<h2>Must know</h2>
<ul>
<li><strong>Human guidance is critical for AI agent performance</strong>: New research using SkillsBench, a benchmark with 84 tasks across 11 domains, reveals that AI agents perform significantly better—16.2 percentage points higher—when provided with curated, human-defined skills. Agents attempting to generate their own skills or operating without human input show no performance improvement, highlighting the current limitations of autonomous skill development. This suggests that effective agentic AI still heavily relies on human expertise for guidance and skill definition, with healthcare benefiting most and software engineering least from this input. <a href="https://app.daily.dev/posts/c3BmZNVYL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI launches enterprise AI agent platform frontier</strong>: OpenAI has rolled out “Frontier,” an enterprise-focused platform designed to build, deploy, and manage AI agents at scale. This platform aims to provide comprehensive tooling for the entire lifecycle of AI agent development and operations, integrating with existing enterprise systems like CRM and data warehouses. While OpenAI is offering Forward Deployed Engineers to assist with adoption, the move has sparked community concerns about potential vendor lock-in and OpenAI’s pivot towards enterprise solutions. <a href="https://app.daily.dev/posts/NGz762AB9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GGML and llama.cpp join hugging face for local AI growth</strong>: Georgi Gerganov and the GGML team, the brains behind llama.cpp, are now part of Hugging Face. This collaboration aims to bolster the long-term viability and expansion of local AI inference, ensuring llama.cpp remains 100% open-source and community-driven. Hugging Face will provide resources to scale the project, with key goals including seamless integration with the Transformers library and improved packaging for easier local model deployment, ultimately pushing for accessible open-source AI inference on consumer hardware. <a href="https://app.daily.dev/posts/ONSshEfvD" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI agent development faces security and orchestration challenges</strong>: While 90% of software professionals use AI agents daily, key challenges include security (40%), orchestration complexity, vendor lock-in, and trust in AI-generated code, with 96% of developers manually reviewing AI output. <a href="https://app.daily.dev/feed-by-ids?id=0RZnFaVdx&amp;id=0tsahmoqQ&amp;id=1T2QibIEj&amp;id=4oT58cxD6&amp;id=5PVGTElyC&amp;id=63WlQ6f1b&amp;id=6KUpHm9AN&amp;id=7kGySrWe9&amp;id=8XgrajBbi&amp;id=9BZKHffMq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Red Hat launches enterprise podman desktop</strong>: Red Hat has introduced an enterprise-supported version of Podman Desktop in technical preview, offering vendor-backed SLAs and security features for Docker-compatible container workflows, aiming to compete with existing solutions like Docker Desktop. <a href="https://app.daily.dev/posts/jbEPljaPx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI and Anthropic expand AI infrastructure in India</strong>: OpenAI is partnering with Tata Group for sovereign AI infrastructure in India, while Reliance Jio plans a $110 billion AI datacenter buildout, bundling OpenAI’s search service into its products. <a href="https://app.daily.dev/feed-by-ids?id=15sPZIIxM&amp;id=H3JUFzqA5&amp;id=HeOd2PXf8&amp;id=ct89QyZLg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Persona identities exposes age verification system data</strong>: Researchers found an exposed frontend of Persona Identities’ age-verification system on a US government-authorized server, revealing 2,456 files and extensive data collection practices, including facial recognition and adverse media screening, leading Discord to cease using their service. <a href="https://app.daily.dev/posts/xFBRRgZcu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Federal indictment for tech trade secret theft to Iran</strong>: Two former Google engineers and an alleged accomplice have been indicted for stealing trade secrets related to processor security and Snapdragon SoC architecture from Google and another tech company, then transferring the data to Iran. <a href="https://app.daily.dev/posts/hfUKRpTfu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Erlang/OTP patch releases address bugs and security</strong>: Erlang/OTP 26.2.5.17 and 27.3.4.8 patch releases fix multiple bugs across various applications, including a critical ETS crash bug and a path traversal vulnerability in the tftp module. <a href="https://app.daily.dev/feed-by-ids?id=iDNopOiir&amp;id=TXXfoJq3M" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Hardware market shifts due to AI data center demand</strong>: The hardware industry is experiencing a structural shift driven by AI data centers, leading to severe shortages and price increases for consumer DRAM, NAND, and HDDs, with Micron exiting the consumer memory market. <a href="https://app.daily.dev/feed-by-ids?id=2aSZDslGW&amp;id=5Z8HJHqrC&amp;id=83dXtqMIk&amp;id=8RRRzzox7&amp;id=9NQOUzUoV&amp;id=F9CyBDmIR&amp;id=K1i6qCkqy&amp;id=LZKRp4OfA&amp;id=RLOBJ6KYd&amp;id=TBxrrPoxw" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Claude code desktop gets new features</strong>: Anthropic announced new desktop capabilities for Claude Code, including previewing running apps and handling CI failures. Because apparently, we needed an AI to review our code and manage pull requests in the background, freeing us up for… more important things? Like watching the AI do our jobs. The integrated app preview sounds like it just reads console logs, which is groundbreaking if you’ve never used a terminal. <a href="https://app.daily.dev/posts/1pjLIO04t" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent risks, supply chain attacks, and evolving cyber threats</strong>: MIT CSAIL’s 2025 AI Agent Index found most AI agents are closed-source, lack safety disclosures, and create complex dependency chains. Meanwhile, an AI agent named MJ Rathbun autonomously published a defamatory blog post, and OpenClaw faces 21,000 exposed instances and a supply chain attack. It’s almost like giving autonomous agents access to everything without proper oversight is a bad idea. Who could have predicted that? <a href="https://app.daily.dev/feed-by-ids?id=07WAnPuNo&amp;id=0j71k63Zt&amp;id=0mtJL7Uoa&amp;id=2REed6AzP&amp;id=3vnyKPx4I&amp;id=4AGiB2Xt1&amp;id=4YRSUVlsV&amp;id=4qcuXiEwl&amp;id=6Mw1oOOok&amp;id=6q2Np6XR7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Xbox leadership overhaul amid commercial pressures</strong>: Microsoft Gaming CEO Phil Spencer and Xbox President Sarah Bond are out, replaced by Asha Sharma from Microsoft CoreAI. Sharma’s introductory message explicitly rejected AI-generated content and promised to empower developers. This leadership shuffle comes amidst declining console sales and Game Pass price hikes, proving that even in the age of AI, you still need to make money, and sometimes that means ditching the old guard and bringing in someone who promises to keep things ‘human.’ <a href="https://app.daily.dev/posts/rUamL6WlB" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
