🤖 AI Products · $25/mo

Can I vibe code Lovable?

Describe an app, get a deployed full-stack web app.

🔴Not reallyest. a weekend for a toy; the product: a company

The meta question — vibe-coding a vibe-coding tool. The chat UI is trivial; the product is an industrial agent harness: sandboxed execution for thousands of concurrent builds, error-recovery loops, framework scaffolds, integrated auth/database/hosting, and model orchestration tuned by a team full-time. Here's the twist though: you personally don't need to build Lovable — you already have Claude Code or Codex, which is the engine with you as the harness.

The one-shot prompt
Skip cloning Lovable — instead, build me a personal 'app spawner' that leans on the agent I already have. A small dashboard where I enter an app idea and it: creates a new local project folder from my preferred Next.js template, writes my idea into IDEA.md with a structured brief (features, pages, data model guess), initializes git, and prints the exact Claude Code / Codex kickoff prompt to paste. Track spawned projects in a list with status notes. Be honest in the README: this is a project bootstrapper, not an app builder — the agent is the builder.
Paste into Claude Code, Codex, or your agent of choice — then steer. Learn the moves in Prompt Engineering.
What you build

Realistically: a thin chat wrapper that forwards prompts to an agent SDK against a repo — a toy demonstrating why the real orchestration is the product.

What you give up
  • Sandboxed multi-tenant execution and preview infrastructure
  • Recovery loops that turn model failures into shipped apps
  • Integrated one-click auth, database, and deploy
  • Continuous harness tuning as models change monthly
Why people still pay

Non-coders pay for the guarantee that something deployable comes out the other end, every time, with zero setup. That reliability layer is the company.

infrastructureagent harness
Related
Actually build it

A prompt is the first move, not the whole game. Course 01 teaches you to take a prompt like this one to a shipped, working app — reviewing, correcting, and steering the agent the whole way.

Start Course 01 →