ids?id=AAuuxbltu&amp;id=CtW0rQUoT&amp;id=N5siIFPEw&amp;id=fSewiiKKS&amp;id=jk8Zg0MX3&amp;id=nvKgzmU4A" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Shopify API adds fulfillment order weight field</strong>: Shopify’s Admin GraphQL API will include a new field in its 2026-04 release candidate to return the total weight of unfulfilled line items, aiding weight-based fulfillment decisions. <a href="https://app.daily.dev/posts/gtFPn8t2q" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Fedora Atomic comes to mobile with Pocketblue</strong>: Pocketblue is a work-in-progress project providing Fedora Atomic images for mobile devices like Xiaomi Pad 5/6 and OnePlus 6/6T, featuring toolbox support and FEX-EMU for x86 apps on ARM. <a href="https://app.daily.dev/posts/hapTm7n1X" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Architectural decay is real, just build better guardrails</strong>: Apparently, software systems just <em>drift</em> from their original design because of “collective amnesia” and “architectural decay.” The solution? Design systems where the “correct architectural path is the easiest to follow.” Because, you know, developers are just mindless drones who can’t possibly remember what they built last week, and documentation is for suckers. Just build it right the first time, problem solved! <a href="https://app.daily.dev/feed-by-ids?id=8ksTqV5kA&amp;id=SbkVDoSXT&amp;id=YhsHrjBDf" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Alibaba’s new AI image model struggles with basic tasks</strong>: Alibaba’s Quinn Image 2.0, an AI image generation model, promises improved typography and 2K resolution. However, one user’s attempt to generate a YouTube thumbnail resulted in aspect ratio issues and questionable output quality. So, while the marketing department is busy hyping up “advancements,” actual users are still struggling with basic functionality. Maybe it’s not the AI that doesn’t understand the assignment, but the people selling it. <a href="https://app.daily.dev/posts/zRxAENzbc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated bug reports are DDoSing open source</strong>: The creator of cURL, Daniel Stenberg, claims AI-generated security reports are “DDoSing” open source projects, with only 1 in 20-30 reports being accurate. This has led cURL to shut down its bug bounty program. So, AI is now so good at generating plausible-sounding garbage that it’s actively harming the very open-source community it’s supposed to be helping. Progress, right? <a href="https://app.daily.dev/posts/2BTVSGYkh" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
