Cloudflare's vinext: A Next.js-compatible framework built with AI on Vite

What vinext is
vinext (pronounced "vee-next") is a drop-in replacement for Next.js built on Vite that deploys to Cloudflare Workers with a single command. One engineer and an AI model rebuilt it from scratch in one week for about $1,100 in tokens.
The deployment problem vinext solves
Next.js has deployment challenges in serverless ecosystems. While tools like OpenNext exist to adapt Next.js output for platforms like Cloudflare, Netlify, or AWS Lambda, they require reverse-engineering Next.js's build output, leading to fragile implementations that break between versions. Next.js's adapters API is still early and doesn't solve development-time issues - next dev runs exclusively in Node.js, preventing testing of platform-specific APIs like Durable Objects, KV, or AI bindings during development.
How vinext works
Instead of adapting Next.js output, vinext reimplements the Next.js API surface directly on Vite as a plugin. This includes routing, server rendering, React Server Components, server actions, caching, and middleware. Vite's Environment API enables the output to run on any platform.
Setup is straightforward:
npm install vinextReplace next with vinext in your scripts while keeping app/, pages/, and next.config.js unchanged.
Commands:
vinext dev- Development server with HMRvinext build- Production buildvinext deploy- Build and deploy to Cloudflare Workers
Performance benchmarks
Benchmarks compare vinext against Next.js 16 using a shared 33-route App Router application, with TypeScript type checking and ESLint disabled in Next.js's build to match Vite's behavior.
Production build time:
- Next.js 16.1.6 (Turbopack): 7.38s baseline
- vinext (Vite 7 / Rollup): 4.64s (1.6x faster)
- vinext (Vite 8 / Rolldown): 1.67s (4.4x faster)
Client bundle size (gzipped):
- Next.js 16.1.6: 168.9 KB baseline
- vinext (Rollup): 74.0 KB (56% smaller)
- vinext (Rolldown): 72.9 KB (57% smaller)
These benchmarks measure compilation and bundling speed, not production serving performance. The test uses a single 33-route app, not a representative sample of all production applications.
Current status
Cloudflare already has customers running vinext in production. The full methodology and historical results are public.
📖 Read the full source: HN AI Agents
👀 See Also

Dart AI productivity app review with OpenClaw integration
A user reports switching from Things to Dart AI for productivity, finding it better for implementing Getting Things Done methodology with full OpenClaw access, despite UI issues and initial setup complexity.

PicoClaw Fails to Build F1 AI Agent, Burns $20 in API Credits
A developer attempted to build an F1 information bot using PicoClaw on a Raspberry Pi Zero 2W, but the tool defaulted to version 11, generated hallucinated Python code, and consumed $20 in DeepSeek API credits without producing a working solution.

MCP Server Connects Claude to CellarTracker Wine Inventory
A developer built an MCP server that connects Claude directly to CellarTracker accounts, allowing conversational queries about wine inventory, tasting notes, purchase history, and drinking windows without manual CSV exports.

Self-Evolving Skill pattern validation: 5-round experiment results
A developer tested the Self-Evolving Skill design pattern for Claude Code with a 5-round experiment on a MySQL database with 29 tables and 590MB of smart building management data. Key results include a 63.6% Five-Gate rejection rate, incremental convergence, and 100% accuracy with no incorrect knowledge surviving.