<h2>Must know</h2>
<ul>
<li><strong>AI agents are overwhelming open source maintainers</strong>: Open-source maintainers are drowning in low-quality, AI-generated pull requests, leading projects like Godot, cURL, and Ghostty to consider closing off external contributions. This ‘slop’ is eroding the collaborative spirit of open source, as AI agents, often misconfigured, generate nonsensical changes that waste human review time. GitHub is reportedly working on mitigations, but the core issue highlights a growing tension between AI-driven code generation and the human effort required for quality control. <a href="https://app.daily.dev/feed-by-ids?id=JAZbbup3r&amp;id=UT5jXEM9J&amp;id=BL3FTkDC5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft copilot faces antitrust scrutiny and security flaws</strong>: Microsoft’s aggressive integration of Copilot into Windows 11 and Microsoft 365 is drawing antitrust investigations from the FTC, which views it as potential illegal bundling. Simultaneously, a confirmed bug allows Microsoft 365 Copilot Chat to bypass Data Loss Prevention (DLP) policies, exposing confidential emails. This highlights a critical security vulnerability where AI tools, operating within trusted environments, can inadvertently act against established security controls, raising concerns about the ‘confused deputy problem’ in AI-powered systems. <a href="https://app.daily.dev/feed-by-ids?id=29INM1IxL&amp;id=GzISnirqE&amp;id=nnFweCk7C" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Chrome zero-day vulnerability actively exploited</strong>: Google patched a high-severity zero-day vulnerability, CVE-2026-2441, in Chrome’s CSS engine, confirming active exploitation in the wild. The use-after-free bug allows remote code execution via crafted HTML pages. This incident underscores the ongoing memory safety challenges in C+±based browser engines and the critical importance of rapid patching, especially as discussions continue around migrating to safer languages like Rust in Chromium. <a href="https://app.daily.dev/posts/ISqaRw1RF" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Nvidia’s GPU fractioning boosts LLM throughput</strong>: NVIDIA Run:ai’s GPU fractioning allows 0.5 GPU shares to deliver 77% of full-GPU token throughput and 86% of concurrent users, optimizing LLM inference on H100 and HGX B200 clusters. <a href="https://app.daily.dev/posts/LsiRfCVWP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks zerobus ingest now generally available</strong>: Databricks has made Zerobus Ingest generally available, allowing direct event pushes into managed Delta tables, bypassing traditional message buses and reducing streaming infrastructure complexity. <a href="https://app.daily.dev/posts/0RtYZ2hnI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux kernel CVE assignment process detailed</strong>: Greg Kroah-Hartman explained the Linux kernel CVE team’s process, focusing on bugfix commits for stable backports and emphasizing that CVSS scores are often misleading for kernel vulnerabilities. <a href="https://app.daily.dev/posts/8whl1CT57" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Deno sandbox offers instant linux microvms</strong>: Deno Deploy now supports instant Linux microVMs via <code>deno sandbox</code>, provisioning cloud-hosted VMs with SSH access in under a second and featuring a secrets mechanism that prevents raw secret exposure. <a href="https://app.daily.dev/posts/7Ar3LQ3uM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kyverno 1.17 promotes cel-based policy engine to ga</strong>: Kyverno 1.17 has promoted its CEL-based policy engine to v1 (GA), stabilizing policy types and adding new namespaced mutation and generation policies for multi-tenancy. <a href="https://app.daily.dev/posts/fnlW905uT" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Tailscale peer relays and services now generally available</strong>: Tailscale has announced the general availability of Peer Relays for customer-deployed, high-throughput relay nodes and Tailscale Services for app-aware connectivity with enhanced control and Kubernetes Operator support. <a href="https://app.daily.dev/feed-by-ids?id=Q0uzlfbeJ&amp;id=j2xXHzw1I" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rust kernel support for freebsd nearing stability</strong>: FreeBSD’s Q4 2025 status report indicates that Rust kernel programming interfaces (KPIs) are expected to be stable enough for developers to experiment with new kernel code by early 2026. <a href="https://app.daily.dev/posts/H5Wur3zjw" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI will solve all your supply chain problems</strong>: Apparently, Claude Opus 4.6, connected via MCP to a database, can now diagnose all your supply chain failures and resolve disputes between warehouse and transportation teams. Because, you know, human planners are just too busy looking at ‘static dashboards’ instead of getting ‘data-driven answers through natural language.’ So, just fire all your logistics experts and let the AI handle it; what could possibly go wrong with a non-technical planner getting answers from a black box? <a href="https://app.daily.dev/posts/gdtsbiLsr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI will make you a 10x developer, just learn these tools</strong>: A software engineer turned manager is breathlessly proclaiming that developers must adopt AI tools like Kiro and Claude Code by 2026 or risk career stagnation, citing a ‘shifted productivity baseline.’ Meanwhile, a veteran developer with 20+ years of experience contends that LLMs, while generating code easily, do not improve ‘taste’ or skill, leading to a flood of derivative, low-quality applications. So, either you become an AI-powered coding god, or you’re obsolete. No pressure, just your entire career on the line, all thanks to tools that might just make everything worse. <a href="https://app.daily.dev/feed-by-ids?id=037a5YIjx&amp;id=3NTUXfPkW&amp;id=Ixa8rNXuv&amp;id=WHxN1Y9gh&amp;id=bzMzHxxko&amp;id=lOJYvaow4&amp;id=nsjV829m0&amp;id=wj4NMewlB&amp;id=zbh2s29Vr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI will make your code high quality, just ask it</strong>: High-quality software is defined by how easy it is to change, but LLMs currently lack the ability to build incrementally on prior learning. So, while AI can churn out code faster than you can say ‘technical debt,’ it’s not actually improving the fundamental quality that matters. It’s like getting a fancy new car that looks great but breaks down every other week because the engine was built by a committee of robots who never talked to each other. <a href="https://app.daily.dev/posts/vtcZQsWCv" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
