<h2>Must know</h2>
<ul>
<li><strong>OpenAI introduces ads to free and Go tiers</strong>: OpenAI is rolling out targeted ads in the free and Go tiers of ChatGPT for US users, a move driven by significant financial losses and low paid subscriber numbers. While OpenAI assures ads won’t influence chatbot responses or compromise user data, privacy advocates, as noted by independent reporting, are wary of the potential for problematic incentives. This marks a strategic shift for the company, with CEO Sam Altman previously calling ads a “last resort” for monetization. <a href="https://app.daily.dev/posts/DQh9IAXNl" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mandiant releases Net-NTLMv1 rainbow tables</strong>: Mandiant has publicly released rainbow tables capable of cracking Net-NTLMv1 authentication hashes in under 12 hours using consumer hardware. This protocol has been deprecated as insecure for over two decades, yet remains prevalent. Mandiant’s release, hosted on Google Cloud, aims to compel organizations to finally disable Net-NTLMv1 and implement monitoring, as credential theft via this method is now “trivial.” <a href="https://app.daily.dev/posts/2q2phjvWS" 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. In parallel, Cloudflare acquired The Astro Technology Company, ensuring the Astro framework remains open-source and platform-agnostic. Both moves, according to company statements, aim to boost reliability and support the developer ecosystem. <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 firms accelerate healthcare expansion amid concerns</strong>: OpenAI acquired health startup Torch, Anthropic launched a healthcare-specific Claude model, and MergeLabs raised significant funding, signaling a rapid expansion of AI in healthcare despite concerns about misinformation and security vulnerabilities. <a href="https://app.daily.dev/posts/YNV6XIM0c" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google previews AI SQL generation in BigQuery</strong>: Google is rolling out ‘Comments to SQL’ in BigQuery, using Gemini AI to convert natural-language instructions in SQL comments into executable queries, aiming to accelerate data analysis. <a href="https://app.daily.dev/posts/5AcBUpc2i" 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 (lazy) loading for MCP tools, significantly 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>Nvidia DGX Spark offers local AI compute</strong>: Nvidia’s DGX Spark is a compact AI workstation delivering one petaFLOP, designed for AI developers needing local compute for prototyping and fine-tuning models up to 70B parameters without cloud dependencies. <a href="https://app.daily.dev/posts/Xk9eNeVDY" 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>Rails updates include error-aware retries and PostgreSQL mapping</strong>: Rails introduced error-aware retry strategies for ActiveJob, a public API for custom PostgreSQL type mappings, and enhanced security features like improved CSRF protection and FIPS-compliant checksums for ActiveStorage. <a href="https://app.daily.dev/feed-by-ids?id=hNvkDDcdl&amp;id=LkAtCVwbE" 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 the archive/zip library and Request.ParseForm, alongside ecosystem updates like enhanced struct pretty-printing and FFmpeg integration. <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>Cursor CEO claims AI built a browser in a week</strong>: Cursor’s CEO, Michael Truell, is touting that hundreds of GPT-5.2 agents generated over 3 million lines of code for a web browser in a week, making “meaningful progress.” However, the codebase reportedly doesn’t compile, has 34 errors and 94 warnings, and lacks any verifiable demo. Sounds like another classic case of AI hype outpacing actual, you know, working software. <a href="https://app.daily.dev/feed-by-ids?id=UbZlY7YDE&amp;id=qHCb0ov8q&amp;id=V3KHuMgUU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI chatbot saves insurance agents 3 minutes a day</strong>: Researchers developed Axlerod, an LLM-based chatbot, to assist insurance agents, claiming it saves a whopping 2.42 seconds per search, or about 3 minutes daily. At an annual cost of $252 for a Gemini Business seat, this translates to a thrilling $121 in labor savings. Apparently, the future of AI is saving us enough time to grab an extra coffee, while existing automation already handles the easy stuff. <a href="https://app.daily.dev/posts/p47qntiOJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Open source projects drowning in AI-generated PRs</strong>: Tldraw, an open-source project, is temporarily closing external pull requests because of an “influx of low-quality AI-generated contributions.” It seems AI is so good at coding now that it’s generating so much garbage, maintainers can’t tell the difference between a human and a bot. Maybe we should just let the AI review the AI’s code, what could possibly go wrong? <a href="https://app.daily.dev/posts/EuNu5at2M" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
