<h2>Must know</h2>
<ul>
<li><strong>AI agents are a security nightmare</strong>: Multiple reports highlight critical security vulnerabilities in AI agent platforms and applications. Moltbook, a social network for AI agents, exposed 1.5 million API keys and private messages due to misconfigured databases. Similarly, the OpenClaw AI agent project, despite its popularity, has critical flaws including data exfiltration and exposed credentials, with thousands of instances discoverable online. These incidents underscore the risks of prioritizing development speed over secure defaults in AI-coded applications, especially given the proliferation of non-human identities in enterprise systems. (Sources: Schneier on Security, Cisco, Vectra, 1Password) <a href="https://app.daily.dev/feed-by-ids?id=fuL6WfKHt&amp;id=iW2dnVT4l&amp;id=nVdzT2QRW&amp;id=IVsgr78CO&amp;id=SulJvcnOu&amp;id=kgAIr1pL1&amp;id=puWFcrOO1&amp;id=zG1YpV8p9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft’s AI spending spree and quality control issues</strong>: Microsoft’s stock dropped following reports of $37.5 billion in AI-related capital expenditure, raising questions about whether AI demand will justify the massive spending. This aggressive AI push coincides with a troubling trend of increased emergency patches for Windows, suggesting quality control issues. While Microsoft is enhancing Azure PostgreSQL with AI features, the broader strategy of embedding Copilot everywhere feels forced, leading to user backlash and a review of AI integrations in Windows 11. (Sources: #0PRBbgRZc, #JtPRqmAkm, #H1aWwdchu, #0jt1kGbdr, #8b12pEovl, #mR53bImT0) <a href="https://app.daily.dev/feed-by-ids?id=isSX8DfZa&amp;id=0PRBbgRZc&amp;id=JtPRqmAkm&amp;id=H1aWwdchu&amp;id=0jt1kGbdr&amp;id=8b12pEovl&amp;id=mR53bImT0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Raspberry Pi prices jump due to AI memory demand</strong>: Raspberry Pi has increased prices for models with 2GB or more memory, with hikes ranging from $10 to $60. The primary culprit is the soaring cost of LPDDR4 memory, driven by insatiable demand from AI infrastructure. This marks a continuation of pricing adjustments, indicating ongoing supply chain pressures in the memory market, directly impacting hobbyists and developers relying on these popular single-board computers. <a href="https://app.daily.dev/posts/Rk4TtZc2A" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>OpenAI standardizes AI agent workflows</strong>: OpenAI released the Open Responses specification to standardize agentic AI workflows, aiming to eliminate API fragmentation and allow developers to switch between models without rewriting integration code. <a href="https://app.daily.dev/posts/yIvolp9mz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New open-source AI models released</strong>: Moonshot AI released Kimi K2.5, an open-source multimodal AI model with 1 trillion parameters, featuring native vision-language understanding and an Agent Swarm capability, while AWS added DeepSeek OCR, MiniMax M2.1, and Qwen3-VL-8B-Instruct to SageMaker JumpStart. <a href="https://app.daily.dev/feed-by-ids?id=45kpQHTQG&amp;id=CoNCaDnyA&amp;id=L2SMBoUpw&amp;id=qG0ZTfTpe&amp;id=qKtGa6cI1&amp;id=ucSoigGp2&amp;id=wEnEYbP9z&amp;id=yKJwfVHPS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>MongoDB Atlas launches AI embedding and reranking API</strong>: MongoDB Atlas introduced a public preview of its Embedding and Reranking API, integrating Voyage AI’s search models directly into the platform to simplify building Retrieval Augmented Generation (RAG) systems. <a href="https://app.daily.dev/posts/Ow0bpZEWc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Firefox 148 to introduce granular AI controls</strong>: Firefox 148 will feature a new AI Controls panel, allowing users to disable AI features at a granular level or block them entirely, following user backlash against intrusive AI functionalities. <a href="https://app.daily.dev/posts/K3PNUlsQj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub addresses low-quality and AI-generated contributions</strong>: GitHub is implementing repository-level controls and planning AI-powered triage tools to combat the surge of low-quality and AI-generated pull requests overwhelming open-source maintainers. <a href="https://app.daily.dev/posts/jwftmMKjS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New tools for AI agent development and infrastructure</strong>: Neuron AI (PHP framework) simplifies RAG agent building, Polyvia (Visual Knowledge Index) enables AI agents to reason across visual data, and Agentuity (cloud platform) supports long-running AI agents with native agent-to-agent communication. <a href="https://app.daily.dev/feed-by-ids?id=GJcC4La6R&amp;id=IvsJ5kU09&amp;id=Pd8Tj3t8V&amp;id=PxHnBAZAP&amp;id=nortfP3DR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux From Scratch drops SysVinit support</strong>: Linux From Scratch (LFS) will become systemd-only starting with version 13.0, citing overwhelming volunteer workload and increasing upstream dependencies on systemd from major desktop environments. <a href="https://app.daily.dev/posts/ohWaYbFxO" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI coding tools: the productivity paradox</strong>: Apparently, AI coding tools are supposed to make us all super-productive, but some experts are now claiming they actually <em>reduce</em> overall productivity by making complexity cheap and compounding organizational chaos. It’s almost like blindly generating code without understanding the underlying architecture or security implications isn’t a shortcut to maintainable software. Who knew? <a href="https://app.daily.dev/feed-by-ids?id=9HijPHSEH&amp;id=AIQanc1ui&amp;id=NEIMwAdHs&amp;id=k9rFokMvG&amp;id=q4q4xGaBF&amp;id=1ESLU5xBs&amp;id=Akml65ddg&amp;id=CUMsxmpoK&amp;id=GOFzDmvEo&amp;id=O9ssUnmJv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Manis AI: build an app in 8 minutes, deploy in one click</strong>: A new no-code platform, Manis AI, claims you can build and deploy a mobile app in 8 minutes using AI orchestration. They even demonstrated a ‘Bill Sniper’ app from a single prompt. Because, you know, the hard part of app development has always been the typing, not the actual design, testing, maintenance, or, you know, making something useful that people actually want. What could possibly go wrong? <a href="https://app.daily.dev/posts/mN5yBPuXJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare’s ‘production-grade’ matrix homeserver</strong>: Cloudflare proudly announced a ‘production-grade’ Matrix homeserver built on Workers, only for the Matrix co-founder to point out it lacked critical features like federation, authentication, and encryption. Turns out, it was mostly unreviewed AI-generated code with ‘TODO’ comments in critical sections. They later added a disclaimer, reclassifying it as a ‘proof of concept.’ It’s almost like you can’t just slap ‘AI’ on something and call it production-ready. <a href="https://app.daily.dev/feed-by-ids?id=46vOfTKm4&amp;id=IgyIzqLcs&amp;id=Iq2RG6GgT&amp;id=bREDMGnKC&amp;id=d6wB2Rm36" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
