<h2>Must know</h2>
<ul>
<li><strong>TypeScript 6 released as a stepping stone to Go rewrite</strong>: TypeScript 6 has been released, primarily as a transitional step to prepare for a complete rewrite of the compiler in Go, aiming to address significant performance bottlenecks in large applications. Developers are strongly encouraged to upgrade to TypeScript 6 to ensure compatibility with the upcoming TypeScript 7, which will feature the Go-based compiler. Key changes include enabling strict mode by default and updating module resolution to ES modules, as reported by Bruno Couriol and others. <a href="https://app.daily.dev/posts/tiMDdyFcG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare R2 introduces local uploads for faster writes</strong>: Cloudflare has launched “Local Uploads” for its R2 object storage, now in open beta. This feature significantly reduces cross-region write latency by allowing data to be written to the nearest Cloudflare PoP and then asynchronously replicated to the bucket’s home region. Objects become immediately accessible after the initial write, without waiting for full replication, which is a game-changer for geographically distributed workloads, according to Renato Losio. <a href="https://app.daily.dev/posts/XbGrpoJDs" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI unveils AI-driven harness engineering</strong>: OpenAI has introduced “Harness Engineering,” an internal AI-driven methodology that uses Codex agents to automate large-scale software development tasks like code generation, testing, and CI workflows. A five-month internal experiment reportedly resulted in a beta product with a million lines of code generated without manual writing. This shifts engineers’ roles from coding to designing environments and providing structured feedback, as noted by Martin Fowler. <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>Linux 7.0 and desktop environments see updates</strong>: Linux 7.0 is getting initial DRM kernel graphics driver fixes for AMDGPU and renewed patch activity for eCryptfs, while GNOME 50 beta includes variable refresh rate and fractional scaling. <a href="https://app.daily.dev/feed-by-ids?id=2U7gOjLpG&amp;id=BZn5Av50T&amp;id=LutMi5Kcz&amp;id=MGUcuQgst&amp;id=gNaBG2osm&amp;id=u9gGdIsNO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AWS enables event-driven architectures for RDS SQL server</strong>: AWS now enables Lambda function triggers directly from RDS for SQL Server database events, allowing for event-driven architectures using stored procedures and CloudWatch Logs. <a href="https://app.daily.dev/posts/KzhIuYzeh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic and Figma integrate claude code for design workflow</strong>: Anthropic and Figma have integrated Claude Code to convert production code into editable Figma designs, streamlining design-to-code and code-to-design workflows. <a href="https://app.daily.dev/posts/aCaE5xJ9z" 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 has been released, featuring runtime-level async improvements, Zstandard compression support, and new C# 15 language features. <a href="https://app.daily.dev/feed-by-ids?id=QBAQQNpLN&amp;id=xPZZ4nMO0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GrapheneOS offers privacy-focused android alternative</strong>: GrapheneOS continues to be a strong privacy-focused Android alternative, offering a de-Googled experience with options for sandboxed Google Play, despite some trade-offs. <a href="https://app.daily.dev/feed-by-ids?id=840Dw8MXb&amp;id=OwmBaq9Nm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NTransformer boosts local LLM inference speed</strong>: NTransformer, a C++/CUDA LLM inference engine, claims an 83x speedup for Llama 3.1 70B on a single RTX 3090, using adaptive caching and layer skipping. <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><a href="http://Lobste.rs" target="_blank" rel="noopener nofollow">Lobste.rs</a> migration to SQLite hits performance snag</strong>: <a href="http://Lobste.rs" target="_blank" rel="noopener nofollow">Lobste.rs</a> attempted a live migration from MariaDB to SQLite but quickly reverted due to severe performance degradation, with plans to retry after addressing the issues. <a href="https://app.daily.dev/posts/RLQeRSeS0" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Altman hand-waves AI energy concerns</strong>: OpenAI CEO Sam Altman is out here dismissing claims of high AI water usage as “completely untrue,” while simultaneously acknowledging total energy consumption is a “legitimate concern.” He then tries to justify it by saying training a human takes “enormous resources.” So, AI is energy-efficient, but also needs nuclear power, and also humans are expensive. Got it. <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>Gemini 3.1 pro: benchmark king, real-world flop</strong>: Google’s Gemini 3.1 Pro is apparently a benchmark champion, scoring 100% on a “skateboarding spatial reasoning benchmark.” Yet, in the real world, it’s plagued by unreliable tool calling, looping, and hallucinated packages. So, it’s great at skateboarding, but can’t tie its own shoes. Classic Google, benchmaxing at the expense of actual usability. <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>AI agent security is a dumpster fire</strong>: A BBC reporter got hacked via an AI vibe coding platform called Orchards, which claims a million users and lists Google, Uber, and Amazon as clients. Meanwhile, OpenClaw, an open-source AI agent, had 512 vulnerabilities, including a critical RCE. But don’t worry, Singapore launched a governance framework! Because regulations always stop the actual hacks, right? <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>
</ul>
