<h2>Must know</h2>
<ul>
<li>Apple taps Google Gemini for Siri and AI features: Apple is reportedly partnering with Google for a multi-year, non-exclusive deal to integrate Google’s Gemini AI models and cloud infrastructure into Siri and other AI features. This move, which could cost Apple around $1 billion annually, follows Apple’s evaluation of other models from OpenAI and Anthropic. The revamped Siri, powered by Gemini, is expected to launch later in 2026, with Apple emphasizing its commitment to privacy despite leveraging Google’s external infrastructure, as reported by multiple tech outlets. <a href="https://app.daily.dev/posts/joVK7B34Z" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Hyperscale AI data centers demand next-gen nuclear power: The insatiable energy demands of hyperscale AI data centers, housing hundreds of thousands of GPUs, are pushing the limits of current power grids. These facilities, backed by giants like OpenAI, Google, Microsoft, and Meta, each require over a gigawatt of electricity, largely sourced from fossil fuels. This massive consumption is accelerating interest in advanced nuclear options like small modular reactors (SMRs) to provide cleaner, more reliable power, according to industry reports. <a href="https://app.daily.dev/feed-by-ids?id=bf0xB32y7&amp;id=qk040xtMi" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Anthropic launches agentic AI for file management: Anthropic has introduced Cowork, a new feature in the Claude Desktop app for macOS, allowing Claude Max subscribers to use agentic AI for autonomous file management. This tool enables Claude to read, edit, and create files within a specified folder via a chat interface, handling tasks like organizing receipts and drafting reports without requiring command-line knowledge. While promising for productivity, it’s still a research preview with documented risks like prompt injection, as noted in Anthropic’s announcement. <a href="https://app.daily.dev/posts/TLkxIySjV" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li>AI coding agents get smarter with semantic search: AI coding agents like Claude Code are getting a significant upgrade with tools like grepai, which enables semantic search for code by intent rather than keywords, reducing token consumption by 39% and improving reasoning turns by 26% compared to traditional grep-based methods. <a href="https://app.daily.dev/posts/jcpFkjQkx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Postman acquires Fern for API documentation and SDK automation: Postman has acquired Fern, a startup specializing in automating API documentation and SDK generation in nine programming languages, aiming to reduce developer friction and support costs by providing a seamless, end-to-end API lifecycle platform. <a href="https://app.daily.dev/posts/9UqlZV9z7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>AWS CloudWatch gets a unified observability revamp: AWS has revamped CloudWatch into a unified observability platform, consolidating operational, security, and compliance logs across multiple AWS accounts, supporting OCSF and OpenTelemetry standards, and offering natural language queries, positioning it as a cost-effective alternative to Splunk and Datadog for AWS-centric organizations. <a href="https://app.daily.dev/posts/Z4eMJMdxN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Kubernetes v1.35 introduces workload-aware scheduling: Kubernetes v1.35 is out, bringing workload-aware scheduling with gang scheduling to prevent partial resource usage and deadlocks, alongside opportunistic batching for faster identical pod scheduling and improved service account token handling. <a href="https://app.daily.dev/feed-by-ids?id=R6fTuHjeX&amp;id=zTULpLxtX" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Linux 6.19-rc5 brings NVIDIA driver and RISC-V security fixes: Linux 6.19-rc5 has been released, featuring a fix for newer NVIDIA GPUs in the Nouveau open-source driver, addressing a RISC-V security vulnerability, and resolving a software bug in the Rust Binder, with two-thirds of the patch focused on driver updates. <a href="https://app.daily.dev/posts/MbuLWzK6Y" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Budgie desktop shifts to Wayland and Qt6: The Budgie 10.10 desktop environment has transitioned to native Wayland support, dropping X11-specific components, and Budgie 11 is being re-architected with the Qt6 toolkit for improved modularity and maintainability, while IceWM continues its X11-focused development. <a href="https://app.daily.dev/posts/peCpqhbhc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Auto-CPUFreq 3.0 for Linux adds enhanced battery controls: Auto-CPUFreq 3.0 for Linux laptops now offers user-controlled CPU turbo mode override via CLI and GUI, improved battery handling for multi-battery setups, and ASUS laptop battery charge threshold support, allowing users to balance battery life and performance. <a href="https://app.daily.dev/posts/umWMqgJxW" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li>AI code generation: cheap code, expensive reliability: Apparently, LLM tools like Claude Code make code generation fast and accessible, even for ‘non-developers.’ But here’s the kicker: building reliable, maintainable, and scalable software is still expensive and requires actual expertise. Salesforce even laid off 4,000 experienced staff to replace them with AI, only to reverse course three months later because the automated systems couldn’t match human effectiveness. So, while code might be cheap, competence certainly isn’t. <a href="https://app.daily.dev/feed-by-ids?id=BCKlippws&amp;id=aKgPqgLac" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Anthropic’s OAuth restrictions: a bold move or a blunder?: Anthropic, in its infinite wisdom, decided to restrict OAuth access for third-party coding agents like OpenCode, forcing them to use higher-cost API access. This, of course, is framed as an attempt to ‘control the value chain’ and avoid being a ‘commodity model provider.’ Meanwhile, OpenAI, ever the opportunist, publicly announced support for using ChatGPT subscriptions in OpenCode, effectively inviting disgruntled Anthropic users to jump ship. It’s almost like treating your developer ecosystem like a cash cow isn’t always the smartest long-term strategy. <a href="https://app.daily.dev/feed-by-ids?id=8dtXe7P09&amp;id=eRi1VQ5SP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>AI agent security: just box it up and call it a day!: With AI agents increasingly achieving goals via tool usage, traditional security is apparently useless. The new hotness? ‘Deterministic security boxing,’ which means sticking agents in network-isolated environments with strict policies, regardless of prompt or exploit. Because, you know, external controls are always stronger than internal alignment. Never mind that 86% of organizations are already experiencing AI-related incidents; just put it in a box, and all your problems will magically disappear. What could possibly go wrong? <a href="https://app.daily.dev/feed-by-ids?id=CuSSpmwf6&amp;id=EDYtMcTfL&amp;id=JuPUyAfNx" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>