<h2>Must know</h2>
<ul>
<li><strong>OpenAI Codex architecture and JSON RPC protocol</strong>: OpenAI revealed that the primary engineering hurdle for Codex was the orchestration layer rather than the underlying model. To handle complex agent interactions like mid-task approval pauses and streaming progress, they bypassed the Model Context Protocol in favor of a custom JSON-RPC App Server protocol. This architecture allows a single codebase to power the CLI, VS Code extension, and web interfaces simultaneously while mitigating context window growth through prefix caching. <a href="https://app.daily.dev/posts/IaNEGzsz8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vercel acquires new.website to bolster v0 agents</strong>: Vercel has acquired the startup new.website to integrate its production-ready primitives directly into the v0 platform. The move aims to make v0 fully agent-aware, allowing AI to autonomously handle core web features like forms, content management, and data fetching. According to Vercel, the incoming team will focus on evolving v0 from a UI generator into a comprehensive agentic development environment. <a href="https://app.daily.dev/posts/TXo5OfLTR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Auth0 launches official Gemini CLI extension</strong>: Auth0 has released an official MCP server extension for the Gemini CLI, enabling developers to manage tenant authentication via a single command. Once authenticated through a device code flow, the CLI maintains persistent context of applications, logs, and API connections. This eliminates the need for manual environment variable setup or repeating tenant details in every prompt, significantly streamlining agentic identity management. <a href="https://app.daily.dev/posts/PSlwFRCWg" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Bun v1.3.11 adds native OS cron scheduling</strong>: The update introduces Bun.cron for cross-platform task scheduling alongside ANSI-aware string slicing and over 100 Node.js compatibility fixes. <a href="https://app.daily.dev/posts/DuG54Q8tB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ConnectRPC enables typed SDK generation from GraphQL</strong>: WunderGraph’s new protocol layer compiles GraphQL operations into Protobufs, serving them as gRPC or JSON endpoints with deterministic client generation. <a href="https://app.daily.dev/posts/Cgx6Z8AyK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Claudeck provides a local browser UI for Claude Code</strong>: A new open-source tool offers a parallel chat grid and cost analytics for Claude Code, running entirely locally via WebSocket without telemetry. <a href="https://app.daily.dev/posts/DXO1Zn3rb" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Heroku CLI v11 completes ESM migration</strong>: The architectural overhaul moves to oclif v4 and Node.js 22, delivering faster command execution and a new semantic color system. <a href="https://app.daily.dev/posts/2swWgPS9c" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vite 8 replaces legacy bundlers with Rolldown</strong>: The release swaps esbuild and Rollup for the Rust-based Rolldown, claiming up to 30x faster production builds while keeping plugin compatibility. <a href="https://app.daily.dev/posts/QG6z68w6t" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Atomic converts markdown notes into AI knowledge graphs</strong>: This open-source personal knowledge base uses sqlite-vec for vector search and an MCP server for deep Claude integration. <a href="https://app.daily.dev/posts/y5RZTff20" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>IntelliJ IDEA makes core web features free</strong>: Starting with v2026.1, JetBrains is moving JavaScript, TypeScript, and CSS support from the Ultimate edition to the free tier. <a href="https://app.daily.dev/posts/qYptuJh4r" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The zero dollar AI coding stack is a rate limit trap</strong>: Clickbait guides are currently pushing ‘fully free’ $0 AI workflows that cobble together eight different free tiers and ‘secret’ APIs. While technically possible for a weekend hobby project, any professional developer will hit rate limits or time-limited trials within the first hour of actual work. Stop pretending that managing a dozen different logins and API keys is a viable alternative to paying for a reliable tool. <a href="https://app.daily.dev/posts/MXcayKavV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Generative Engine Optimization is just SEO with more hallucination</strong>: Marketing firms are desperately trying to make ‘GEO’ (Generative Engine Optimization) happen, selling tools to ‘optimize brand presence’ in ChatGPT and Perplexity answers. It is essentially the same old SEO snake oil rebranded for the LLM era, promising to track citations that the models frequently fabricate anyway. Spending enterprise budgets to ‘game’ a non-deterministic black box is a fantastic way to set money on fire. <a href="https://app.daily.dev/posts/6t97IdU4o" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vibe coding is not a replacement for system design</strong>: The ‘vibe coding’ trend—where you just prompt until it works—is creating a massive wave of unmaintainable ‘AI slop’ in production. While tools like Google Stitch and v0 make for great demos, they lack the judgment for original product design and often produce bland, over-engineered results. If you aren’t moving your judgment upstream to the specification level, you aren’t ‘engineering’; you’re just gambling with a very expensive autocomplete. <a href="https://app.daily.dev/feed-by-ids?id=MDkRe4LKP&amp;id=dX9Yva0Ku" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
