<h2>Must know</h2>
<ul>
<li><strong>Google rolls out Gemini 3.1 Pro preview</strong>: Google has released Gemini 3.1 Pro in public preview, touting it as their latest agentic coding and state-of-the-art reasoning model. It boasts a 1 million token input context window and significant performance gains on benchmarks like ARC-AGI-2, scoring 77.1% compared to Gemini 3 Pro’s 31.1%. This release intensifies the AI model race, with Google positioning it as a cost-effective alternative to competitors like Anthropic’s Opus 4.6. <a href="https://app.daily.dev/posts/PjE6Xo1XG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents shift developer roles and introduce new risks</strong>: AI coding tools are transforming developer roles, with 93% of developers now using AI assistants, shifting focus from writing code to reviewing and orchestrating AI agents. While some report significant productivity gains, concerns are rising about AI-generated code introducing subtle production bugs and the ‘AI slop problem’ in open source, where low-quality AI-generated PRs burden maintainers. The Moonwell DeFi protocol even lost $1.78 million due to a misconfigured price oracle, with GitHub records indicating the code was co-authored by Claude Opus 4.6. <a href="https://app.daily.dev/feed-by-ids?id=037a5YIjx&amp;id=0Q86tGAco&amp;id=0tYkAozTe&amp;id=1v9gMZMqJ&amp;id=2LocixQWJ&amp;id=2WWQ65Wm6&amp;id=2Xamq3Lpe&amp;id=3Kdb5qCcA&amp;id=4CjaWv3pP&amp;id=6KUpHm9AN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>MySQL governance under scrutiny as community pushes for independent foundation</strong>: Over 248 database professionals, including engineers from Percona and MariaDB, have signed an open letter urging Oracle to transfer MySQL governance to an independent non-profit foundation. Citing declining commit activity, a shrinking contributor pool, and a lack of AI-era features, the community seeks to revitalize MySQL development and address growing fragmentation across its forks. This move highlights ongoing concerns about Oracle’s control over the popular open-source database. <a href="https://app.daily.dev/posts/GY7vcPtOu" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI agents can now control iOS and Android devices</strong>: Callstack’s open-source <code>agent-device</code> CLI tool allows AI agents to programmatically control iOS and Android devices, including simulators, emulators, and physical hardware, supporting actions like tapping, filling, scrolling, and inspecting accessibility trees. <a href="https://app.daily.dev/posts/Ekfc0OFB7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Red Hat Lightspeed uses AI for RHEL image creation</strong>: Red Hat Lightspeed Model Context Protocol (MCP) enables natural language-driven RHEL image creation and management, allowing users to converse with an LLM to build blueprints, apply security profiles, and push images to cloud providers. <a href="https://app.daily.dev/posts/CtJea7QHd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>BigQuery gets autonomous embedding generation</strong>: Google Cloud’s BigQuery now supports autonomous embedding generation, automatically maintaining embedding columns on tables and simplifying semantic search with a new <code>AI.SEARCH</code> function that integrates with Vertex AI embedding models. <a href="https://app.daily.dev/posts/BhFMZ3Oac" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vitest 3.x significantly outperforms Jest 30</strong>: Vitest 3.x shows dramatically faster performance than Jest 30, with 5.6x faster cold starts and 28x faster watch mode re-runs in monorepos, along with 57% lower peak memory usage, making it a strong contender for JavaScript testing. <a href="https://app.daily.dev/posts/T3G4VUzdk" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Node.js memory usage halved with <code>node-caged</code> Docker image</strong>: Cloudflare, Igalia, and the Node project collaborated on <code>node-caged</code>, a Docker image with V8 pointer compression that can halve Node.js memory usage in production, alongside npm v11.10.0 and BrowserPod 1.0 for in-browser Node.js sandboxing. <a href="https://app.daily.dev/posts/Mm7b9KCfF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>WinterTC standardizes JavaScript runtimes</strong>: WinterTC (Ecma TC55) is standardizing a unified API surface across JavaScript runtimes like Node, Deno, Bun, and Cloudflare Workers, aiming for true ‘write once, run anywhere’ isomorphic JavaScript by treating the browser as the baseline. <a href="https://app.daily.dev/posts/hSdXmiGXO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Tailscale workload identity federation now generally available</strong>: Tailscale’s workload identity federation is now generally available, allowing CI/CD pipelines and cloud workloads to authenticate using short-lived OIDC tokens instead of static API keys, enhancing security and simplifying credential management. <a href="https://app.daily.dev/posts/ekx9IIs1v" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Claude Pro integrates with PowerPoint, because that’s what we needed</strong>: Anthropic announced Claude Pro now integrates with Microsoft PowerPoint, allowing users to ‘pull context’ from daily tools directly into presentations. Because clearly, the biggest problem with presentations was the lack of AI-powered context pulling, not, you know, the presentations themselves. One source even claims Claude Pro is ‘getting bigger every week,’ which sounds suspiciously like a growth spurt in marketing fluff. <a href="https://app.daily.dev/posts/oLQ4V0bPw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Trajectory Explorer: the ‘first sane way’ to inspect AI agents</strong>: @benhylak announced Trajectory Explorer, a new tool for navigating AI agent traces, claiming it’s the ‘first sane way’ to inspect AI agent decisions. Because before this, we were all just wildly guessing what our AI agents were doing, apparently. The promotional language is strong with this one, suggesting a groundbreaking solution to a problem that may or may not have been that insane to begin with. <a href="https://app.daily.dev/posts/AzPQg0qJn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Link11 launches AI management dashboard for crawler traffic control</strong>: Link11, an IT security provider, launched an AI Management Dashboard to give companies ‘dedicated visibility and control’ over AI crawler traffic. Because apparently, AI crawlers are so special they need their own dashboard, distinct from all the other bots. It’s almost like they’re trying to sell us a solution to a problem they just invented, complete with ‘forensic drilldowns’ and ‘audit-ready event logs’ for your very important AI traffic. <a href="https://app.daily.dev/posts/zxo3wKZNk" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
