<h2>Must know</h2>
<ul>
<li><strong>Next.js 16.2 and the rise of TanStack Start</strong>: Next.js 16.2 has landed with a 400% faster dev server and a move from JSON.parse to recursive walks for 60% faster HTML rendering. Interestingly, some senior engineers are already jumping ship to TanStack Start (React 19 + Vite), citing doubled shipping velocity after escaping Next.js abstraction overhead. Vercel is leaning into the ‘agentic’ future by including an <a href="http://agents.md" target="_blank" rel="noopener nofollow">agents.md</a> context file by default in all new projects. <a href="https://app.daily.dev/feed-by-ids?id=pLH5NluEB&amp;id=4nwi9O4T1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI acquires Astral to swallow the Python ecosystem</strong>: OpenAI is acquiring Astral, the team behind the high-performance Rust-based Python tools uv and Ruff. While OpenAI promises to maintain these as open-source, the move strategically integrates critical developer infrastructure directly into the Codex ecosystem. This mirrors Anthropic’s acquisition of Bun, signaling a land grab where AI giants own the very compilers and package managers developers use daily. <a href="https://app.daily.dev/feed-by-ids?id=lqMrRsYud&amp;id=G7iZQZ26W" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cursor releases Composer 2 to challenge Claude Opus 4.6</strong>: Cursor has launched Composer 2, an in-house model featuring ‘self-summarization’ that compresses its own context window to 1,000 tokens to minimize errors. It reportedly edges out Anthropic’s Claude Opus 4.6 on Terminal-Bench 2.0 while being significantly cheaper at $0.50 per million input tokens. This release marks Cursor’s third proprietary model in five months, focusing on long-horizon coding tasks that require hundreds of sequential actions. <a href="https://app.daily.dev/posts/mIKxoNmwG" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>LangChain launches Open SWE for sandboxed coding agents</strong>: An MIT-licensed framework that codifies agent architectures from Stripe and Ramp, running tasks in isolated cloud sandboxes with full shell access. <a href="https://app.daily.dev/posts/eZFXj4R96" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google Stitch 2.0 pivots to AI native design canvas</strong>: The experimental tool now features an ‘agent manager’ for parallel UI exploration and exports <a href="http://design.md" target="_blank" rel="noopener nofollow">design.md</a> rules for Claude Code and Cursor integration. <a href="https://app.daily.dev/posts/Ke9ZMEd7T" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Pulumi OPA support reaches stable 1.1.0</strong>: Now features full parity with TypeScript/Python SDKs and a Gatekeeper compatibility mode to shift Kubernetes policy enforcement left into the preview stage. <a href="https://app.daily.dev/posts/2G4tIOqfZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Durable migrates 3 million users to Vercel with 6 engineers</strong>: The AI business builder used coding agents to rewrite their entire platform, processing 360 billion tokens annually while cutting infra costs by 4x. <a href="https://app.daily.dev/posts/zNuNcsVh2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Monty is a Python interpreter in Rust for AI safety</strong>: A purpose-built, sandboxed interpreter with 6µs cold starts designed to execute LLM-generated code without the overhead of full Python environments. <a href="https://app.daily.dev/posts/ZfyVdsLNL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>React 19 patterns fix async await caching bugs</strong>: New community patterns for shared async task caches are emerging to prevent duplicate network requests when multiple views hit the same resource. <a href="https://app.daily.dev/posts/jlHdGWXDO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>VitePress documentation theme comes to Jekyll</strong>: A new gem brings the polished VitePress UI, including local search and dark mode, to Ruby-based documentation sites without requiring Node.js. <a href="https://app.daily.dev/posts/x8ytGnN9G" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The vibe coding productivity mirage</strong>: Cisco executives are bragging about porting 500,000 lines of 15-year-old code to Go in a week using ‘vibe coding.’ While they claim it’s the future, they conveniently mention it still needs ‘extensive validation’ before production, which is corporate speak for ‘the AI hallucinated half the business logic and we’re terrified to deploy it.’ <a href="https://app.daily.dev/feed-by-ids?id=0hn0P8EbP&amp;id=LJipAqA3r" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Context rot is the new bit rot</strong>: Researchers have quantified ‘context rot,’ proving that 60% of facts are irrecoverable after a single compression pass in frontier models. Despite the marketing hype around 2M+ token windows, your expensive LLM is silently violating constraints with full confidence after just a few turns of conversation. <a href="https://app.daily.dev/posts/X2JhFEKeS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PwC threatens staff with AI or the door</strong>: PwC’s US CEO claims senior staff who aren’t ‘AI-first’ will be replaced, despite their own research showing half of businesses see zero benefit from AI. It’s a bold strategy to pivot to a subscription-based AI billing model while simultaneously admitting that AI-generated code is flooding open-source with low-quality garbage. <a href="https://app.daily.dev/feed-by-ids?id=f99oheuSo&amp;id=hisjcCaKo" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
