codevibing

Level Up

You're already vibe coding. You've built things, deployed things, had those magic sessions where everything flows. You've also had sessions where nothing works and you want to throw your laptop.

This module is about naming the patterns so you can use them deliberately.


Know your archetype

From analyzing 10,497 prompts across 68 days of intensive vibe coding, three distinct styles emerged. Most people lean toward one:

The Scholar

Deep domain work, research, archives

You start by immersing Claude in the problem domain. You don't say "build me an app" — you say "help me understand this space." The building follows naturally from shared understanding.

Signature moves:

  • "What do you notice about this?"
  • "How would an expert approach this?"
  • "What's missing from the current experience?"

Strength: Produces deeply thoughtful, domain-appropriate work. Risk: Can get lost in research and never ship.

The Aesthete

Design, visual work, brand

You show, don't tell. Reference sites beat descriptions. Then you iterate fast with vibe words: "warmer", "more breathing room", "too corporate."

Signature moves:

  • Share reference URLs before describing anything
  • Use evocative language: "ethereal", "grounded", "luminous"
  • Cycle fast: prompt → see result → "now try X" → see result

Strength: Produces beautiful, emotionally resonant work. Risk: Can over-iterate and lose the thread.

The Sprinter

Deadline-driven, intensive builds

You blend research and building in the same session. You don't separate phases. A single session might go: find a paper → summarize it → draft how it fits → build the component → test it → move to the next section.

Signature moves:

  • External deadlines create focus
  • Never separate "research phase" from "build phase"
  • Let accumulated context do the heavy lifting

Strength: Ships fast, dense output. Risk: Can burn through context window and lose quality at the end.

>Self-assessment

Which archetype resonates most? Think about your last three building sessions. Were you researching deeply (Scholar), iterating on aesthetics (Aesthete), or blasting through tasks (Sprinter)? There's no wrong answer — this is about self-awareness, not optimization.

~

Self-assessment

Which style resonates most with how you naturally work?

No right answer. Both ends are valid.


What 10,497 prompts reveal

Here's what the data shows about real vibe coding at scale:

The arc is always the same: Experimentation → Foundation → Application. First you explore what's possible. Then you build the core thing. Then you ship it to people. Every project follows this at some scale.

Parallel streams are natural. Never one project at a time — always 3-5. Context-switching isn't a bug. It's the native mode.

Projects spawn projects. A translation tool spins off a children's book engine. A website becomes a platform. The work is generative, not linear.

Intensity is tidal. 300-500 prompts/day during shipping. 40-100 during exploration. Don't fight the rhythm.

The most common prompt openers: "this", "i", "yes". Not elaborate instructions. Just presence and direction.


Named patterns you might be missing

Strategic Ambiguity

The most common mistake is over-specification. Precise prompts produce brittle results. The art is knowing what to leave open.

  • Describe the destination, not the route
  • Name the vibe, not the pixels
  • Specify constraints, not implementations

Claude fills ambiguity with competence. Your job is to create the right-shaped space for that competence to flow into.

The Delegation

"Work on the todos for X"

Hand over the steering wheel entirely. This only works after you've built enough shared context. Early in a session: dangerous. After 30 minutes of collaboration: powerful.

The Vibe Shift

"This is too [X]. Make it more [Y]."

Fast iteration with feeling-words. "Too busy, simplify." "Too cold, warm it up." "Too corporate, make it human." Four words that carry more intent than a paragraph of specifications.

The Soft Reset

"Let's step back. The core goal is X."

When you've gone down a wrong path and need to recover without starting over. Restate the destination, not the failed route.

The Constraint Addition

"Same thing, but [constraint]."

Add constraints one at a time, not all at once. "Same layout, but mobile-first." "Same logic, but handle the error case." Incremental narrowing beats upfront specification.


Anti-patterns

Over-specification: "Create a React component with TypeScript using Tailwind CSS that has a blue background (#3B82F6) and padding of 16px..." The more you specify, the less room for Claude to apply judgment. Save precision for actual constraints.

Premature architecture: "First, let's plan the folder structure and database schema..." Start with the thing you actually want. Structure emerges from need.

Research/build separation: "First research everything, then we'll build." Blend them. Context accumulated during research makes building better.

Critique Exercise

Both prompts ask for the same thing: a dashboard for tracking daily habits. Which prompt will produce a better result? Make your prediction.

Clarity of intentRoom for AI judgmentLikely output quality

Power user: beyond prompts

CLAUDE.md mastery

Your CLAUDE.md isn't just preferences — it's a self-portrait. Write it as "here's how I think and work," not "here are your rules." The better Claude understands you, the less you need to prompt.

MCP servers

Connect Claude to external tools — databases, APIs, browser devtools. This turns Claude from "AI that writes code" to "AI that operates your infrastructure."

Skills

Reusable prompt patterns saved as files. When you find yourself giving the same instructions across projects, extract them into a skill.

Hooks

Auto-run commands on events. Run tests before commits. Notify yourself after deploys. Lint on save. The automation layer on top of the conversation.