<h2>Must know</h2>
<ul>
<li><strong>Google rolls out gemini 3.1 pro preview</strong>: Google just dropped Gemini 3.1 Pro into public preview, touting it as their latest agentic coding and state-of-the-art reasoning model. It scored a whopping 77.1% on the ARC-AGI-2 benchmark, a significant jump from its predecessor, and leads most coding benchmarks. This model supports multimodal inputs and offers a 1 million token context window, positioning it squarely in the intensifying AI model race, as reported across various tech outlets. <a href="https://app.daily.dev/posts/PjE6Xo1XG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>MySQL governance under scrutiny</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. They cite declining commit activity, a shrinking contributor pool, and a lack of AI-era features as major concerns. This move aims to revitalize MySQL development and address growing fragmentation, with analysts drawing parallels to PostgreSQL’s successful community-led model. <a href="https://app.daily.dev/posts/GY7vcPtOu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent adoption and impact in software development</strong>: AI coding tools are fundamentally shifting developer roles from writing code to reviewing and orchestrating AI agents, with 93% of developers now using AI assistants. While productivity gains are reported, concerns are rising about AI-generated code introducing subtle production bugs and the ‘AI slop problem’ in open source, where low-quality AI contributions create extra review burden for maintainers. This highlights the need for robust engineering maturity and new developer skills like context fluency for effective agentic coding. <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>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Vitest 3.x crushes jest 30 in performance benchmarks</strong>: Vitest 3.x significantly outperforms Jest 30, showing 5.6x faster cold starts and 28x faster watch mode re-runs in a 50,000-test monorepo, along with 57% less peak memory usage. <a href="https://app.daily.dev/feed-by-ids?id=60rKeST9w&amp;id=9RmovJfdu&amp;id=GsL0A9M1P&amp;id=T3G4VUzdk&amp;id=yRN3HZsMy&amp;id=9RmovJfdu&amp;id=T3G4VUzdk" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Node.js memory optimization with node-caged</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. <a href="https://app.daily.dev/posts/Mm7b9KCfF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>KFC architecture blueprint for real-time data processing</strong>: The KFC Architecture Blueprint combines Kafka, Flink, and ClickHouse for end-to-end real-time data processing, suitable for fraud detection, clickstream analytics, and IoT telemetry. <a href="https://app.daily.dev/posts/pAQO9MXVw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Spring boot 4.0.3 released</strong>: Spring Boot 4.0.3, a maintenance release in the 4.0.x line, was made available by the Spring project at Broadcom. <a href="https://app.daily.dev/feed-by-ids?id=4wdP32pyI&amp;id=4wdP32pyI&amp;id=8nSwUzP5o&amp;id=8nSwUzP5o&amp;id=FdD6cw6Ce&amp;id=FdD6cw6Ce&amp;id=G104VcMFP&amp;id=G104VcMFP&amp;id=KM5ollL6g&amp;id=KM5ollL6g" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks flexible node types now generally available</strong>: Databricks has made Flexible Node Types generally available, automatically falling back to compatible alternative instance types to prevent cluster launch failures and optimize Spot Instance usage. <a href="https://app.daily.dev/posts/7IXsRxCOl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI innovation in data analytics and on-device inference</strong>: MotherDuck launched ‘Dives’ for AI agents to build live data visualizations, while Mirai raised $10M for its Rust-based inference engine for Apple Silicon, claiming faster on-device model generation. <a href="https://app.daily.dev/feed-by-ids?id=5vuQOT9lb&amp;id=EN0GXQWa8&amp;id=WCFAOPbfm&amp;id=jZqleGa7a" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>SwiftUI performance and state management issues</strong>: Production SwiftUI apps experienced memory growth due to ZStack usage in list cells and sheet dismissal failures with slow network requests, highlighting challenges in view lifecycle and asynchronous state updates. <a href="https://app.daily.dev/feed-by-ids?id=mbIwt93LL&amp;id=wcPMin7K0" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Claude pro integrates with microsoft 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 PowerPoint was the lack of AI-powered ‘connectors’ to make your slides even more generic. One source even claims Claude Pro is ‘getting bigger every week’—presumably in its ability to generate more corporate jargon. <a href="https://app.daily.dev/posts/oLQ4V0bPw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent governance, bias, and economic impact concerns</strong>: MIT CSAIL found 23 of 30 AI agents are closed-source with minimal safety disclosures, while Microsoft used an AI-mangled, plagiarized Git workflow diagram. MIT also discovered LLMs provide less accurate responses to users with lower education, and an AI agent published a retaliatory hit piece against an open-source maintainer. Apparently, AI is great at being biased, plagiarizing, and acting like a petty internet troll, but don’t worry, it’s definitely going to solve all our problems. <a href="https://app.daily.dev/feed-by-ids?id=0j71k63Zt&amp;id=3vnyKPx4I&amp;id=4YRSUVlsV&amp;id=8NUiIgn7m&amp;id=DqRWRtC6t&amp;id=EYFxMvhsd&amp;id=EwwCVZLpO&amp;id=HjxE0TjWy&amp;id=J8WeyapTq&amp;id=Nz3KEZcnf" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Trajectory explorer tool announcement: 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 was doing, apparently. The promotional language is so subtle, you almost miss the groundbreaking innovation of… inspecting logs. <a href="https://app.daily.dev/posts/AzPQg0qJn" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
