<h2>Must know</h2>
<ul>
<li><strong>AI agent security is a dumpster fire</strong>: Multiple reports highlight catastrophic security flaws in AI agents. A BBC reporter was hacked via an AI coding platform, and an open-source agent, OpenClaw, had 512 vulnerabilities, including critical remote code execution. Even a financial AI project contained a backdoor. This ‘lethal trifecta’ of data access, untrusted content, and external communication means traditional security models are completely inadequate, as sources like WIRED and the BBC point out. <a href="https://app.daily.dev/feed-by-ids?id=1HpSP7L9S&amp;id=1TelNKsBD&amp;id=9xLrRFK2b&amp;id=FuKqPJxtz&amp;id=Jm3weKMWA&amp;id=ZmcjIu8ai&amp;id=em7afQ5sm&amp;id=kdJWfknAe&amp;id=ohhVPexbM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI’s financial woes and Gemini’s practical problems</strong>: OpenAI is reportedly facing significant financial challenges, projecting $44 billion in losses before 2029 and seeing key investors and talent pull back, with some analysts predicting they could run out of money by 2027. Meanwhile, Google’s Gemini 3.1 Pro, despite benchmark dominance, is plagued by real-world usability issues like unreliable tool calling and looping behavior, suggesting Google is ‘benchmaxing’ at the expense of practical application, according to one critical analysis. <a href="https://app.daily.dev/feed-by-ids?id=95MvV75SN&amp;id=mGa3eARS3&amp;id=ogUHau6h0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>TypeScript 6 prepares for a Go rewrite</strong>: TypeScript 6 has been released as a transitional update, primarily to pave the way for a complete rewrite of its compiler in Go, aiming to fix severe performance bottlenecks in large applications. Developers are strongly encouraged to upgrade and resolve deprecation warnings, as this release enables strict mode by default and updates module resolution to ES modules, preparing for the more performant TypeScript 7. <a href="https://app.daily.dev/posts/tiMDdyFcG" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Anthropic and Figma integrate for design-to-code workflow</strong>: Anthropic and Figma have partnered to allow Claude Code to convert production code into editable Figma designs and round-trip updates back to the codebase, streamlining designer-developer collaboration. <a href="https://app.daily.dev/posts/aCaE5xJ9z" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare R2 launches local uploads for faster writes</strong>: Cloudflare’s R2 object storage now offers ‘Local Uploads’ in beta, reducing cross-region write latency by up to 75% by writing data to the nearest PoP and asynchronously replicating it to the home region. <a href="https://app.daily.dev/posts/XbGrpoJDs" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AWS enables event-driven architectures for RDS SQL Server</strong>: AWS now allows Lambda functions to be triggered directly from RDS for SQL Server database events, enabling event-driven architectures by parsing SQL Server’s ERRORLOG for specific messages. <a href="https://app.daily.dev/posts/KzhIuYzeh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NTransformer speeds up local LLM inference</strong>: NTransformer, a new C++/CUDA engine, claims an 83x speedup for Llama 3.1 70B on a single RTX 3090, using adaptive caching and layer skipping for efficient local LLM inference. <a href="https://app.daily.dev/feed-by-ids?id=Ahx4zgQks&amp;id=DyVC0qyKG&amp;id=Fkac7nXWM&amp;id=K01E44V9h&amp;id=WoJMuyqAQ&amp;id=mulSXvfKt" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI introduces harness engineering for AI-driven development</strong>: OpenAI unveiled ‘Harness Engineering,’ an internal AI-driven methodology using Codex agents to automate code generation, testing, and CI, shifting engineers’ roles to designing environments and providing feedback. <a href="https://app.daily.dev/posts/dPCNsAxds" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI data centers explore high-temperature superconductors</strong>: Hyperscalers like Microsoft are investing in high-temperature superconductors (HTS) for data centers to address power constraints, as HTS cables offer near-zero resistance and higher capacity than copper. <a href="https://app.daily.dev/posts/jbBqj45d9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>.NET 11 Preview 1 released with async improvements and C# 15</strong>: .NET 11 Preview 1 is out, featuring runtime-level async improvements, Zstandard compression support, and new C# 15 language features, signaling the direction for the next major .NET release. <a href="https://app.daily.dev/posts/QBAQQNpLN" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Unity CEO says AI will make entire games</strong>: Unity’s CEO is apparently going to unveil a beta that lets developers prompt full casual games into existence with natural language. Because, clearly, the world is clamoring for more Flappy Bird clones generated by an LLM. While rapid prototyping is cool, let’s not pretend AI is replacing actual game designers anytime soon; quality, polish, and design cohesion still require human brains, not just fancy prompts. <a href="https://app.daily.dev/posts/VkFal2xDM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is exposing design’s craft crisis</strong>: According to some, AI tools aren’t creating a design crisis; they’re just exposing how many designers can’t evaluate AI output, spot accessibility issues, or understand technical trade-offs because the industry decided they don’t need to code. So, instead of learning how things actually work, we’re just going to blame the tools for revealing the emperor has no clothes. Classic. <a href="https://app.daily.dev/posts/44OJ4cBTv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google VP warns about AI startup viability</strong>: Google Cloud’s VP is warning that AI startups built on ‘thin IP’ like LLM wrappers or aggregators are doomed, comparing them to early AWS resellers who got squeezed out. He’s bullish on ‘vibe coding platforms,’ though. So, basically, if your AI startup isn’t doing something truly innovative, or if it’s just a fancy wrapper, you’re probably toast. But don’t worry, there’s always ‘vibe coding’ to save the day. <a href="https://app.daily.dev/posts/djwfpUzVm" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
