<h2>Must know</h2>
<ul>
<li>Cloudflare reimplements Next.js on Vite to kill deployment friction: Cloudflare has unveiled ‘vinext’, an experimental project that bypasses Turbopack to run the Next.js API surface on Vite. Developed in just one week using AI, it claims to pass 94% of Next.js tests while delivering 4x faster builds and 57% smaller client bundles. This is a direct shot at Vercel’s perceived platform lock-in, aiming to make Next.js truly portable across non-Vercel environments. <a href="https://app.daily.dev/posts/kuvbBXUUR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Alibaba releases Qwen 3.5 with Sonnet 4.5 performance for local GPUs: The new Qwen3.5-35B-A3B model uses a hybrid Gated Delta Network and MoE architecture to activate only 3B parameters per token, making it highly efficient for consumer hardware. It reportedly surpasses GPT-5-mini and Claude Sonnet 4.5 in visual reasoning and knowledge benchmarks while supporting a 1-million-token context window. The model is open-sourced under Apache 2.0, providing a high-performance alternative for self-hosted agentic workflows. <a href="https://app.daily.dev/posts/DZM9ficRE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>NVIDIA pushes AI-native 6G and autonomous telco stacks: Ahead of MWC 2026, NVIDIA announced the Aerial Omniverse Digital Twin for simulating 6G networks and a 30B-parameter ‘Large Telco Model’ for automated fault remediation. Partners like T-Mobile and SoftBank are already moving AI-RAN from labs to field trials, achieving 36 Gbps throughput on GH200 servers. This shift signals the transition of RAN from fixed hardware to software-defined, AI-orchestrated workloads. <a href="https://app.daily.dev/feed-by-ids?id=N9cHfOw6g&amp;id=central_news_id_2%2C+%22pf0UFZRkY%22%0A++++++++++%5D%0A++++++++%7D%0A++++++%5D%0A++++%7D%2C%0A++++%7B%0A++++++%22title%22%3A+%22Good+to+know%22%2C%0A++++++%22items%22%3A+%5B%0A++++++++%7B%0A++++++++++%22title%22%3A+%22Linux+7.0-rc2+arrives+with+unusually+large+filesystem+and+BPF+updates%22%2C%0A++++++++++%22body%22%3A&amp;id=Linus+Torvalds+noted+the+release+is+bulkier+than+usual+due+to+pent-up+work+from+the+extended+6.19+cycle%2C+with+a+stable+release+expected+in+April.&amp;id=post_ids%22%3A+%5B%0A++++++++++++%221hIBgMnIM" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Timber AOT compiler converts ML models to native C99 for 336x speedups: By eliminating the Python runtime, Timber achieves microsecond latency for XGBoost and scikit-learn models on edge devices. <a href="https://app.daily.dev/posts/xW6jc9SaR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Argo CD 3.3 introduces PreDelete hooks and shallow cloning: The update enables safer GitOps teardowns and significantly faster repository fetching for large monorepos. <a href="https://app.daily.dev/posts/J931yCKTd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Systemd v260 officially drops support for SysV init scripts: This marks a final shift in Linux init system dependencies, forcing a move to native service units. <a href="https://app.daily.dev/posts/T0tOfwEhy" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Numpy-ts ports 94 percent of the NumPy API to TypeScript: The library is zero-dependency and tree-shakeable, bringing type-safe scientific computing to Node.js and the browser. <a href="https://app.daily.dev/posts/S646eUkOf" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>PHP-JS extension embeds Google V8 engine directly into PHP: This allows bidirectional variable mapping and server-side JavaScript execution within existing PHP environments. <a href="https://app.daily.dev/posts/zAEWLZlNT" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Android 17 Beta 1 optimizes UI performance with MessageQueue rewrite: The update focuses on reducing missed frames and introduces stricter background audio focus restrictions. <a href="https://app.daily.dev/posts/INaqwNVxY" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li>The Model Context Protocol is just overengineered piping: Critics on Hacker News are correctly pointing out that MCP is a solution looking for a problem, adding flaky initialization and auth headaches to workflows that a simple CLI handles better. Why build a specialized server for an LLM when they are already trained to use the composable, debuggable tools we’ve used for decades? It’s another layer of abstraction designed to lock you into a specific ‘agentic’ ecosystem. <a href="https://app.daily.dev/posts/lMlswjQmz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>OpenAI’s flexible ethics on military contracts: While Sam Altman calls the Pentagon’s blacklisting of Anthropic a ‘scary precedent,’ OpenAI reportedly signed a Department of War deal that conveniently omitted the specific autonomous weapon safeguards Anthropic was fighting for. It’s a classic move: publicly signal support for safety while privately accepting the contract terms your competitor refused on principle. Apparently, ‘red lines’ are more like ‘suggestions’ when a massive government contract is on the table. <a href="https://app.daily.dev/feed-by-ids?id=xURdvgDug&amp;id=VNmdoI3Re" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>The AWS SDK for .NET is a million-violation dumpster fire: A recent audit of the AWS SDK revealed nearly a million code quality violations, including guaranteed memory leaks and thousands of missing unit tests. For a service that powers half the internet, the SDK’s internal state is a masterclass in ‘move fast and break things’ technical debt. It’s a sobering reminder that just because a library is ‘official’ doesn’t mean it isn’t held together by duct tape and swallowed exceptions. <a href="https://app.daily.dev/posts/mIeiA4aYa" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>