<h2>Must know</h2>
<ul>
<li><strong>AI code generation: hype vs. reality</strong>: Despite claims of AI saving developers 10+ hours weekly, many engineering teams report AI-generated code often compiles but lacks architectural context, leading to significant refactoring or even complete rewrites. According to multiple reports, this can degrade project velocity and introduce technical debt, especially when AI produces ‘realistic’ but non-existent API elements. The real productivity gains from AI emerge only when it’s integrated across the entire development lifecycle, not just for isolated coding tasks. <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&amp;id=gMHZMigM0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mandiant releases NTLMv1 rainbow tables, urges deprecation</strong>: Mandiant has publicly released rainbow tables capable of cracking Net-NTLMv1 password hashes in under 12 hours using consumer hardware, hosted on Google Cloud. This move aims to force organizations to finally deprecate the 20-year-old, vulnerable protocol, which Mandiant notes is still widely used despite Microsoft’s recommendations. The release includes guides for security teams to detect and disable NTLMv1, emphasizing the triviality of credential theft when this legacy protocol 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: Microsoft, SentinelOne, and Palo Alto Networks have withdrawn from the 2025 MITRE ATT&amp;CK Evaluations, citing resource prioritization for product development. MITRE’s CTO admitted the 2025 tests became more demanding, expanding to cloud environments and alert volume tracking. Industry experts are calling the evaluations ‘vendor theater,’ suggesting they serve PR more than genuine security improvement, prompting MITRE to plan a reboot of its vendor forum in 2026. <a href="https://app.daily.dev/feed-by-ids?id=c0ZYfkcgP&amp;id=hu11lcW1J" 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 AI data center boom is creating a critical shortage of skilled workers, with demand for electricians, engineers, and technicians far outstripping supply, prompting major tech companies and educational institutions to launch new training programs. <a href="https://app.daily.dev/posts/JE89HgTw1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google and Anthropic launch new AI assistant features</strong>: Claude released Co-work for desktop automation, while Google introduced Personal Intelligence for Gemini, integrating Gmail, Photos, YouTube, and Search for contextual assistant responses, intensifying competition in the AI assistant ecosystem. <a href="https://app.daily.dev/posts/3e5iSak1o" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA boosts local AI performance on RTX GPUs</strong>: NVIDIA announced significant local performance upgrades for AI video/image generation on RTX GPUs, including 3x faster video generation and 60% VRAM reduction with NVFP4/FP8 quantization, and 35% faster SLM inference via RTX acceleration for Ollama and llama.cpp. <a href="https://app.daily.dev/feed-by-ids?id=1q51Gs3MM&amp;id=JItr1KihQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-powered accelerator management at Berkeley Lab</strong>: Lawrence Berkeley National Laboratory is using an LLM-driven Accelerator Assistant on NVIDIA H100 GPUs to autonomously troubleshoot and manage their particle accelerator, reducing setup time by 100x and demonstrating a blueprint for AI in complex scientific infrastructure. <a href="https://app.daily.dev/posts/ZzMUF9PNc" 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 leaked chat logs that revealed the group’s structure. <a href="https://app.daily.dev/posts/O0PAVFWl9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>US GPU supply constraints and AI export tariffs</strong>: Nvidia and Asus confirmed RTX 5070 Ti and 5060 Ti GPUs are not cancelled, but persistent supply shortages are projected until 2027, while new US export controls mandate a 25% tariff on AI chips for China, significantly impacting Nvidia’s market share there. <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>Bullshit police</h2>
<ul>
<li><strong>AI agents: just APIs with a fancy new hat?</strong>: Apparently, ‘AI agents’ are the hot new thing, but some cynics are pointing out that these ‘novel’ technologies are just a rebranding of APIs, web automation, and semantic web standards that have been around for over a decade. It seems the real ‘innovation’ is in attracting new investment rounds and financialization, rather than any groundbreaking technical advancements. Who knew a good marketing team could turn old wine into ‘agentic’ gold? <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 ‘unsustainable’ but SLMs will save us all (and make us rich)</strong>: Research now claims that large language models are ‘economically unsustainable’ due to inference costs, but fear not, because ‘specialized Small Language Models’ (SLMs) with a mere 7 million parameters can apparently outperform their bigger, dumber cousins. We’re told the future is a ‘digital factory’ of cheap SLMs, democratizing AI and reducing costs by ‘several orders of magnitude.’ Sounds like another round of ‘this time it’s different’ for the AI gold rush. <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>
<li><strong>AppGuard says AI malware detection is ‘fundamentally flawed’</strong>: In a bold move, AppGuard is calling out AI-based malware detection as ‘fundamentally flawed’ because, apparently, AI can’t parse ‘infinite possibilities.’ Instead of fancy AI, they’re pushing a ‘default-deny’ approach to reduce the attack surface. It’s almost as if some companies are realizing that throwing ‘AI’ at every problem isn’t a silver bullet, and sometimes, good old-fashioned controls are just… better. <a href="https://app.daily.dev/posts/M7evoWVEa" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
