<h2>Must know</h2>
<ul>
<li><strong>AI agent security is a dumpster fire</strong>: Recent incidents highlight catastrophic security flaws in AI agents. A BBC reporter was hacked via an AI coding platform, and an open-source agent, OpenClaw, revealed 512 vulnerabilities, including critical RCE. This ‘lethal trifecta’ of data access, untrusted content, and external communication means traditional security models are woefully inadequate, as reported by multiple sources. <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>TypeScript 6 prepares for a Go rewrite</strong>: TypeScript 6 has been released as a transitional version, primarily to pave the way for a complete rewrite of the compiler in Go, aiming to tackle significant performance bottlenecks. 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. <a href="https://app.daily.dev/posts/tiMDdyFcG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI introduces harness engineering</strong>: OpenAI has unveiled ‘Harness Engineering,’ an internal AI-driven methodology that uses Codex agents to automate large-scale software development tasks like code generation and testing. This approach shifts engineers’ roles to designing environments and specifying intent, with OpenAI claiming a beta product of one million lines of code generated without manual writing. <a href="https://app.daily.dev/posts/dPCNsAxds" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Cloudflare R2 local uploads reduce latency</strong>: Cloudflare launched ‘Local Uploads’ for R2 object storage, 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 function triggers directly from RDS for SQL Server database events, enabling event-driven architectures by parsing ERRORLOG messages to invoke Lambda functions via CloudWatch and SNS. <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 C++/CUDA engine, claims an 83x speedup for Llama 3.1 70B on a single RTX 3090, using adaptive caching and layer skipping to enable 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>Sandkasten offers self-hosted AI agent sandboxes</strong>: Sandkasten provides self-hosted Linux sandboxes for AI agents using native Linux primitives for isolation, offering persistent shell sessions and integration with OpenAI Agents SDK and LangChain. <a href="https://app.daily.dev/posts/kdJWfknAe" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Go structured concurrency patterns</strong>: While Go’s <code>go</code> statement is unstructured, patterns like <code>errgroup</code> and <code>WaitGroup</code> can achieve structured concurrency, though goroutines require explicit <code>ctx.Done()</code> checks for cancellation unlike Python or Kotlin. <a href="https://app.daily.dev/posts/2zOi0Da7K" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ward is a new Laravel security scanner in Go</strong>: Ward, an open-source security scanner written in Go, performs static analysis on Laravel projects to report vulnerabilities, supporting custom rules and offering a terminal UI for findings. <a href="https://app.daily.dev/posts/dji7is7Et" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>.NET 11 Preview 1 brings async improvements and Zstandard</strong>: .NET 11 Preview 1 introduces 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/feed-by-ids?id=QBAQQNpLN&amp;id=xPZZ4nMO0" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic and Figma integrate Claude Code for design-to-code</strong>: Anthropic and Figma are touting a new workflow where Claude Code converts production code into editable Figma designs and back again. Because apparently, the ‘simple prompt’ to turn live pages into Figma files is going to magically streamline developer-designer workflows, rather than just adding another layer of abstraction to debug when things inevitably go sideways. It’s always ‘simple’ until it’s not. <a href="https://app.daily.dev/posts/aCaE5xJ9z" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google VP warns about AI startup viability</strong>: Google Cloud VP Darren Mowry is warning that AI startups focused on LLM wrappers or aggregators are under threat, drawing parallels to early AWS resellers. So, the big cloud providers are now telling us that building on their platforms without ‘deep differentiation’ is a bad idea? Shocking. It’s almost like they want all the margin for themselves, while simultaneously being ‘bullish’ on ‘vibe coding platforms.’ Pick a lane, Google. <a href="https://app.daily.dev/posts/djwfpUzVm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI’s financial woes and Gemini’s ‘benchmaxing’</strong>: OpenAI is reportedly facing a ‘dire financial situation’ with projected $44 billion in losses and investors pulling back, leading to ‘financial desperation’ ads in ChatGPT. Meanwhile, Google’s Gemini 3.1 Pro is ‘benchmaxing’ with top scores but is plagued by ‘unreliable tool calling’ and ‘looping behavior’ in the real world. It seems the AI giants are either running out of money or releasing models that look good on paper but can’t tie their own shoes. Peak innovation, folks. <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>
</ul>
