Atomic Spin

CSO Online

Daily Dose of Data Science | Avi Chawla | Substack

Hacker News

IEEE Spectrum

InfoQ

<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 even provided guides on how security teams can acquire NTLMv1 hashes via authentication coercion attacks, emphasizing that credential theft is trivial when this legacy protocol is in use. It’s a public service, they say, to push for better security hygiene. <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 became more demanding, expanding to cloud environments and alert volume tracking. Industry experts are calling the whole process ‘vendor theater,’ suggesting it’s more about PR than actual security improvement. MITRE plans to reboot its vendor forum in 2026 to realign with industry needs, but this exodus from major players is a significant blow to the evaluation’s perceived credibility. <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>AI and non-human identity management for cloud security</strong>: Managing Non-Human Identities (NHIs)—machine credentials like tokens and keys—is becoming critical for cloud data security, especially with the rise of AI. Both CSO Online and other industry sources emphasize that a comprehensive, AI-driven lifecycle approach, from discovery to remediation, is essential. Fragmented tools won’t cut it; organizations need integrated platforms to reduce data breach risks, ensure compliance, and automate security in complex environments like healthcare and financial services. CISOs need to get involved to bridge the gap between development, security, and R&amp;D teams. <a href="https://app.daily.dev/feed-by-ids?id=lq3eNhRBB&amp;id=rc0IgUSh5" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>NVIDIA expands Enterprise AI Factory with BlueField cybersecurity</strong>: NVIDIA’s Enterprise AI Factory now integrates BlueField for hardware-accelerated, zero-trust cybersecurity and infrastructure offloading, partnering with nine ecosystem vendors to secure AI workloads at speed. <a href="https://app.daily.dev/posts/SCEwsKfyS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Pulumi adds native support for Terraform and HCL</strong>: Pulumi now natively supports Terraform and HCL, allowing direct execution and state management within Pulumi Cloud, with a financial program to ease migration for HashiCorp users. <a href="https://app.daily.dev/posts/brDXTtnxJ" 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 massive demand for skilled workers—electrical engineers, HVAC technicians, and construction specialists—with companies recruiting from diverse industries and launching training programs to fill the gap. <a href="https://app.daily.dev/posts/JE89HgTw1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LLMOps course focuses on production challenges</strong>: A new LLMOps course highlights strategies for configuring and orchestrating large language models in production, covering decoding, generation parameters, and lifecycle management to optimize cost, reduce hallucinations, and manage latency. <a href="https://app.daily.dev/posts/1q51Gs3MM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA boosts AI video generation 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, enabling 4K AI video creation on consumer hardware. <a href="https://app.daily.dev/posts/JItr1KihQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong><a href="http://LWN.net" target="_blank" rel="noopener nofollow">LWN.net</a> hit by DDoS attack</strong>: <a href="http://LWN.net" target="_blank" rel="noopener nofollow">LWN.net</a> is experiencing a severe DDoS attack from tens of thousands of scrapers, causing major slowdowns and highlighting the challenge content sites face with automated content theft and search engine prioritization of scraped material. <a href="https://app.daily.dev/posts/hTdHReUkc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Bad JSON costs enterprises millions</strong>: Malformed JSON and schema drift are costing enterprises millions annually through silent data corruption, emphasizing the need for defensive programming, schema validation, and shift-left testing to prevent costly downstream failures. <a href="https://app.daily.dev/posts/MaYOMR00v" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI agents: old wine in new bottles, or actual progress?</strong>: Apparently, ‘AI agents’ are the hot new thing, but some folks are pointing out that it’s mostly just a rebranding of APIs and automation that’s been around for a decade. While developers are enthusiastically building Pokémon Trainer AIs and using Claude Code for syntopical reading, the skeptical voices suggest this ‘agentic technology’ hype is more about attracting new investment than genuine technical breakthroughs. So, are we just calling old dogs new names, or is there actually something revolutionary happening beyond the marketing spin? <a href="https://app.daily.dev/feed-by-ids?id=4rJ96gwiU&amp;id=6IAZSWIz1&amp;id=6XeVWOBXO&amp;id=77jDjhI5C&amp;id=sH4hATTfN&amp;id=OemgIV2rY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI code generation: the gift that keeps on giving… technical debt</strong>: Engineering teams are reporting that AI-generated code, while compiling initially, often lacks architectural context and creates poor state management, leading to major refactoring or even complete rewrites. ChatGPT is apparently a master of generating ‘realistic’ but non-existent API elements, causing integration failures that only surface during review. So, while the hype machine claims AI saves developers 10+ hours a week, the reality is often a trade-off between initial speed and a mountain of silent technical debt. Maybe we should call it ‘AI-assisted job security for refactoring specialists.’ <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=DSeK6EEPJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA’s retail AI blueprints: democratizing AI or just more vendor lock-in?</strong>: NVIDIA is pushing ‘open-source AI blueprints’ for retail, like the Multi-Agent Intelligent Warehouse and Catalog Enrichment, claiming they’ll ‘democratize AI deployment’ and reduce costs. Meanwhile, their survey conveniently found 91% of companies are using or exploring AI and 90% are increasing budgets. It’s a classic move: release ‘open’ tools that just happen to run best on your hardware, then tout the ‘democratization’ while subtly locking everyone into your ecosystem. Nothing says ‘open’ like a proprietary hardware requirement, 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>
