<h2>Must know</h2>
<ul>
<li><strong>Mandiant drops NTLMv1 rainbow tables, urges immediate deprecation</strong>: Mandiant has publicly released rainbow tables capable of cracking Net-NTLMv1 password hashes in under 12 hours using consumer hardware. This move, detailed by Mandiant, aims to force organizations to finally deprecate the 20-year-old, vulnerable protocol, which is still surprisingly widespread. They’ve provided tools and guidance for security teams to detect and disable NTLMv1, emphasizing that credential theft is trivial when this legacy authentication is in use. <a href="https://app.daily.dev/posts/2q2phjvWS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Major vendors ditch MITRE ATT&amp;CK evaluations</strong>: Microsoft, SentinelOne, and Palo Alto Networks have pulled out of the 2025 MITRE ATT&amp;CK Evaluations, citing resource prioritization for product development. MITRE’s CTO admitted the 2025 tests are more demanding, now including cloud environments and alert volume tracking. Industry experts are calling the evaluations ‘vendor theater,’ suggesting they’re more for PR than genuine security improvement, as reported by CSO Online. <a href="https://app.daily.dev/feed-by-ids?id=c0ZYfkcgP&amp;id=hu11lcW1J" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Bluetooth ‘WhisperPair’ flaw exposes millions of devices</strong>: A significant Bluetooth vulnerability, dubbed ‘WhisperPair,’ affects hundreds of millions of accessories using Google’s Fast Pair system. Researchers from KU Leuven found that widespread improper implementation of Google’s security specifications allows attackers within Bluetooth range to hijack devices, injecting audio or enabling microphones. Google is working on patches, but many low-end peripherals may never receive updates, leaving users exposed. <a href="https://app.daily.dev/feed-by-ids?id=1QuQlmOmw&amp;id=onISZaYqR" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>NVIDIA pushes AI gaming and streaming with DLSS 4.5</strong>: NVIDIA announced DLSS 4.5 with Dynamic Multi Frame Generation for up to 5 additional frames per rendered frame, enabling 240+ FPS path-traced gaming on RTX 50 Series GPUs, alongside new G-SYNC Pulsar monitors and expanded GeForce NOW cloud gaming. <a href="https://app.daily.dev/posts/jXjzYY9vI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI data centers face skilled worker shortage</strong>: The booming AI data center construction market is creating a critical shortage of skilled workers, including engineers and technicians, with the U.S. needing nearly 400,000 more construction workers by 2033, prompting new training programs from tech giants and educational institutions. <a href="https://app.daily.dev/posts/JE89HgTw1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Black Basta ransomware leader added to Most Wanted list</strong>: Ukrainian and German authorities have identified Oleg Nefedov as the alleged leader of the Black Basta ransomware group, adding him to the EU Most Wanted list and an INTERPOL Red Notice, following a leak of internal chat logs in early 2025. <a href="https://app.daily.dev/posts/O0PAVFWl9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA expands Enterprise AI Factory with BlueField cybersecurity</strong>: NVIDIA’s Enterprise AI Factory now integrates BlueField cybersecurity and infrastructure acceleration, offloading networking, storage, and security tasks to dedicated processors, enabling zero-trust hardware-accelerated security for AI workloads. <a href="https://app.daily.dev/posts/SCEwsKfyS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LLMOps course focuses on production-ready LLMs</strong>: A new LLMOps course is available, focusing on strategies for configuring and orchestrating large language models in production, covering topics like decoding strategies, cost optimization, and evaluation system design, distinct from traditional MLOps. <a href="https://app.daily.dev/posts/1q51Gs3MM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare R2 SQL adds aggregation support</strong>: Cloudflare has introduced aggregation support to R2 SQL, allowing developers to run analytical queries like SUM, COUNT, and AVG directly on R2-stored data with GROUP BY and HAVING clauses, eliminating the need for separate data warehouses. <a href="https://app.daily.dev/posts/buc7GV2gZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Porffor compiler targets WebAssembly for JavaScript</strong>: Porffor is an ahead-of-time JavaScript compiler that targets WebAssembly and native code, promising 50x smaller binaries and significant performance gains for use cases like serverless functions and game engines, though it currently lacks garbage collection and web APIs. <a href="https://app.daily.dev/posts/xTAvQVuAU" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI-generated code: the gift that keeps on giving… technical debt</strong>: Multiple engineering teams are reporting that AI-generated code, while compiling initially, often lacks architectural context and creates poor state management, leading to major refactoring that can take longer than writing it from scratch. Apparently, the ‘decent toy version’ a Google engineer claimed Claude Code could produce in an hour isn’t quite production-ready, despite the hype that AI would eliminate developers. Who knew that complex system design required, you know, actual thought? <a href="https://app.daily.dev/feed-by-ids?id=STWE5tRXe&amp;id=UAphs9F5B&amp;id=euzGVQGzQ&amp;id=gMHZMigM0&amp;id=m6P4U0J0b&amp;id=wCQfiCXDa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents: just APIs with a fancy new hat</strong>: Sources are pointing out that ‘AI agents’ are mostly a rebranding of existing API-centric automation, driven by new investment rounds rather than genuine technical breakthroughs. It seems the industry is just slapping ‘agentic technology’ on everything to justify the next funding bubble. Meanwhile, the real advancements are still in building machine-readable APIs and semantic infrastructure, not in the marketing department’s latest buzzword bingo. <a href="https://app.daily.dev/feed-by-ids?id=4rJ96gwiU&amp;id=6IAZSWIz1&amp;id=6XeVWOBXO&amp;id=77jDjhI5C&amp;id=sH4hATTfN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LLMs are ‘economically unsustainable’ (unless you’re NVIDIA)</strong>: Research suggests large language models with over 100 billion parameters are ‘economically unsustainable’ due to inference costs, pushing a trend towards smaller, specialized SLMs. Conveniently, NVIDIA is simultaneously releasing open-source AI blueprints for retail and touting how their GPUs accelerate everything from video generation to SLM inference. So, LLMs are too expensive for everyone else, but somehow NVIDIA’s hardware is the magic bullet for making AI ‘democratized’ and ‘cost-effective.’ Right. <a href="https://app.daily.dev/feed-by-ids?id=J1BKrq2u6&amp;id=J1IZlkzNv&amp;id=Yuaq0Gyje&amp;id=dliP781v1" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
