<h2>Must know</h2>
<ul>
<li><strong>OpenAI introduces ads and a new tier to boost revenue</strong>: OpenAI is rolling out targeted ads in its free and Go ($8/month) ChatGPT tiers for US users, a move CEO Sam Altman previously called a “last resort.” This comes as only 5% of users subscribe to paid plans and the company reported over $11.5 billion in losses in Q3 2025. While OpenAI assures ads won’t influence chatbot answers or sell user data, privacy advocates, as noted by independent reporting, warn of problematic incentives similar to social media, potentially eroding user trust. <a href="https://app.daily.dev/posts/DQh9IAXNl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI firms accelerate healthcare expansion amid major concerns</strong>: OpenAI acquired health startup Torch for $100 million, Anthropic launched a Claude AI model for healthcare, and MergeLabs raised $250 million, signaling a rapid expansion of AI in healthcare. However, this ‘gold rush’ brings significant risks, including AI model hallucination leading to medical misinformation and critical security vulnerabilities for sensitive patient data, according to sector reports. The industry is moving fast, but the potential for harm is substantial. <a href="https://app.daily.dev/posts/YNV6XIM0c" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mandiant releases Net-NTLMv1 rainbow tables to force deprecation</strong>: Mandiant has publicly released rainbow tables capable of cracking Net-NTLMv1 authentication hashes in under 12 hours on consumer hardware. This move aims to compel organizations to finally retire the 20-year-deprecated and insecure Microsoft protocol, which Mandiant describes as making credential theft ‘trivial.’ Security professionals are urged to disable Net-NTLMv1 via Group Policy and monitor for its continued use, as highlighted by technical press. <a href="https://app.daily.dev/posts/2q2phjvWS" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Google previews AI SQL generation in BigQuery</strong>: Google is testing ‘Comments to SQL’ in BigQuery, using Gemini AI to convert natural-language instructions in SQL comments into executable queries, aiming to speed up data analysis. <a href="https://app.daily.dev/posts/5AcBUpc2i" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare launches resilience initiative and acquires Astro</strong>: Cloudflare introduced ‘Code Orange: Fail Small’ to prevent outages through phased rollouts and acquired Astro Technology Company, keeping Astro open-source and platform-agnostic. <a href="https://app.daily.dev/feed-by-ids?id=OdNHxfcwg&amp;id=yRAOJZmiR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-driven browser development generates millions of lines of code</strong>: Cursor’s CEO claims a large-scale AI system, using hundreds of GPT-5.2 agents, generated over 3 million lines of code for a functional web browser in one week, though it lacks production-quality features like security. <a href="https://app.daily.dev/feed-by-ids?id=UbZlY7YDE&amp;id=qHCb0ov8q" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Claude Code implements dynamic tool loading for MCP servers</strong>: Claude Code introduced dynamic tool loading for MCP servers, reducing token usage by up to 95% and improving tool selection accuracy by only loading relevant tools on demand. <a href="https://app.daily.dev/feed-by-ids?id=WwZicivI3&amp;id=uq0isMFwF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LLM Compressor updates for AI model quantization</strong>: LLM Compressor 0.9.0 introduces refactored attention and KV cache quantization supporting multiple schemes (FP4, INT8, FP8), integrates Intel’s AutoRound, and adds experimental MXFP4 support. <a href="https://app.daily.dev/feed-by-ids?id=Y9B2mh3yd&amp;id=YCoHMIgPO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI and non-human identity management in cloud security</strong>: AI-driven solutions are enhancing non-human identity (NHI) management in cloud security by providing predictive capabilities and adaptive monitoring for machine credentials, though oversight is crucial to prevent new vulnerabilities. <a href="https://app.daily.dev/feed-by-ids?id=lq3eNhRBB&amp;id=rc0IgUSh5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Go 1.26 RC2 released with security updates</strong>: Go 1.26 Release Candidate 2 is out, bringing security fixes for <code>archive/zip</code> and <code>Request.ParseForm</code>, with similar updates also released for Go 1.25.6 and Go 1.24.12. <a href="https://app.daily.dev/posts/DFYbtJ1rQ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI coding hype vs. reality</strong>: Cursor’s CEO is touting an AI-generated web browser with ‘meaningful progress’ after AI agents supposedly wrote millions of lines of code in a week. Yet, the codebase doesn’t compile, has dozens of errors, and there’s no working demo. It seems ‘meaningful progress’ now means ‘we got AI to spit out a lot of broken text,’ which is a bold new definition of success. <a href="https://app.daily.dev/feed-by-ids?id=7Sqt9u8B2&amp;id=V3KHuMgUU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The AI investment pendulum is about to swing back, apparently</strong>: Some ‘experts’ are claiming the AI investment bubble is worse than the dot-com era, with VCs pouring billions into startups with negative margins and no path to profitability. They say CES 2026 was full of vaporware and basic chatbots rebranded as ‘revolutionary AI agents.’ So, basically, the same old story, just with more buzzwords and bigger numbers. Who could have predicted that? <a href="https://app.daily.dev/feed-by-ids?id=dosALRqYx&amp;id=6a57VDcl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI will fix the talent landscape, or so they say</strong>: The job market is apparently a mess, but don’t worry, AI-powered skills-based matching is here to save us all by moving beyond degrees and job titles. LinkedIn is using LLMs to infer skills and make career pivots visible. Because nothing says ‘unbiased’ and ‘fair’ like an algorithm trained on existing data, which we all know is perfectly equitable. What could possibly go wrong? <a href="https://app.daily.dev/posts/ERw5pwvMm" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
