<h2>Must know</h2>
<ul>
<li><strong>Pentagon demands military access to AI models</strong>: The Pentagon is pushing major AI companies, including Anthropic, OpenAI, Google, and xAI, to 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, as reported by Ars Technica. <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>Agentic AI is the next big thing, and it’s already here</strong>: Agentic AI, the evolution beyond generative AI, involves autonomous agents that can accomplish multi-step goals independently. Major tech companies like Microsoft, Google, IBM, and Amazon are developing orchestration platforms for these agents, with Microsoft even transforming Windows into an agentic operating system. OpenAI recently hired the creator of OpenClaw, a personal AI agent, to lead its next-gen agent development, signaling a significant shift towards more autonomous AI systems. <a href="https://app.daily.dev/posts/2BTVSGYkh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Open source is getting DDoSed by AI-generated bug reports</strong>: Daniel Stenberg, the creator of cURL, stated that AI-generated security reports are overwhelming open-source projects, with only 1 in 20-30 reports being accurate. This deluge has forced cURL to shut down its bug bounty program. While AI-powered analysis tools have found legitimate bugs, AI-generated code contributions remain unreliable, creating a significant burden for maintainers. <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>LLM inference speeds get a boost with new modes</strong>: Anthropic and OpenAI have introduced 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 with a smaller model. <a href="https://app.daily.dev/posts/e1zDCJ67S" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 brings M.2 power management and dynamic NFS threads</strong>: The latest Linux kernel, 7.0, introduces a power sequencing driver for PCIe M.2 connectors, enabling better management of M.2 interfaces on Device Tree platforms, and adds dynamic thread pool sizing for NFS servers to optimize resource utilization. <a href="https://app.daily.dev/feed-by-ids?id=2zVNOvhqh&amp;id=8Lxd8Rnjy" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New HTML archival format aims for static, single-file efficiency</strong>: Gwtar, a new HTML archival format, was introduced to create static, single-file web archives that are efficient and lazy-load assets via HTTP range requests into an embedded tarball, addressing future compatibility concerns. <a href="https://app.daily.dev/posts/t2oeOs48x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Gradium’s Moshi model advances audio AI with small team efficiency</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 expertise can outperform larger labs. <a href="https://app.daily.dev/posts/iRysiiuhz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Software architecture needs guardrails against decay</strong>: Software systems tend to drift from their original design, leading to ‘collective amnesia’; combat this by designing systems where the correct architectural path is the easiest to follow, using contracts and modular interfaces from day one. <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>Google Search Console integrates with Claude Code for SEO</strong>: A new Python MCP server, mcp-gsc, integrates Google Search Console with Claude Code, allowing SEO analysis and fixes within a single conversational interface, eliminating the need to switch between dashboards and code editors. <a href="https://app.daily.dev/posts/rHujhRHLJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LocalStack drops community edition, raises developer concerns</strong>: LocalStack is discontinuing its open-source Community Edition, consolidating into a single Docker image requiring user registration and a paid subscription for full access, sparking developer concerns about its open-source commitment. <a href="https://app.daily.dev/posts/fsky5EERm" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI image generation still struggles with basic assignments</strong>: Alibaba’s Quinn Image 2.0 boasts improved typography and 2K resolution, but one user found it couldn’t even generate a decent YouTube thumbnail, struggling with aspect ratios and output quality. So much for ‘improved typography’ when it can’t even understand the assignment. Maybe it needs a prompt injection to grasp basic design principles. <a href="https://app.daily.dev/posts/zRxAENzbc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Developers are having an existential crisis over AI</strong>: Apparently, software developers are experiencing ‘Deep Blue’ anxiety, fearing AI will devalue their skills because coding agents can now produce working software autonomously. One article even suggests over-reliance on AI could lead to ‘cognitive atrophy.’ Relax, folks, it’s just a fancy autocomplete. Your ‘passion’ for programming isn’t going anywhere, unless you let it. <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>Palantir’s data practices spark controversy, again</strong>: Palantir is under fire for its $4 million contract with NYC hospitals, granting it access to protected health information, while simultaneously suing a Swiss magazine for reporting on its attempts to secure government contracts. Because nothing says ‘we respect your privacy’ like collecting patient data and then trying to silence journalists. The ‘Streisand Effect’ is strong with this one. <a href="https://app.daily.dev/feed-by-ids?id=i5hXbkiCe&amp;id=uL7dyR24I" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
