<h2>Must know</h2>
<ul>
<li><strong>AI coding tools: the good, the bad, and the ugly</strong>: Salesforce deployed Cursor AI to 20,000 engineers, reporting double-digit gains in cycle time and code quality, with 90% adoption in months. However, a CMU study on 807 GitHub repos found AI coding assistants like Cursor increased initial productivity by 30-40% but this vanished by month three, replaced by a 30% rise in static analysis warnings and over 40% increase in code complexity. This suggests AI tools are great for initial boosts but can degrade code quality long-term, creating ‘context collapse’ that might even worsen future AI model training. Meanwhile, OpenAI went all-in on the Rust TUI framework Ratatouille, using it for their Codex, showing that even the big players are leveraging open-source for their AI tooling. <a href="https://app.daily.dev/feed-by-ids?id=7KqSlBkW9&amp;id=EZUXhYwRZ&amp;id=a6juBspr2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents are changing developer roles, but not replacing them</strong>: The rise of AI coding agents is shifting developer roles from writing code to overseeing AI, much like VCs manage portfolios. This means strong programming fundamentals, architectural skills, and critical thinking are more important than ever. Junior developers face higher entry barriers as AI handles entry-level tasks, while experienced engineers must adapt to managing and testing AI-generated code. This transformation requires a ‘VC mindset’ to direct multiple AI agents, emphasizing the need for rigorous testing and new standards to ensure safety and reliability, as reported by Code to the Moon. <a href="https://app.daily.dev/feed-by-ids?id=BJsnyHclD&amp;id=HWjGJHh6I" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare patches WAF bypass, Fortinet targeted by SSO exploits</strong>: Cloudflare fixed a critical vulnerability in its Web Application Firewall (WAF) ACME validation logic that allowed attackers to bypass security rules and access origin servers directly. The flaw, reported via their bug bounty program, was due to improper validation of ACME challenge tokens. Separately, Arctic Wolf observed an automated cyberattack campaign targeting Fortinet FortiGate devices, exploiting SSO accounts to make unauthorized firewall changes, create persistence, and exfiltrate data. These attacks closely resemble a December 2025 campaign exploiting SSO authentication bypass vulnerabilities, highlighting ongoing threats to critical network infrastructure. <a href="https://app.daily.dev/feed-by-ids?id=PoZiSIcPJ&amp;id=R9RchEmNx" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Amazon launches Bedrock AgentCore for AI agent deployment</strong>: Amazon introduced Bedrock AgentCore, a platform for deploying and operating AI agents with zero infrastructure management, offering modular services like Runtime, Memory, Gateway, Code Interpreter, and Observability, along with a CLI toolkit for rapid creation and integration with frameworks like LangGraph and CrewAI. <a href="https://app.daily.dev/posts/PbCI8QH1q" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks scales feature flagging to 300M evaluations/sec</strong>: Databricks developed SAFE, an in-house feature flagging platform handling over 300 million evaluations per second with ~10μs p95 latency, managing 25,000 active flags across hundreds of services by decoupling code deployment from feature enablement. <a href="https://app.daily.dev/posts/EfIBAJOrr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Linux patch boosts NVMe performance by 15%</strong>: Intel engineers released a Linux kernel patch that improves NVMe storage performance by up to 15% on high core count processors by optimizing IRQ affinity and grouping CPUs by cluster within NUMA domains. <a href="https://app.daily.dev/posts/CI1vi8cFn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PyTorch 2.10 expands AMD ROCm and Intel GPU support</strong>: PyTorch 2.10 was released, primarily enhancing support and performance for AMD ROCm and Intel GPUs to improve hardware compatibility in the deep learning ecosystem. <a href="https://app.daily.dev/posts/qtIrJcChn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Shadcn Studio MCP: AI-powered UI builder for developers</strong>: Shadcn Studio MCP is an AI-powered tool that integrates with IDEs to generate production-ready UI components, blocks, and pages using shadcn/ui for React, Next.js, Vite, and Laravel, offering lifetime access and unlimited project usage. <a href="https://app.daily.dev/feed-by-ids?id=84wxnOe0I&amp;id=ncQILOQ9m" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Curiosum sponsors Hologram Elixir framework</strong>: Curiosum is now the main sponsor of Hologram, an Elixir framework that transpiles Elixir code to JavaScript for browser-side execution, aiming to accelerate its development and adoption for pure Elixir end-to-end applications. <a href="https://app.daily.dev/feed-by-ids?id=ePz5bRXJH&amp;id=jyQtn5H4O" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>jQuery 4.0 drops legacy browser support, adds security</strong>: jQuery 4.0 has been released, dropping support for Internet Explorer 10 and Microsoft Edge Legacy, integrating Trusted Types and Content Security Policy, and migrating to ES modules, resulting in a 3KB size reduction. <a href="https://app.daily.dev/posts/uMYmsakGM" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI will eliminate immigration, says Palantir CEO</strong>: Palantir CEO Alex Karp, speaking at the World Economic Forum, boldly declared that AI will eliminate the need for large-scale immigration in Western economies, except for specialized skills. This visionary insight comes despite industry reports from Deutsche Bank, Gartner, and PwC all signaling that AI’s ‘honeymoon’ is over, with most CEOs seeing ‘zero payoffs’ from their AI investments. But hey, why let pesky facts get in the way of a good narrative, especially when you’re selling AI solutions? <a href="https://app.daily.dev/posts/3s1vrBGbY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI world models: the next big hallucination?</strong>: Apparently, AI ‘world models’ are the future, promising content creation speed by accepting ‘reduced compute efficiency’ and ‘learned hallucinations.’ So, we’re trading the control and efficiency of traditional game engines for faster asset generation that might just make things up. It’s like saying, ‘Why bother with reality when you can have a really fast, slightly unhinged version of it?’ <a href="https://app.daily.dev/posts/OIj9HQhHA" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ChatGPT UI replication: looks good, breaks fast</strong>: ChatGPT can now replicate Apple demo UIs visually, which sounds impressive until you realize it completely misses the underlying system constraints. The generated implementations ‘break under performance load’ because ChatGPT lacks ‘deep framework comprehension.’ So, it’s great at making things look pretty, but don’t expect them to actually work. Maybe it’s just trying to teach us a lesson about judging books by their covers, or UIs by their pixels. <a href="https://app.daily.dev/posts/LQOYGhWvV" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
