<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 apparently includes autonomous weapons and mass surveillance. Anthropic is resisting these demands, particularly on autonomous weapons, risking a $200 million contract cancellation. This dispute highlights the ethical tightrope AI developers walk, especially when their tech is reportedly used in operations like the capture of Venezuelan President Nicolás Maduro, as detailed 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 foundation with OpenAI’s backing. This move signals a significant push into agentic AI, where autonomous agents accomplish multi-step goals, a field major tech companies like Microsoft are also heavily investing in, transforming operating systems like Windows into agentic platforms. <a href="https://app.daily.dev/posts/2BTVSGYkh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 and Ubuntu 26.04 bring core infrastructure 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 release also adds dynamic thread pool sizing for NFS servers, optimizing resource utilization, and multi-cluster support for exFAT, improving sequential read performance. Ubuntu 26.04 LTS is further refining the ecosystem by 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" 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 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>Combatting architectural decay with structural guardrails</strong>: Software systems inevitably drift from their original design, so architectural guardrails that make the correct path the easiest to follow are more effective than documentation alone. <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>Small team outperforms big AI labs in audio 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, showcasing small team efficiency in audio AI. <a href="https://app.daily.dev/posts/iRysiiuhz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>US considers easing China tech restrictions, India tightens deepfake rules</strong>: The US government is considering reversing some tech restrictions on Chinese companies, potentially removing Alibaba, Baidu, and BYD from military-linked lists, while India introduces strict regulations for social media platforms to remove AI-generated deepfakes within two hours. <a href="https://app.daily.dev/posts/d9SXCk8xt" 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, enabling lazy-loading and gigabyte-sized web page archiving without special server software. <a href="https://app.daily.dev/posts/t2oeOs48x" 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 in DependencyInjection and cURL copy/paste in WebProfiler. <a href="https://app.daily.dev/posts/tiyCa1O2y" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PostgreSQL JDBC fixes date regression</strong>: PostgreSQL JDBC 42.7.10 was released to fix a regression related to proleptic dates, addressing issues introduced in previous versions. <a href="https://app.daily.dev/posts/8PETdQgUV" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Developers get “Deep Blue” anxiety over AI, told to chill</strong>: Apparently, software developers are experiencing “Deep Blue” anxiety because AI coding agents can now produce working, tested software autonomously. The author of this piece, however, reassures us that accumulated developer experience is still valuable. So, don’t worry, your years of debugging obscure memory leaks still count for something, even if a bot can now write your boilerplate faster than you can open your IDE. <a href="https://app.daily.dev/posts/DwIZWcSFk" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Pharmacy API vulnerability allowed super admin access</strong>: A security researcher found critical API vulnerabilities in Dava India Pharmacy, allowing unauthorized super admin access to 883 stores, nearly 17,000 customer orders, and the ability to create 100% discount coupons. The issue was responsibly disclosed and fixed, with no reported data leaks. So, another day, another reminder that if you’re running an online pharmacy, maybe double-check your API security before someone gives away all your meds for free. <a href="https://app.daily.dev/posts/yhJRbeJMO" 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 and 2K resolution. However, one user’s attempt to generate a YouTube thumbnail resulted in issues with aspect ratio and output quality, suggesting the model might not “understand the assignment.” It seems even with all the AI hype, getting a bot to make a decent thumbnail is still harder than it looks. <a href="https://app.daily.dev/posts/zRxAENzbc" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
