Tracer bullet

Tracer bullet

Code First, Then Expand
You code the seed, AI helps it grow. Tracer bullets are for when you're the expert with a specific architectural vision. You build the minimal working proof-of-concept first, then let AI help you expand it systematically while preserving your design decisions.
Example: Basic real-time sync (your working foundation)
// Basic operational transform for text sync
function applyTransform(doc, op) {
  if (op.type === 'insert') {
    return doc.slice(0, op.pos) + op.text + doc.slice(op.pos);
  }
  // ... handle delete, etc
}

io.on('connection', (socket) => {
  socket.on('edit', (op) => {
    // Transform against concurrent ops
    const transformed = transform(op, pendingOps);
    applyToDocument(transformed);
    socket.broadcast.emit('edit', transformed);
  });
});
• When to use: Custom architectures, niche frameworks, or when you have strong implementation opinions
• Why it works: You control the architectural decisions; AI handles systematic expansion and refinement
• Skip it when: Using common patterns the AI knows well, or when you want AI to explore alternatives
Build Your Tracer Bullet Prompt
Expand your working code systematically

Describe your system - be specific about the architectural choices you want to demonstrate.

Required: Describe your existing working code. What have you already built and proven works?

What features should AI help you add systematically to your working foundation?

What architectural decisions are non-negotiable? What patterns must AI follow?

Your expansion prompt will appear here

Show AI your working code and get systematic guidance to grow it while preserving your architecture.