# Explore Exploit Cycles

> Use when managing growth experiments, when a product area faces diminishing returns, or when deciding whether to generalize or specialize in career or product strategy

- Skill: `coowoolf/explore-exploit-cycles` (Agent Skill)
- Install (CLI): `npx skillmds@latest add coowoolf/explore-exploit-cycles`
- Raw SKILL.md: https://api.skillmd.com/api/skills/coowoolf/explore-exploit-cycles/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Marketing & Growth
- Author: Coowoolf (https://skillmd.com/u/coowoolf)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/coowoolf/explore-exploit-cycles

---


# Explore and Exploit Framework

## Overview

A cyclical approach to growth that alternates between discovering new opportunities (**Explore**) and maximizing their value (**Exploit**) to prevent stagnation in local maxima.

**Core principle:** Recognize when returns diminish (saturation) and deliberately return to exploration.

## The Cycle

```
        ┌─────────────────┐
        │   EXPLORATION   │
        │ (Find Mountain) │
        └────────┬────────┘
                 │
                 ▼
        ┌─────────────────┐
        │   VALIDATION    │
        │ (Prove Insight) │
        └────────┬────────┘
                 │
                 ▼
        ┌─────────────────┐
        │  EXPLOITATION   │
        │ (Scale/Optimize)│
        └────────┬────────┘
                 │
                 ▼
        ┌─────────────────┐
        │   SATURATION    │◄──── Signal to restart
        │(Diminishing ROI)│
        └────────┬────────┘
                 │
                 └──────────────► Back to EXPLORE
```

## Mode Characteristics

| Mode | Mindset | Activities |
|------|---------|------------|
| **Explore** | Divergent | Find new levers, user psychology, hypotheses |
| **Exploit** | Convergent | Scale proven insight, optimize, A/B test |

## When to Switch Modes

| Signal | Action |
|--------|--------|
| Experiments hitting < 1% lifts | → Explore |
| Found high-leverage insight | → Exploit |
| Team feels "stuck" | → Explore |
| Clear winner validated | → Exploit |

## Common Mistakes

- **All Explore**: Scattershot without scaling wins
- **All Exploit**: Stuck in local maximum, diminishing returns
- **No oscillation**: Failing to recognize saturation signals

## Real-World Example

Chess.com "explored" why users reviewed games (finding they did it after wins, not losses), then "exploited" by redesigning to celebrate wins—increasing engagement 25%.

---

*Source: Albert Cheng (Chess.com, Duolingo, Grammarly) via Lenny's Podcast*