<h2>Must know</h2>
<ul>
<li><strong>Google rolls out Gemini 3.1 Pro preview</strong>: Google just dropped Gemini 3.1 Pro in public preview, their latest agentic coding and reasoning model. It scored 77.1% on ARC-AGI-2, a huge jump from 31.1% for Gemini 3 Pro, and leads most coding benchmarks. This model supports multimodal inputs, a 1 million token context window, and new features like configurable thinking levels, all while being cheaper than Anthropic’s Opus 4.6. This is a significant move in the AI model race, with early adopters already reporting gains in complex problem-solving and code generation. <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, lack of AI-era features, and an opaque roadmap as major concerns. The community wants to revitalize MySQL development and ensure its future relevance, but analysts are skeptical Oracle will cede control, especially over trademarks and the release pipeline. <a href="https://app.daily.dev/posts/GY7vcPtOu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>FBI warns of surge in ATM jackpotting attacks</strong>: The FBI issued a security alert after ATM jackpotting attacks surged in 2025, with over 700 incidents stealing more than $20 million. Attackers gain physical access, install malware like Ploutus, and force cash dispensing by exploiting the XFS API, often on Windows-based ATMs. The FBI urges financial institutions to monitor for specific digital and physical indicators of compromise, highlighting the lucrative and hard-to-detect nature of these attacks. <a href="https://app.daily.dev/posts/LFNwj8vdP" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI agent adoption and impact in software development</strong>: AI coding tools are shifting developer roles towards reviewing and orchestrating AI agents, with 93% of developers using AI assistants, though productivity gains have plateaued at around 10% and AI-authored code accounts for 26.9% of production code. <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>BigQuery now supports autonomous embedding generation</strong>: Google Cloud’s BigQuery now offers autonomous embedding generation, automatically maintaining embedding columns and simplifying semantic search with a new <code>AI.SEARCH</code> function, integrating with Vertex AI embedding models. <a href="https://app.daily.dev/posts/BhFMZ3Oac" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Apache Kafka 4.2.0 released with share groups and streams improvements</strong>: Apache Kafka 4.2.0 is out, bringing production-ready Share Groups (Kafka Queues), GA status for Kafka Streams server-side rebalance protocol, and improved observability and security features. <a href="https://app.daily.dev/posts/0DcTaIDm9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google Cloud Spanner adds native CQL API support</strong>: Google Cloud Spanner now generally supports native CQL (Cassandra Query Language) API, allowing Cassandra applications to migrate with minimal code changes while gaining Spanner’s global ACID transactions and 99.999% availability. <a href="https://app.daily.dev/posts/CKaLROkeW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New AWS Aurora DSQL connectors simplify IAM authentication</strong>: AWS released new Aurora DSQL connectors for Go, Python, and Node.js that automate IAM authentication, eliminating static credentials and enhancing security by automatically generating IAM tokens for connections. <a href="https://app.daily.dev/posts/tqOM2TAOg" 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 cold starts (5.6x) and watch mode re-runs (28x) compared to Jest 30 in monorepo benchmarks, along with 57% less 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>Docker hardened images for HIPAA compliance</strong>: Medplum, an open-source EHR platform, migrated to Docker Hardened Images (DHI) with minimal code changes, significantly reducing CVEs and simplifying compliance for HIPAA and SOC 2 audits by offloading OS-level hardening. <a href="https://app.daily.dev/posts/tcR1V1vMG" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI agent governance, bias, and economic impact concerns</strong>: MIT CSAIL found 23 of 30 AI agents are closed-source, with only 4 of 13 frontier-autonomy agents disclosing safety evaluations. Meanwhile, Microsoft used an AI-mangled, plagiarized Git workflow diagram, and MIT researchers found LLMs give less accurate responses to users with lower education. It’s almost like these AI companies are more interested in shipping something, anything, than ensuring it’s actually safe or fair. Who needs standards when you have hype? <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</strong>: @benhylak announced ‘Trajectory Explorer,’ a new tool for ‘navigating AI agent traces’ and providing ‘visibility into every decision an AI agent makes.’ They even called it ‘the first sane way’ to inspect AI agent decisions. Because apparently, before this, we were all just blindly trusting our AI overlords without any sane way to see what they were doing. Good to know someone finally brought sanity to the AI agent debugging process. <a href="https://app.daily.dev/posts/AzPQg0qJn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Claude Pro integrates with Microsoft PowerPoint</strong>: Claude AI now integrates with Microsoft PowerPoint for Claude Pro users, allowing them to pull context from daily tools directly into presentations. The announcements are short, promotional, and claim Claude Pro is ‘getting bigger every week.’ Because nothing says ‘groundbreaking AI’ like helping you make slightly less terrible PowerPoint slides. Truly, the future is now. <a href="https://app.daily.dev/posts/oLQ4V0bPw" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
