<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. This initiative aims to create AI that understands the real world, offering an alternative to current large language models (LLMs). The company, co-founded with Alex LeBrun, is reportedly seeking a $3.5 billion valuation and plans to target high-stakes sectors like healthcare and robotics, as reported by various tech outlets. They emphasize persistent memory, reasoning, and safety, though no products or demonstrations have been released yet. <a href="https://app.daily.dev/posts/CvMr0F9gc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Oracle’s Orasort patent expires, open source databases eye integration</strong>: Oracle’s patent for the ‘Orasort’ sorting algorithm has expired after 20 years, opening the door for its adoption in open-source databases. The algorithm, which combines common prefix skipping, adaptive quicksort/radix sort, and key substring caching, claims up to a 5x performance improvement. According to inventor Mark Callaghan, this could significantly boost efficiency for projects like MySQL and PostgreSQL, which are already considering integration. This is a big win for the open-source community, potentially bringing enterprise-grade performance to widely used database systems. <a href="https://app.daily.dev/posts/aheyA2T6F" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>TikTok’s new privacy policy expands data collection</strong>: TikTok has updated its U.S. privacy policy, now explicitly collecting precise GPS location data from U.S. users, tracking AI interactions (prompts and outputs), and expanding 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 disclosure is legally mandated by state laws like CCPA and isn’t a new collection practice, as detailed in various reports. <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>Rust-based deep learning framework Burn offers flexibility</strong>: Burn is a new deep learning framework and tensor library written in Rust, designed for flexibility, efficiency, and portability across various backends like CUDA, ROCm, Metal, and CPU, aiming to bridge the Python/C++ gap in ML. <a href="https://app.daily.dev/posts/EAj5GDC5K" 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, improving functionality and compatibility for Linux users. <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 and system command features</strong>: Newelle, the AI assistant for GNOME, now supports local LLMs via llama.cpp and can execute system commands, expanding its utility and offering users more flexibility and automation capabilities. <a href="https://app.daily.dev/posts/Au9UNs8LJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rails updates bring performance and maintenance improvements</strong>: Rails introduced a 2x speed improvement to <code>String#squish</code>, enhanced error handling for stream rendering, and improved <code>console reload!</code> to prevent query cache artifacts, alongside other generator and performance optimizations. <a href="https://app.daily.dev/feed-by-ids?id=WT6Ixi3bR&amp;id=Ji6vBmphI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GNU C Library 2.43 enhances C23 support and system calls</strong>: GNU C Library 2.43 has been released, expanding C23 language support, adding new system call wrappers like <code>mseal</code> and <code>openat2</code> for Linux, and improving FMA performance on AMD Zen processors. <a href="https://app.daily.dev/posts/4kkaXLeFv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mobile MCP platform enables cross-device AI automation</strong>: Mobile MCP is a new Model Context Protocol server designed to let LLMs and AI agents automate and interact with native iOS and Android apps across simulators, emulators, and real devices, integrating with AI coding assistants like Claude and Copilot. <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>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI agent skills automate database migration in minutes, or so they say</strong>: Neo4j is touting specialized AI agent skills that supposedly automate database migration tasks, claiming to reduce weeks of work to mere minutes. They even provided an example of migrating 94+ Cypher queries across 24+ Lambda functions in 6 minutes. While the idea of AI handling tedious migrations is appealing, one has to wonder about the edge cases and the inevitable ‘human in the loop’ that always seems to pop up when these ‘miracle’ solutions hit the real world. <a href="https://app.daily.dev/posts/znLf6Yx7x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding tools: the new ‘skilled new engineer’ that needs hand-holding</strong>: Apparently, AI coding assistants are now to be treated like ‘skilled new engineers’ who require ‘targeted, relevant inputs’ and not ‘exhaustive data dumps.’ This comes after developers found AI tools often fail when used as a last resort or with too much context. So, after all the hype about AI writing 90% of our code, it turns out we still need to babysit them, provide focused prompts, and manage their ‘over-configuration’ issues. Sounds less like a revolution and more like another junior dev to train. <a href="https://app.daily.dev/feed-by-ids?id=BCMHZegjN&amp;id=LMNhcIFC9&amp;id=kVmMac2gZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Software estimation: still a political game, not engineering</strong>: A recurring theme in the tech world is that software estimation is ‘fundamentally impossible’ due to unpredictable unknowns, and it’s really just a ‘political tool’ for managers to negotiate resources. Engineers are advised to focus on communicating ‘unknowns’ and building trust, rather than providing accurate timelines. So, after decades, we’re still pretending estimates are about engineering while everyone knows it’s just a dance to get things done. Good to know some things never change. <a href="https://app.daily.dev/feed-by-ids?id=HxseyW07U&amp;id=vTe4dHYnt" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
