<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 reportedly 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 OpenAI, Google, and xAI facing similar pressures. <a href="https://app.daily.dev/feed-by-ids?id=VtyOOEppn&amp;id=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&amp;id=szGxppNbh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Combating architectural decay and performance regressions</strong>: Software systems inevitably drift from their original design, leading to architectural decay. To combat this, design systems where the correct architectural path is the easiest to follow, using contracts as constraints and building modular interfaces from day one. Automated performance guardrails integrated into CI/CD pipelines are also crucial, especially with AI-generated code, which often prioritizes syntax over efficiency. <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>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI model inference speeds get a boost</strong>: Anthropic and OpenAI announced distinct “fast modes” for their coding models, with Anthropic achieving 2.5x speed by reducing batch sizes at 6x cost, and OpenAI achieving 15x speed using specialized Cerebras chips with a smaller model. <a href="https://app.daily.dev/posts/e1zDCJ67S" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Small team achieves breakthrough in audio AI latency</strong>: Gradium’s Moshi, a 7-billion parameter conversational AI model, achieved 160ms latency, demonstrating that small, specialized teams can outperform major AI labs in audio AI. <a href="https://app.daily.dev/posts/iRysiiuhz" 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’s public hospital system and the UK’s NHS, with activists demanding termination due to concerns over patient data security 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>Google search console integrates with claude code for SEO</strong>: A Python MCP server, <code>mcp-gsc</code>, integrates Google Search Console with Claude Code, enabling SEO analysis and fixes within a single conversational interface. <a href="https://app.daily.dev/posts/rHujhRHLJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New HTML archival format gwtar introduced</strong>: Gwtar, a new HTML archival format, creates a single polyglot file combining HTML+JavaScript with a tarball of assets, enabling lazy-loading and archiving gigabyte-sized web pages without special server software. <a href="https://app.daily.dev/posts/t2oeOs48x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Former L3Harris executive sold zero-days to Russia</strong>: A former L3Harris cyber subsidiary 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>Rust’s boring nature is its strength</strong>: Rust’s “boring” nature, stemming from its compiler catching critical errors at compile time, is highlighted as its greatest strength, leading to predictable and reliable code. <a href="https://app.daily.dev/posts/CrbhK2GEI" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Developers fear AI will steal their jobs</strong>: Apparently, software developers are experiencing “Deep Blue” anxiety because AI coding agents like Claude Opus 4.5/4.6 and GPT-5.2/5.3 can now produce working, tested software autonomously. The article suggests this is causing “genuine mental anguish” in developer communities. Because, you know, the ability to write boilerplate code means all that accumulated experience and understanding of complex systems just vanishes. Right. <a href="https://app.daily.dev/posts/DwIZWcSFk" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding assistants will make you dumb</strong>: An opinion piece warns that over-reliance on AI coding assistants might lead to “cognitive atrophy” in programmers, suggesting they could lose problem-solving abilities. The author draws an analogy to London cab drivers and GPS users, implying that convenience comes at the cost of diminished skills. Because, clearly, using tools to be more efficient is the same as losing your brain. Next, they’ll tell us calculators make us bad at math. <a href="https://app.daily.dev/posts/qtVi65Njq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Alibaba’s new AI image model struggles with basic tasks</strong>: Alibaba released Quinn Image 2.0, an AI image generation model, claiming improved typography, 2K resolution, and faster inference. However, one user’s attempt to generate a YouTube thumbnail resulted in issues with aspect ratio and output quality, suggesting it’s not as intuitive or capable as advertised. It seems even AI models can’t always “understand the assignment,” despite all the hype. <a href="https://app.daily.dev/posts/zRxAENzbc" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
