<h2>Must know</h2>
<ul>
<li><strong>OpenAI launches enterprise AI agent platform</strong>: OpenAI has rolled out “Frontier,” an enterprise-focused platform designed to help companies build, deploy, and manage AI agents at scale. This platform aims to integrate with existing systems like CRMs and data warehouses, offering tools for the full lifecycle of AI agent development. However, the move has sparked community concerns about potential vendor lock-in, with some comparing it to Anthropic’s Claude Cowork, as reported by Sergio De Simone. <a href="https://app.daily.dev/posts/NGz762AB9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google engineers indicted for stealing trade secrets to Iran</strong>: A federal grand jury has indicted two former Google engineers, sisters Samaneh and Soroor Ghandali, along with an alleged accomplice, Mohammadjavad Khosravi, for stealing trade secrets related to processor security and Snapdragon SoC architecture. Prosecutors allege the data was transferred to Iran using methods like Telegram and photographing screens to evade detection. Google’s security monitoring flagged the activity, leading to the charges, which carry significant prison sentences. <a href="https://app.daily.dev/posts/hfUKRpTfu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Human guidance is critical for AI agent performance</strong>: New research using the SkillsBench benchmark reveals that AI agents perform significantly better—an average of 16.2 percentage points higher—when provided with curated, human-designed skills. Agents that generate their own skills or receive no human input show no performance improvement, highlighting the current limitations of AI in autonomous skill development and underscoring the essential role of human expertise in effective agentic AI. <a href="https://app.daily.dev/posts/c3BmZNVYL" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>GGML and llama.cpp join Hugging Face</strong>: Georgi Gerganov and the GGML team, creators of llama.cpp, are joining Hugging Face to ensure the long-term sustainability and growth of local AI inference, with llama.cpp remaining 100% open-source and community-driven. <a href="https://app.daily.dev/posts/ONSshEfvD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Red Hat launches enterprise Podman Desktop</strong>: Red Hat has released an enterprise-supported build of Podman Desktop in technical preview, offering vendor-backed SLAs and security features for corporate developers needing Docker-compatible container workflows, integrating with OpenShift and RHEL. <a href="https://app.daily.dev/posts/jbEPljaPx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>G42 and Cerebras partner for 8 exaFLOPS AI supercomputer in India</strong>: UAE-based G42 and US chipmaker Cerebras are deploying an 8 exaFLOPS AI supercomputer in India, utilizing approximately 64 Cerebras WSE-3 wafer-scale accelerators to serve educational institutions, government, and startups. <a href="https://app.daily.dev/posts/ASzeKt856" 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 departing, with Asha Sharma taking over as the new Xbox boss, amidst declining console sales and controversies over Game Pass pricing. <a href="https://app.daily.dev/posts/rUamL6WlB" 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>Integrated browser in VS Code gets AI features</strong>: VS Code now features an integrated Chromium-based browser with full DevTools and AI-powered capabilities, allowing users to inject HTML, CSS, and screenshots into Copilot chat, and enabling AI agents to autonomously interact with browser tabs for testing. <a href="https://app.daily.dev/posts/tN0TdAcUh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ATM jackpotting attacks surge, netting $20 million</strong>: The FBI warns of a significant increase in ATM jackpotting attacks in 2025, with over 700 incidents resulting in at least $20 million stolen by combining physical access with Ploutus malware to force cash dispensers to eject money without debiting customer accounts. <a href="https://app.daily.dev/posts/LFNwj8vdP" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI agents are transforming software development, or so they say</strong>: According to Docker’s State of Agentic AI report, 90% of software professionals use AI agents daily, and 60% run them in production. Yet, 96% don’t trust AI-generated code without manual review, spending 24% of their week auditing AI output. So, we’re just shifting toil, not eliminating it, and calling it ‘transformation.’ <a href="https://app.daily.dev/feed-by-ids?id=0RZnFaVdx&amp;id=ZMDQf4WZI&amp;id=M3VIxPqTq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent security: the new ‘trust us, we’re selling you more tools’ narrative</strong>: MIT CSAIL found 23 of 30 AI agents are closed-source, lack safety disclosures, and create complex dependency chains with no single accountable entity. But don’t worry, Lasso Security has ‘Intent Deputy’ to monitor for deviations, and Adversa AI has ‘SecureClaw’ to audit agents. Because the solution to opaque, insecure AI is always more opaque, proprietary security tools, right? <a href="https://app.daily.dev/feed-by-ids?id=07WAnPuNo&amp;id=833aZp617&amp;id=ITcMT7cby" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The ‘AI-native mobile development’ hype train is leaving the station</strong>: Apparently, by 2026, we’ll all be doing ‘AI-native mobile development’ with on-device inference, NPU acceleration, and ‘Predictive UX.’ Because nothing says ‘innovation’ like taking existing concepts, slapping ‘AI’ on them, and pretending it’s a revolutionary new paradigm. Just remember to bring your own NPU, because your phone probably doesn’t have one yet. <a href="https://app.daily.dev/posts/QWkRl6v8e" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
