<h2>Must know</h2>
<ul>
<li><strong>Pentagon demands military access to AI models</strong>: The Pentagon is demanding major AI companies, including Anthropic, OpenAI, Google, and xAI, 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 highlights a growing tension between ethical AI development and national security interests, as reported by sources like Ars Technica. <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 new foundation with OpenAI’s support. 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>Linux 7.0 brings M.2 power management and exFAT improvements</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 on x86_64. This update, along with dynamic thread pool sizing for NFS servers and multi-cluster support for exFAT, significantly enhances hardware compatibility and performance for Linux systems. Ubuntu 26.04 LTS is also splitting its large <code>linux-firmware</code> package into 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>AI coding models get speed boosts with trade-offs</strong>: Anthropic and OpenAI announced new “fast modes” for their coding models, with Anthropic achieving 2.5x speed by reducing batch sizes at 6x 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>Claude Code gets Google Search Console integration</strong>: A new Python MCP server, <code>mcp-gsc</code>, integrates Google Search Console with Claude Code, allowing 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>Small team builds low-latency conversational AI</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, 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>New HTML archival format Gwtar released</strong>: A new HTML archival format called Gwtar was introduced, aiming to solve the “trilemma” of being static, single-file, and efficient by combining an HTML+JavaScript header with an embedded tarball of assets. <a href="https://app.daily.dev/posts/t2oeOs48x" 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’s modern version, allowing code execution via malicious Markdown links, which Microsoft has since patched. <a href="https://app.daily.dev/feed-by-ids?id=AAuuxbltu&amp;id=CtW0rQUoT&amp;id=fSewiiKKS&amp;id=jk8Zg0MX3&amp;id=nvKgzmU4A" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Palantir under fire for data contracts</strong>: Palantir is facing scrutiny over its data contracts with NYC 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>US considers easing tech restrictions on Chinese firms</strong>: The US government is considering reversing some tech restrictions on Chinese companies, potentially removing Alibaba, Baidu, and BYD from military-linked lists and lifting bans on certain telecom companies. <a href="https://app.daily.dev/posts/d9SXCk8xt" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Developers get “Deep Blue” anxiety over AI coding</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. Because, you know, the entire craft of software engineering is just typing code, and certainly not problem-solving, system design, or understanding complex requirements. Good thing we have articles to reassure us that our “accumulated experience” is still valuable, even if we’re just babysitting AI now. <a href="https://app.daily.dev/feed-by-ids?id=DwIZWcSFk&amp;id=D9oPYSzsO&amp;id=G48nTExd2&amp;id=NY3oBIud2&amp;id=SKlOXm4Tn&amp;id=jGb0yTdHY&amp;id=sNnVdBWnz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding assistants will make your brain mush</strong>: An opinion piece warns that over-reliance on AI coding assistants might lead to “cognitive atrophy” in programmers, suggesting we’ll lose our problem-solving skills if we don’t actively practice coding logic. The author compares it to London cab drivers’ hippocampi vs. GPS users. Because, clearly, using a tool to automate repetitive tasks is the same as outsourcing your brain entirely. 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 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’s attempt to generate a YouTube thumbnail resulted in issues with aspect ratio and output quality, suggesting it’s not quite the intuitive, magical tool advertised. It seems even with “2.0” in the name, some AI models still struggle to understand the assignment. <a href="https://app.daily.dev/posts/zRxAENzbc" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
