<h2>Must know</h2>
<ul>
<li><strong>Mandiant exposes legacy Net-NTLMv1 vulnerabilities</strong>: Mandiant has released rainbow tables that can crack Net-NTLMv1 authentication hashes in under 12 hours on consumer hardware. This protocol has been deprecated for over 20 years due to its insecurity, yet remains prevalent in many organizations. Mandiant’s move aims to push companies to finally disable Net-NTLMv1 and implement monitoring, as credential theft via this method is now ‘trivial,’ according to their report. <a href="https://app.daily.dev/posts/2q2phjvWS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI introduces targeted ads in ChatGPT free and Go tiers</strong>: OpenAI is rolling out targeted advertising for US users on its free and Go ($8/month) ChatGPT tiers, with ads appearing at the bottom of conversations. This strategic shift, despite CEO Sam Altman previously calling ads a ‘last resort,’ is driven by financial strain, with OpenAI reporting over $11.5 billion in Q3 2025 losses. The company assures ads won’t influence chatbot answers or sell user data, but privacy advocates warn of problematic incentives, as reported by TechCrunch. <a href="https://app.daily.dev/posts/DQh9IAXNl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare launches resilience initiative and acquires Astro</strong>: Following two major global outages in late 2025, Cloudflare has launched ‘Code Orange: Fail Small,’ a resilience initiative mandating phased configuration rollouts and automatic rollbacks to prevent cascading failures. Concurrently, Cloudflare acquired The Astro Technology Company, an open-source framework, allowing Astro to focus on framework improvement without monetization pressures. Both moves aim to enhance reliability and support the developer ecosystem, as detailed in Cloudflare’s official announcements. <a href="https://app.daily.dev/feed-by-ids?id=OdNHxfcwg&amp;id=yRAOJZmiR" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI-generated code: the good, the bad, and the ugly</strong>: While AI tools like GitHub Copilot and Claude Code can accelerate development, they often lack architectural intent, leading to code quality issues and the need for careful context management and upfront design decisions, as highlighted by SoftwareMill and Rebeloper. <a href="https://app.daily.dev/feed-by-ids?id=SRm19EN1b&amp;id=nLHgst6Y1&amp;id=S37WCeeOE" target="_blank" rel="noopener nofollow">Read more</a></li>
<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>Open source projects struggle with AI-generated PRs</strong>: Tldraw, an open-source project, is temporarily closing external pull requests due to an influx of low-quality AI-generated contributions, highlighting a growing challenge for maintainers. <a href="https://app.daily.dev/posts/EuNu5at2M" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LLM Compressor 0.9.0 enhances model quantization</strong>: LLM Compressor 0.9.0 introduces refactored attention and KV cache quantization, supporting multiple schemes like FP4, INT8, and FP8, along with Intel’s AutoRound algorithm and improved MoE model calibration. <a href="https://app.daily.dev/posts/Y9B2mh3yd" 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 archive/zip and Request.ParseForm, with similar updates also applied to 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>
<li><strong>AMD EPYC 8004 ‘Siena’ CPUs improve with new Linux stacks</strong>: AMD EPYC 8004 ‘Siena’ processors show significant performance gains on newer Linux distributions like Ubuntu 25.10 with Linux 6.18 LTS and GCC 15.2, benefiting cost-optimized server deployments. <a href="https://app.daily.dev/posts/85IZs9OZ7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kafka 4.0 ends ZooKeeper support, Confluent offers migration tool</strong>: Apache Kafka 4.0 will drop ZooKeeper support, requiring migration to KRaft-based clusters, for which Confluent has released Kafka Copy Paste (KCP), an open-source tool to simplify the transition to Confluent Cloud. <a href="https://app.daily.dev/posts/fOVQvObt4" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI-driven browser development: 3 million lines of code, zero working product</strong>: Cursor CEO Michael Truell proudly announced that hundreds of GPT-5.2 agents generated over 3 million lines of code for a web browser in a week. The catch? The codebase doesn’t compile, has 34 errors and 94 warnings, and there’s no working demo. Apparently, ‘meaningful progress’ now means generating a mountain of non-functional code, as reported by V42saD54U. Who needs a working product when you have a great story? <a href="https://app.daily.dev/feed-by-ids?id=UbZlY7YDE&amp;id=qHCb0ov8q&amp;id=V3KHuMgUU&amp;id=V42saD54U" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI chatbot saves insurance agents a whopping 3 minutes a day</strong>: Researchers developed an LLM-based chatbot, Axlerod, to help insurance agents, claiming it saves them an average of 2.42 seconds per search, totaling about 3 minutes daily. For a system costing $252 annually per user, this translates to a thrilling $121 in labor savings. The Register points out that existing automation already handles most routine queries, leaving this AI to tackle the ‘complex’ task of saving pennies. <a href="https://app.daily.dev/posts/p47qntiOJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI healthcare gold rush: because what could go wrong?</strong>: OpenAI acquired health startup Torch, Anthropic launched a healthcare-specific Claude model, and MergeLabs raised $250 million, all within a week. This ‘AI healthcare gold rush’ is apparently booming, despite ‘major concerns’ about AI hallucinating in clinical settings and security vulnerabilities with sensitive patient data. Because nothing says ‘trustworthy medical advice’ like an AI that occasionally makes things up, as noted by YNV6XIM0c. <a href="https://app.daily.dev/posts/YNV6XIM0c" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
