<h2>Must know</h2>
<ul>
<li><strong>AI agents are exposing cloud infrastructure weaknesses</strong>: Autonomous AI systems are revealing critical flaws in current cloud infrastructure, particularly in networking segmentation, identity-based security, and cost controls. These agents operate at machine speed, highlighting the need for fine-grained network policies, strict identity governance with short-lived credentials, and real-time cost monitoring with automated guardrails, according to InfoWorld. Organizations must treat agentic AI as a fundamental operational shift, not just an incremental feature, to prevent outages and runaway expenses. <a href="https://app.daily.dev/posts/1w09x1vo7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Go developers are happy, but not with AI coding tools</strong>: A recent Go developer survey found 91% satisfaction with the language, primarily for CLI tools and API services. However, only 13% are ‘very satisfied’ with AI coding tools, citing non-functional code generation (53%) and poor code quality (30%) as main complaints. Developers value AI for boilerplate, unit tests, and autocompletion, but the overall sentiment, as reported in Golang Weekly, suggests AI isn’t quite hitting the mark for Go’s pragmatic community. <a href="https://app.daily.dev/posts/yyZ0pKuL1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft gave BitLocker keys to the FBI</strong>: In a first publicly known instance, Microsoft provided BitLocker recovery keys to the FBI to decrypt three laptops during a fraud investigation. This happened because users, by default, store these keys in Microsoft’s cloud. Security experts warn this practice increases privacy risks and potential hacker access if Microsoft’s cloud is compromised. This incident highlights the trade-off between data recoverability and privacy, impacting BitLocker’s suitability for high-privacy needs. <a href="https://app.daily.dev/posts/6Aq315WBk" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Vercel optimizes deployments with smarter function caching</strong>: Vercel now skips function uploads when code hasn’t changed, reducing build times by 400ms-5s by injecting deployment-specific environment variables at runtime instead of build time, with Next.js support coming soon. <a href="https://app.daily.dev/posts/sygra8Fax" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NetBox Labs releases AI coding agent best practices</strong>: NetBox Labs published a repository of best practices for AI coding agents like Claude Code and Cursor, offering guidance on authentication, API usage, and integration patterns to help engineers build better NetBox integrations. <a href="https://app.daily.dev/posts/oJ6EYnP2o" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks Delta Sharing now supports Apache Iceberg</strong>: Databricks has integrated Apache Iceberg into Delta Sharing, allowing users to share Delta tables with Iceberg-compatible clients like Snowflake and Trino, eliminating the need for manual data exports and replication. <a href="https://app.daily.dev/posts/hqDVaLN8F" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Python ecosystem gets performance upgrades and new tools</strong>: Python is seeing a new JIT compiler for performance gains, Pandas version 3, new GUI tools for SQLite, and Zed, a Rust-based text editor, indicating a focus on efficiency and tooling diversification. <a href="https://app.daily.dev/posts/0TjKgLuas" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Spring AI 2.0.0-M2 brings null safety and new integrations</strong>: The latest Spring AI milestone introduces null safety across core APIs using JSpecify, updates the Model Context Protocol SDK, and adds new vector store integrations with Amazon S3, Infinispan, and Amazon Bedrock Knowledge Base. <a href="https://app.daily.dev/posts/7qw3Pandm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ESLint v10.0.0-rc.1 released for testing</strong>: ESLint v10.0.0-rc.1 is out, featuring built-in TypeScript type definitions for Espree and ESLint Scope, fixes to the no-shadow rule, and RuleTester improvements, with a migration guide available for users. <a href="https://app.daily.dev/posts/APrgmrJOW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Tailscale updates for security and integration</strong>: Tailscale v1.92.5 introduces workload identity federation API integration, changes default state file encryption for Windows and Linux clients, and improves Kubernetes operator behavior for certificate renewal. <a href="https://app.daily.dev/posts/Ktk2P1Mk4" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI coding agents: the new typing monkeys?</strong>: Anthropic CEO Dario Amodei boldly predicted AI could handle ‘most or all’ software engineering tasks in 6-12 months, claiming their engineers just ‘edit AI-generated code.’ Meanwhile, Zed’s Blog and other practitioners are out here saying LLMs automate typing, not thinking, and you still need to rigorously review every line. So, either Anthropic’s got some magic beans, or the rest of us are still stuck doing the actual brain work. <a href="https://app.daily.dev/feed-by-ids?id=0fyNmKnQE&amp;id=AnhMtsEYf&amp;id=DczaKkmGw&amp;id=KfJ8s7nDm&amp;id=NagkQXPrh&amp;id=dSFznJ0W1&amp;id=zJhxrzVIY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Nvidia’s copper conundrum: a simple math error, or a sign of the times?</strong>: Nvidia, the AI darling, published a paper claiming a gigawatt-scale data center would need 500,000 tons of copper. Turns out, it was a unit conversion error, and the real number is closer to 200 metric tons. The article, with a delightful touch of sarcasm, compares this to the Gimli Glider and a lost NASA Mars probe, suggesting a widespread ‘innumeracy’ in tech. Maybe they should ask an AI to double-check their math next time? <a href="https://app.daily.dev/posts/kgMlKXNFr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>TikTok’s new privacy policy: more data, more ‘US ownership’ fun</strong>: TikTok, now under ‘US ownership,’ has updated its privacy policy to collect precise GPS location, track all AI user interactions, and expand ad network data sharing. Users were, of course, required to accept these changes to continue using the app. It’s almost as if ‘US ownership’ means ‘more data for us,’ and privacy is just a suggestion, not a right. But hey, at least it’s not China getting all that sweet, sweet data, right? <a href="https://app.daily.dev/posts/BFPMM0rX7" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
