<h2>Must know</h2>
<ul>
<li><strong>Yann LeCun launches AMI Labs to build ‘world models’</strong>: Yann LeCun, a prominent figure in AI, has launched AMI Labs, a new startup focused on developing ‘world models’ for intelligent systems. The company aims to offer an alternative to current large language models by emphasizing persistent memory, reasoning, and safety. With Alex LeBrun as CEO, AMI Labs is reportedly seeking funding at a $3.5 billion valuation, targeting high-stakes applications in healthcare and robotics, as reported by various sources. This move signals a potential shift in AI development focus, moving beyond the current generative AI paradigm. <a href="https://app.daily.dev/posts/CvMr0F9gc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding tools are getting smarter, but still need you</strong>: AI coding assistants are evolving, with tools like Claude Code reportedly gaining a ‘Swarms’ feature that allows the AI to act as a team lead, coordinating specialist agents for complex tasks. This, along with the release of open-source alternatives like Craft Agents, suggests a move towards more sophisticated AI orchestration in development. However, experts emphasize that effective use still requires developers to provide focused context and iterative refinement, treating the AI as a skilled engineer rather than an oracle, as noted in developer commentary. <a href="https://app.daily.dev/feed-by-ids?id=BCMHZegjN&amp;id=LMNhcIFC9&amp;id=kVmMac2gZ&amp;id=ClAzJhvfM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>TikTok’s new privacy policy expands data collection</strong>: TikTok’s updated U.S. privacy policy now explicitly collects precise GPS location data from U.S. users, tracks AI interactions (prompts and outputs), and expands its ad network data use for targeted advertising across the web. Users must accept these terms to continue using the app. While concerns about ‘immigration status’ collection were raised, privacy lawyers clarify this is a legal disclosure requirement, not a new collection practice, as detailed in various reports. The significant changes are the new GPS and AI interaction data collection. <a href="https://app.daily.dev/feed-by-ids?id=BFPMM0rX7&amp;id=nbFKCoxnX" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Oracle’s Orasort patent expires, open source benefits</strong>: Oracle’s patent for the ‘Orasort’ sorting algorithm has expired, opening the door for its adoption in open-source databases like MySQL and PostgreSQL, promising up to 5x performance improvements. <a href="https://app.daily.dev/posts/aheyA2T6F" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LLM observability needs a new approach</strong>: Traditional observability methods are falling short for LLM-powered applications, requiring new approaches to track prompt versions, agent pipelines, token usage, and information retrieval processes, with eBPF emerging as a key tool for infrastructure-level monitoring. <a href="https://app.daily.dev/posts/2NT8yV81n" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 6.19 kernel adds ASUS Armoury driver</strong>: The Linux 6.19 kernel now includes the ASUS Armoury driver, enhancing support for ASUS ROG gaming devices, including ROG Ally handhelds and various ROG laptops. <a href="https://app.daily.dev/posts/rvv41Cn0C" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GNOME’s Newelle AI assistant gets local LLM support</strong>: Newelle, the AI assistant for GNOME, now supports local LLMs via llama.cpp and can execute system commands, expanding its utility and reducing cloud dependence. <a href="https://app.daily.dev/posts/Au9UNs8LJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mobile MCP platform enables cross-device AI automation</strong>: The Mobile MCP (Model Context Protocol) server allows LLMs and AI agents to automate and interact with native iOS and Android apps across devices, providing a platform-agnostic interface for tasks from UI testing to multi-step automation. <a href="https://app.daily.dev/posts/F3rJpKxUE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Wix releases React Native UI component library</strong>: Wix has launched React Native UI Lib, a new UI component library offering pre-built, customizable components with integrated theme management and modifier-based styling for React Native applications. <a href="https://app.daily.dev/posts/3XpEcphch" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rails gets performance and maintenance updates</strong>: Rails introduced a 2x speed improvement to <code>String#squish</code>, enhanced error handling for stream rendering, and improved console <code>reload!</code> functionality, alongside other updates for better performance and developer experience. <a href="https://app.daily.dev/feed-by-ids?id=WT6Ixi3bR&amp;id=Ji6vBmphI" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI will replace you, says survey (again)</strong>: A new report claims 80% of workers are worried about AI taking their jobs, yet only 23% actually understand which roles are at risk. This sounds less like a groundbreaking insight and more like a rehash of every ‘AI is coming for your job’ headline since 2015. The article offers vague ‘strategies for future-proofing careers,’ which probably means ‘learn to prompt better’ or ‘become an AI ethicist.’ <a href="https://app.daily.dev/posts/VJ0dfMh36" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Software estimation is political, not technical (shocking!)</strong>: Apparently, software project estimation is ‘fundamentally impossible’ because it’s all about ‘unpredictable, unknown work’ and managers using estimates as ‘political tools.’ Who knew that the art of guessing how long something takes, especially when you don’t know what ‘something’ is, isn’t a precise science? The solution? ‘Build trust with management’ and ‘communicate unknowns.’ Groundbreaking stuff, really. <a href="https://app.daily.dev/feed-by-ids?id=HxseyW07U&amp;id=vTe4dHYnt" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated content for kids: What could go wrong?</strong>: Sparkli, a startup by ex-Google employees, just secured $5 million to create an AI-driven educational app for kids aged 5-12, generating multimedia content on-demand in two minutes. They claim ‘built-in safety features,’ but with AI’s track record, one has to wonder if ‘on-demand’ also means ‘on-demand hallucinations’ or ‘on-demand inappropriate content.’ What could possibly go wrong with an AI generating educational material for children without independent third-party assessment? <a href="https://app.daily.dev/posts/CoMIJ1Dxu" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
