<h2>Must know</h2>
<ul>
<li><strong>AI agents are a security nightmare for IAM</strong>: AI agents are fundamentally breaking traditional Identity and Access Management (IAM) systems. Unlike humans, these agents compile broad, persistent permissions from multiple users, creating untraceable authorization bypasses and significant security risks. Organizations need new protocols to track agent-user-system relationships, treating agents as high-risk entities rather than simple automation tools, according to multiple reports. <a href="https://app.daily.dev/feed-by-ids?id=nUFRFzE1V&amp;id=nUFRFzE1V" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google Cloud opens public datasets via Apache Iceberg</strong>: Google Cloud now offers public datasets through the Apache Iceberg REST Catalog on BigLake, making them accessible as read-only Iceberg tables. This means data engineers and analysts can query datasets like the NYC Taxi data using any compute engine supporting the Iceberg protocol (Spark, Trino, Flink, BigQuery) without any infrastructure setup. It’s all about immediate, easy access to analytical data, as announced by Google Cloud. <a href="https://app.daily.dev/posts/tn371FITl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft Aspire expands beyond .NET, embraces AI/ML</strong>: Microsoft’s Aspire platform is shedding its .NET-only skin, now supporting JavaScript, TypeScript, Python, and Java for distributed application development. This move aims to attract a broader developer base, especially for AI/ML and web projects. Microsoft is even running a six-month hackathon with prizes for AI/ML solutions to drive adoption, as reported by Microsoft. <a href="https://app.daily.dev/feed-by-ids?id=gRAkh5lrI&amp;id=gRAkh5lrI" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>GNOME AI assistant Newelle gets local LLM support</strong>: Newelle, the AI virtual assistant for GNOME, now supports llama.cpp, allowing users to run large language models locally on their devices, alongside existing integrations with cloud APIs like Google Gemini, OpenAI, and Groq. <a href="https://app.daily.dev/posts/Au9UNs8LJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI shifts software engineering focus to SRE</strong>: With AI making code-writing easier, the core value in software engineering is shifting towards operational excellence and reliability, making Site Reliability Engineering (SRE) increasingly critical for maintaining trusted and dependable services. <a href="https://app.daily.dev/feed-by-ids?id=otuC5YedQ&amp;id=otuC5YedQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Oracle’s Orasort patent expires, open source adoption eyed</strong>: Oracle’s ‘Orasort’ sorting algorithm, which offers a 5x performance boost over older methods, is now open source after its patent expired, sparking immediate interest from the MySQL and PostgreSQL communities. <a href="https://app.daily.dev/posts/aheyA2T6F" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Compyle slashes cloud terminal latency to 14ms</strong>: Compyle drastically reduced cloud development environment terminal latency from 200ms to 14ms by implementing a warm pool, removing a socket server middleman, deploying multi-regional sandbox pools, and simplifying architecture. <a href="https://app.daily.dev/posts/2R6Ajv0Sw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Observability for LLMs needs a new playbook</strong>: Traditional observability tools are falling short for LLM-powered applications, requiring new telemetry like prompt versioning, agent pipeline tracing, token usage monitoring, and retrieval analysis to manage costs and ensure reliability. <a href="https://app.daily.dev/posts/2NT8yV81n" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Skip tool open-sourced for cross-platform Swift development</strong>: Skip, a tool that converts Swift code for native iOS and Android apps using SwiftUI and Jetpack Compose, has been fully open-sourced, removing all paywalls and shifting to community funding. <a href="https://app.daily.dev/posts/7snXi8ahF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>DXVK-NVAPI 0.9.1 adds DLSS override, fixes game crashes</strong>: DXVK-NVAPI version 0.9.1 introduces environment variables for manually overriding DLSS scaling ratios and fixes startup crashes in several games, including ‘Middle-earth: Shadow of Mordor’ and ‘The Last of Us Part 2’. <a href="https://app.daily.dev/feed-by-ids?id=6hOCF9MBP&amp;id=6hOCF9MBP&amp;id=6hOCF9MBP" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI coding tools: From skepticism to 90% Claude Code</strong>: A developer, once skeptical, now claims 90% of their code is authored by Claude Code, noting its bug detection and iterative improvements. They argue that debates around LLMs have become ‘tribal and political,’ urging experimentation over entrenched positions. So, apparently, we’re all just waiting for the AI to take over, and anyone who disagrees is just being ‘tribal.’ <a href="https://app.daily.dev/posts/BCMHZegjN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Model Context Protocol: The new buzzword for AI automation</strong>: Mobile MCP is here to let LLMs and AI agents automate native iOS and Android apps, promising a ‘platform-agnostic interface’ and integration with all your favorite AI coding assistants like Claude and Copilot. Then there’s ‘Port of Context’ (pctx), an ‘open-source, vendor-agnostic’ implementation of ‘code mode’ that supposedly slashes token usage by 98.7%. Because, you know, we needed more acronyms and ‘vendor-agnostic’ solutions that still integrate with all the big vendors. <a href="https://app.daily.dev/feed-by-ids?id=F3rJpKxUE&amp;id=F3rJpKxUE&amp;id=mcXaLHnCP&amp;id=mcXaLHnCP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Coi language claims to ‘beat React/Vue’ with WASM</strong>: A new component-based language called Coi is here, targeting WebAssembly and boldly claiming ‘performance superiority’ over React and Vue. Their benchmarks show 100 FPS canvas rendering versus Emscripten’s 40 FPS, thanks to ‘O(1) reactivity’ and no virtual DOM diffing. All these claims, of course, come directly from the Coi project’s own publicity. Nothing like a self-proclaimed champion to get the hype train rolling. <a href="https://app.daily.dev/feed-by-ids?id=iOGOTjzat&amp;id=iOGOTjzat&amp;id=iOGOTjzat" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
