<h2>Must know</h2>
<ul>
<li><strong>Nvidia-OpenAI investment uncertainty and stock drop</strong>: Nvidia’s stock dipped following reports of uncertainty around its planned $100 billion investment in OpenAI. While CEO Jensen Huang initially criticized OpenAI’s strategy, he later confirmed a significant investment, though not exceeding $100 billion. Analysts suggest this uncertainty stems from concerns about circular financing in AI and Nvidia’s strategic balancing act to support OpenAI without funding potential competitors. <a href="https://app.daily.dev/posts/isSX8DfZa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Russia-linked APT28 exploits Microsoft Office zero-day</strong>: The Russia-linked APT28 group is actively exploiting a Microsoft Office zero-day vulnerability (CVE-2026-21509) to target Ukrainian government agencies and EU organizations. The attacks use weaponized documents to establish WebDAV connections, download malware, and deploy the COVENANT framework. CERT-UA warns that slow update cycles will likely lead to increased attacks, highlighting a persistent state-linked cyber threat. <a href="https://app.daily.dev/posts/7E37fbDrv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Raspberry Pi increases prices due to memory demand</strong>: Raspberry Pi has increased prices for models with 2GB or more memory, citing doubled LPDDR4 memory costs driven by AI infrastructure demand. This affects models like the Raspberry Pi 4, 5, Compute Module 4/5, and Pi 500 series, with increases ranging from $10 to $60. Models with 1GB RAM and older LPDDR2 memory are unaffected, but this marks a continuing trend of supply chain pressures in the memory market. <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>Moonshot AI releases Kimi K2.5 multimodal model</strong>: Moonshot AI launched Kimi K2.5, an open-source multimodal AI model with 1 trillion parameters, featuring native vision-language understanding, a 256K context length, and an Agent Swarm capability for task decomposition. <a href="https://app.daily.dev/feed-by-ids?id=45kpQHTQG&amp;id=CoNCaDnyA&amp;id=L2SMBoUpw&amp;id=qG0ZTfTpe&amp;id=qKtGa6cI1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI releases open responses specification for agentic AI workflows</strong>: OpenAI introduced the Open Responses specification to standardize agentic AI workflows, aiming to eliminate API fragmentation across LLM providers and enable developers to switch models without rewriting integration code. <a href="https://app.daily.dev/posts/yIvolp9mz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>MongoDB Atlas launches embedding and reranking API</strong>: MongoDB Atlas released 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>GitHub addresses low-quality contributions and AI-generated code</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>Microsoft Windows emergency patch frequency increases</strong>: Microsoft continues to release frequent out-of-band emergency patches for Windows, a trend coinciding with workforce reductions and increased AI-generated code, raising concerns about quality control. <a href="https://app.daily.dev/posts/H1aWwdchu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Pulsar RAT Windows malware campaign discovered</strong>: Security researchers uncovered a sophisticated Windows malware campaign using Pulsar RAT, a modular remote access trojan that operates primarily in memory to evade detection and exfiltrate data via Discord and Telegram. <a href="https://app.daily.dev/posts/5bhTgZQlt" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rust Coreutils project progresses towards GNU compatibility</strong>: The uutils project, a Rust implementation of GNU coreutils, released version 0.6 with improved compatibility and reduced unsafe code, actively working towards 100% GNU utility compatibility. <a href="https://app.daily.dev/feed-by-ids?id=cuWdYm0D1&amp;id=iBkEhbYAD" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI agent knowledge sharing and framework development</strong>: Apparently, AI agents are now so advanced they need their own social networks and knowledge commons to ‘learn collectively’ and ‘avoid redundant work.’ Because, you know, the real problem with AI isn’t hallucinating or being biased, it’s that they’re not sharing enough cat memes with each other. And of course, we need yet another ‘expert’ ranking of AI agent frameworks, because clearly, the last dozen weren’t enough to solve all our problems. <a href="https://app.daily.dev/feed-by-ids?id=Jq25xqmJD&amp;id=YjbPdZkNb&amp;id=j0iRlHvmi" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Manis AI platform enables rapid mobile app development</strong>: Oh look, another ‘build and launch a mobile app in 10 minutes with no code’ platform. Manis AI promises to turn your single prompt into a Google Play-ready app using ‘AI orchestration’ and ‘specialized sub-agents.’ Because what the world really needs is more hastily built, potentially buggy apps, all thanks to an orchestrator AI that probably just delegates to a bunch of glorified script runners. What could possibly go wrong? <a href="https://app.daily.dev/posts/mN5yBPuXJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI manipulated content persists despite exposure</strong>: Turns out, telling people AI-generated content is fake doesn’t actually make them stop believing it. The article ‘What we’ve been getting wrong about AI’s truth crisis’ suggests our ‘truth crisis’ isn’t about initial deception, but the ‘lasting emotional and cognitive impact’ of manipulated content. So, all those fancy transparency tools like Adobe’s Content Authenticity Initiative? Just a band-aid on a bullet wound, apparently. Good to know our brains are so easily swayed, even when we know better. <a href="https://app.daily.dev/posts/savTmPDZC" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
