<h2>Must know</h2>
<ul>
<li><strong>Hugging Face and llama.cpp team up for local AI</strong>: Georgi Gerganov and the llama.cpp team are joining Hugging Face to ensure the long-term sustainability and growth of local AI inference. The collaboration aims for seamless integration between llama.cpp and the Transformers library, with llama.cpp remaining 100% open-source and community-driven. This move is intended to make open-source AI inference accessible on consumer hardware worldwide, positioning local inference as a competitive alternative to cloud solutions. <a href="https://app.daily.dev/posts/ONSshEfvD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI industry super PACs clash over regulation</strong>: Two rival pro-AI Super PACs are duking it out in a New York congressional race, highlighting a deep split within the AI industry over regulation. One PAC, backed by Andreessen Horowitz and OpenAI’s Greg Brockman, is attacking a candidate who sponsored a bill requiring AI developers to disclose safety protocols. The opposing PAC, funded by Anthropic, supports the candidate, showcasing a battle between minimal regulation and transparency standards. <a href="https://app.daily.dev/posts/cobwYTIA0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>EFF updates open-source policy for AI-assisted code</strong>: The Electronic Frontier Foundation (EFF) has updated its open-source contribution policy to permit AI-generated code, acknowledging that banning it is impractical. The new policy emphasizes accountability: contributors must understand and be able to explain any AI-generated code they submit, and all comments and documentation must be human-authored. This approach aims to mitigate concerns about AI hallucinations and low-quality contributions without resorting to ineffective AI detection. <a href="https://app.daily.dev/posts/W3yzAhjLd" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Claude Code desktop gets major update</strong>: Anthropic released a significant update to Claude Code for desktop, adding an integrated app preview, a ‘Review code’ button for inline comments and auto-fixes, GitHub PR monitoring with optional auto-fix for CI failures, and cross-device session continuity. <a href="https://app.daily.dev/posts/1pjLIO04t" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Claude Code CLI adds git worktree support</strong>: Claude Code CLI now includes built-in git worktree support, enabling multiple AI agents to run in parallel with isolated worktrees, a feature previously exclusive to the desktop app. <a href="https://app.daily.dev/posts/X4UvQOpsB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Red Hat launches enterprise Podman Desktop</strong>: Red Hat introduced an enterprise-supported build of Podman Desktop in technical preview, offering vendor-backed SLAs, security fixes, and policy enforcement for corporate developers needing Docker-compatible container workflows, with tight integration with OpenShift and RHEL. <a href="https://app.daily.dev/posts/jbEPljaPx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Laravel 13 to require PHP 8.3 and add attributes</strong>: Laravel 13, due in Q1 2026, will require PHP 8.3 and introduce first-class PHP 8 Attributes for configuring Eloquent models, queue jobs, and console commands, alongside a new <code>Cache::touch()</code> method. <a href="https://app.daily.dev/posts/noeEN07CU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Java’s 2026 roadmap includes Valhalla and Loom</strong>: Java’s 2026 development roadmap includes Project Valhalla previewing value types, Project Babylon incubating code reflection, and Project Loom finalizing the structured concurrency API, alongside AOT compilation from Project Leyden. <a href="https://app.daily.dev/feed-by-ids?id=sSQwkMVM8&amp;id=fwzcHZTXf" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ruby pushes for AI development with new tools</strong>: Ruby is positioning itself as a strong contender for AI application development, with new libraries like SchnellMCP allowing Ruby scripts to become MCP servers, and Pi for Excel embedding AI agents that connect to MCP servers. <a href="https://app.daily.dev/feed-by-ids?id=5tjEzlHaH&amp;id=J2GdE2JFq&amp;id=aNUYCuspS&amp;id=dWNi83vQW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>VS Code integrates browser and AI agent testing</strong>: VS Code now features an integrated browser with full DevTools and AI-powered features, allowing developers to inject HTML/CSS into Copilot chat and enabling AI agents to autonomously interact with the browser via Playwright/CDP for testing and bug finding. <a href="https://app.daily.dev/posts/tN0TdAcUh" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>OpenAI launches ‘Frontier’ for enterprise AI agents</strong>: OpenAI just dropped ‘Frontier,’ an enterprise platform for building and managing AI agents at scale. They’re promising to integrate with all your existing systems and even send ‘Forward Deployed Engineers’ to hold your hand. Because nothing says ‘open’ like a proprietary platform with built-in vendor lock-in, right? The community is already calling it out, comparing it to Anthropic’s Claude Cowork. Original reports by Sergio De Simone. <a href="https://app.daily.dev/posts/NGz762AB9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Elon Musk delays Grok release over Baldur’s Gate performance</strong>: Apparently, Elon Musk delayed a Grok model release because he wasn’t happy with its answers about Baldur’s Gate, even pulling high-level engineers to fix the issue. Business Insider reported on this critical national security concern. TechCrunch then ran an ‘informal benchmark’ (BaldurBench, naturally) and found Grok is ‘comparable’ to other models. So, all that drama for ‘comparable’ performance. Peak efficiency, folks. <a href="https://app.daily.dev/posts/vj4IIdP4D" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents need human guidance to perform</strong>: New research using ‘SkillsBench’ (because we need more benchmarks) found that AI agents perform significantly better with ‘curated, human-provided skills.’ Without human input, they show ‘no performance improvement.’ So, after all the hype about autonomous agents, it turns out they still need us to tell them what to do. Who could have possibly seen that coming? <a href="https://app.daily.dev/posts/c3BmZNVYL" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
