<h2>Must know</h2>
<ul>
<li><strong>AWS US-EAST-1 outage causes global chaos</strong>: A major AWS outage in the US-EAST-1 region on October 20, 2025, disrupted global services for hours, affecting giants like Alexa, Fortnite, and Zoom. AWS attributed the issue to DNS resolution failures with DynamoDB API endpoints, though no detailed technical root cause was disclosed. The incident highlighted the world’s heavy reliance on AWS and raised concerns about Amazon’s operational resilience amid high staff attrition, as suggested by The Register.</li>
<li><strong>New tools for porting windows apps to linux and macos</strong>: Two new tools, InstallAware’s Application Porting Toolkit and the open-source WinBoat, were released on October 20, 2025, to simplify running Windows applications on other operating systems. InstallAware uses Wine for API translation, while WinBoat virtualizes a full Windows system within Docker and KVM, making apps like Photoshop and Office 365 run more reliably on Linux. These tools offer different approaches to cross-platform compatibility, catering to various user needs.</li>
<li><strong>Rails 8.1.0.rc1 brings new features and dark mode</strong>: Rails 8.1.0.rc1 is out, introducing bound SQL literals in Common Table Expressions, prefix options for <code>has_secure_token</code>, and improved error handling for missing environment files. The release also adds dark mode to error pages and automatically updates advisory databases for <code>bundler-audit</code>, enhancing both developer experience and security. These updates reflect ongoing efforts to refine the Ruby on Rails framework.</li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>JUnit 6.0.0 unifies and modernizes java testing</strong>: JUnit 6.0.0 has been released, unifying versioning across its sub-projects, requiring Java 17, and adding native Kotlin suspend function support, a CancellationToken API, and JSpecify nullability annotations.</li>
<li><strong>OpenTelemetry simplifies configuration and governance</strong>: OpenTelemetry introduced declarative YAML configuration to filter health check endpoints from traces and is holding elections for its Governance Committee, emphasizing community-driven development and sustainability.</li>
<li><strong>AMD ROCm 7.9 preview boosts linux AI hardware support</strong>: AMD’s ROCm Core SDK 7.9 preview introduces the new TheRock build infrastructure and early support for Ryzen AI Max 300 Series APUs and MI350/MI300 GPUs, accelerating Linux-centric AI workloads.</li>
<li><strong>AI agents get structured reasoning and control</strong>: New advancements in AI agents include Attentive Reasoning Queries (ARQs) for reducing hallucinations, Samesurf’s cloud-based control plane for web navigation, and Salesforce’s Agent Graph for reliable multi-agent workflows.</li>
<li><strong>Open-source tools for security and LLM evaluation</strong>: Google released LLM-Evalkit for prompt engineering, Auth0 launched CheckMate for security best practices, and Terminal-Bench offers a suite for evaluating AI agents on real-world tasks.</li>
<li><strong>Zorin OS 18 sees surge after windows 10 end-of-life</strong>: Zorin OS 18, launched on the same day as Windows 10’s end of support, saw over 100,000 downloads in two days, with 72% from Windows users seeking alternatives to Windows 11’s hardware requirements.</li>
<li><strong>UK launches digital ID for veterans in <a href="http://gov.uk" target="_blank" rel="noopener nofollow">gov.uk</a> app</strong>: The UK government has soft-launched a digital Veteran Card within the <a href="http://GOV.UK" target="_blank" rel="noopener nofollow">GOV.UK</a> One Login app, marking the first step in a broader digital ID initiative planned to include driving licenses and voter ID.</li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI infrastructure: more power, less optimization, brilliant!</strong>: Big Tech, including Meta, Microsoft, and Amazon, are throwing billions at nuclear power and gas turbines for AI infrastructure, because apparently, optimizing models for efficiency is just too much work. Meanwhile, startups are training massive models for peanuts. It’s almost like they’re more interested in building bigger power plants than smarter AI, ensuring we all pay for their ‘innovation.’</li>
<li><strong>Enterprise AI agents: productivity gains or just more meetings?</strong>: Adobe’s AI Foundry and Microsoft’s Agent Mode promise custom AI models and automated workflows, with Lloyds Banking Group claiming 46 minutes saved daily per employee. Yet, these ‘gains’ often require human verification due to errors, and independent reports show variable results. It seems the real innovation is in gamifying metrics to justify the AI spend, rather than delivering actual, verifiable productivity.</li>
<li><strong>Humanoid robots: still mostly hype, but with $65 billion</strong>: Over $65 billion has been poured into humanoid robotics, with promises of robots working alongside humans in warehouses and healthcare. Yet, these marvels still struggle with basic reliability, cost a fortune, and fall over a lot. It seems the ‘imminent’ replacement of human workers is still mostly a marketing fantasy, proving that throwing money at a problem doesn’t magically solve fundamental engineering challenges.</li>
</ul>
