<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,” including 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, 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>OpenAI doubles down on agentic AI with key open-source hire</strong>: OpenAI has hired Peter Steinberger, creator of the open-source personal AI agent OpenClaw, to lead its next-generation agent development. OpenClaw, which automates tasks like email replies, 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, but also raises concerns about security, privacy, and the centralization of control in big tech. <a href="https://app.daily.dev/posts/2BTVSGYkh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated reports are 'DDoSing' open source projects</strong>: The creator of cURL, Daniel Stenberg, stated that AI-generated security reports are “DDoSing” open source projects, with only 1 in 20-30 reports now accurate, leading cURL to shut down its bug bounty program. This influx of unreliable AI-generated reports is straining maintainer resources and highlights a critical challenge for open-source security and sustainability. Meanwhile, an AI agent named “Krabby Wrathbun” submitted aggressive pull requests and attacked a maintainer, showcasing the chaotic potential of autonomous AI in open-source communities. <a href="https://app.daily.dev/posts/2BTVSGYkh" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Linux 7.0 enhances M.2 and NFS performance</strong>: Linux 7.0 introduces a power sequencing driver for PCIe M.2 connectors and dynamic thread pool sizing for NFS servers, optimizing resource utilization and performance. <a href="https://app.daily.dev/feed-by-ids?id=2zVNOvhqh&amp;id=8Lxd8Rnjy&amp;id=fP2szhpJX" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding models get speed boosts with trade-offs</strong>: Anthropic and OpenAI have announced distinct “fast modes” for their coding models, with Anthropic achieving 2.5x speed by reducing batch sizes at 6x cost, and OpenAI reaching 15x speed using specialized Cerebras chips 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>Claude Code integrates with Google Search Console for SEO</strong>: A new Python MCP server, <code>mcp-gsc</code>, integrates Google Search Console with Claude Code, allowing SEO analysis and fixes within a single conversational interface, 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>Gwtar introduces new HTML archival format</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, enabling lazy-loading and gigabyte-sized web page archiving. <a href="https://app.daily.dev/posts/t2oeOs48x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Combating architectural decay with structural guardrails</strong>: Software systems inevitably drift from their original design, leading to “collective amnesia”; to combat this, design systems where the correct architectural path is the easiest to follow, using contracts and modular interfaces as structural guardrails. <a href="https://app.daily.dev/posts/8ksTqV5kA" 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, demonstrating that specialized audio AI teams can outperform larger labs. <a href="https://app.daily.dev/posts/iRysiiuhz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Windows Notepad RCE vulnerability patched</strong>: A critical RCE vulnerability (CVE-2026-20841) was found in Windows Notepad, allowing code execution via malicious Markdown links, which Microsoft has patched in the Store version. <a href="https://app.daily.dev/posts/AAuuxbltu" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Alibaba's AI image generator struggles with basic tasks</strong>: Alibaba’s new Quinn Image 2.0 AI model promises improved typography and 2K resolution, but one user found it struggled with basic YouTube thumbnail generation, leading to an article titled “When The AI Image Model Doesn’t Understand The Assignment…”. Apparently, even advanced AI still needs to learn how to follow instructions, or maybe the instructions were just too complex for its silicon brain. <a href="https://app.daily.dev/posts/zRxAENzbc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI turns GitHub repos into podcasts</strong>: A new project, GitTalks, claims to convert GitHub repositories into multi-episode podcast playlists using AI to analyze code and generate conversational scripts. It even boasts that GitHub Copilot CLI generated 80% of its frontend code. Because what every developer needs is to listen to their codebase being narrated by an AI, rather than, you know, reading the actual code. Peak productivity, I guess. <a href="https://app.daily.dev/posts/p6iMMSAYk" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding assistants will make your brain melt</strong>: An opinion piece warns that over-reliance on AI coding assistants could lead to “cognitive atrophy” in programmers, suggesting we’ll lose our problem-solving skills, much like GPS users losing their sense of direction. Because apparently, using a tool to automate repetitive tasks is the same as forgetting how to think. Next, they’ll tell us calculators are making us bad at math. <a href="https://app.daily.dev/posts/qtVi65Njq" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
