<h2>Must know</h2>
<ul>
<li><strong>Pentagon demands military access to AI models</strong>: The Pentagon is strong-arming AI companies, including Anthropic, to allow military use of their technology for “all lawful purposes,” which apparently includes autonomous weapons and mass surveillance. Anthropic is pushing back, risking a $200 million contract. Meanwhile, OpenAI, Google, and xAI are also facing similar demands, with varying levels of compliance. This isn’t just about a contract; it’s about who controls the ethical boundaries of AI when it comes to military applications, as reported by sources like #VtyOOEppn. <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>Linux 7.0 brings M.2 power management and NFS improvements</strong>: Linux 7.0 is rolling out with some notable updates, including a new power sequencing driver for PCIe M.2 connectors, enabling Device Tree platforms to manage M.2 interfaces. It also brings dynamic thread pool sizing for NFS servers, allowing the kernel to automatically adjust threads for better resource utilization. On top of that, Ubuntu 26.04 LTS is splitting its massive <code>linux-firmware</code> package into 17 vendor-specific sub-packages, which should make updates smaller and more targeted, as detailed in sources like #2zVNOvhqh and #8Lxd8Rnjy. <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 suffer from “collective amnesia” and architectural decay as teams and requirements change. To fight this, the advice is to design systems where the correct architectural path is the easiest one to follow, using contracts and modular interfaces from day one. Automated performance guardrails are also becoming critical, especially with AI-generated code, which often prioritizes syntax over efficiency, as highlighted by sources like #8ksTqV5kA and #SbkVDoSXT. <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 coding models get speed boosts with trade-offs</strong>: Anthropic and OpenAI have introduced “fast modes” for their coding models, with Anthropic achieving 2.5x speed by reducing batch sizes at 6x the cost, and OpenAI hitting 15x speed using specialized Cerebras chips but with a smaller, less capable 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 conversational AI latency</strong>: Gradium, a small team from Kyutai, developed Moshi, the first full-duplex conversational AI model to achieve 160ms latency with only 7 billion parameters, showcasing how specialized teams can outperform larger labs in audio AI. <a href="https://app.daily.dev/posts/iRysiiuhz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Gwtar format for efficient HTML archiving</strong>: A new HTML archival format called Gwtar aims to solve the “trilemma” of being static, single-file, and efficient by combining an HTML+JavaScript header with a tarball of assets, allowing lazy-loading and gigabyte-sized web pages in a single file. <a href="https://app.daily.dev/posts/t2oeOs48x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Palantir under fire for data contracts and privacy concerns</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 concerns over patient data security and Palantir’s 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>Rust’s “boring” nature hailed as a strength</strong>: Rust is being described as “boring” by developers, but this is framed as its greatest strength, as its compiler catches critical errors at compile time, leading to predictable and reliable code and fewer late-night debugging sessions. <a href="https://app.daily.dev/posts/CrbhK2GEI" 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 and eliminating the need to switch between multiple tools. <a href="https://app.daily.dev/posts/rHujhRHLJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Symfony 8.1 development progresses with new features</strong>: Symfony 8.1 development saw 120 pull requests merged, including bug fixes for Redis authentication and new features like tag decoration support in DependencyInjection and error indicators in the profiler. <a href="https://app.daily.dev/posts/tiyCa1O2y" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI is “DDoSing” open source with bad security reports</strong>: Apparently, AI is now “DDoSing” open source projects with security reports, according to Daniel Stenberg, creator of cURL. He claims only 1 in 20-30 AI-generated reports are accurate, forcing cURL to shut down its bug bounty program. Because, you know, AI is so good at finding bugs, it’s actually making it harder for humans to find real ones. What a time to be alive. <a href="https://app.daily.dev/feed-by-ids?id=2BTVSGYkh&amp;id=7e4x0vONT&amp;id=KdUKcvk8L&amp;id=UMatWajxo&amp;id=VtyOOEppn&amp;id=vb6x1hImO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Developers fear AI-induced “cognitive atrophy”</strong>: Developers are experiencing “Deep Blue” anxiety, fearing AI will devalue their skills, especially with coding agents like Claude Opus 4.5/4.6 and GPT-5.2/5.3 supposedly producing working software autonomously. One article even warns that over-reliance on AI coding assistants could lead to “cognitive atrophy.” Because, clearly, the only way to be a good programmer is to suffer through every line of code yourself, like some kind of digital Sisyphus. <a href="https://app.daily.dev/feed-by-ids?id=DwIZWcSFk&amp;id=qtVi65Njq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Alibaba’s Quinn Image 2.0 struggles with basic tasks</strong>: Alibaba released Quinn Image 2.0, an AI image generation model, claiming improved typography and 2K resolution. However, one user trying to generate a YouTube thumbnail found it wasn’t intuitive and had issues with aspect ratio and output quality. So much for those “improved” claims; seems like the AI still doesn’t quite “understand the assignment,” as the article so subtly puts it. <a href="https://app.daily.dev/posts/zRxAENzbc" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
