wordle/_bmad-output/planning-artifacts/briefs/brief-wordle-2026-07-07/brief.md

---

title: "Product Brief: Wordle Clone" status: draft created: 2026-07-07

updated: 2026-07-07

Product Brief: Wordle Clone

Executive Summary

A word-guessing game — 5 letters, 6 attempts, green/yellow/gray feedback — built as a hands-on project to learn the BMad method end-to-end. The gameplay feels familiar but introduces two mechanics the original Wordle lacks: computer word ranking (every word pre-solved by algorithms to produce objective difficulty scores) and computer guessing (watch the solver's optimal path, or pit yourself against it in a subgame). V1 ships the core game and solver engine. Post-v1 features are prioritized by real user feedback — ship to friends, ask what they want, build that.

The Problem

I'm learning the BMad method and need a hands-on project. Abstract tutorials don't stick. The project must be well-understood (no domain research rabbit holes), scoped for one person, but feature-rich enough to exercise the full BMad workflow from analysis through implementation. A Wordle clone hits the sweet spot: everyone knows the rules, they're crisp, yet there's genuine room for design decisions around the solver engine, word ranking, and the architecture that ties them together. The codebase itself becomes a reference for future BMad projects.

The Solution

A browser-playable word-guessing game with the familiar Wordle mechanics. Under the hood, a computer solver pre-computes every word using multiple algorithms, producing an objective difficulty score for each. That solver engine is the architectural centerpiece — it drives word ranking, validates the word bank, and will power future features post-v1. The point isn't just to ship a clone — it's to have a working system I understand end-to-end, built with the BMad method, that I can extend later.

What Makes This Different

Two things the original Wordle doesn't do:

Computer word ranking. Every word in the bank is pre-solved by an algorithm, producing an objective difficulty score. That score drives word tiering and fair comparisons. No manual labeling — the data does the work.

Computer guessing. The same solver becomes a teaching tool: after you solve, watch the optimal path unfold. Or play a subgame where you pick a word and the computer cracks it in front of you, revealing strategy.

Everything else in the roadmap — percentile ranking, streak quality, standardized hints, the post-solve learning layer — builds on these two engines.

Who This Serves

Starts with me — I'm learning BMad and need a real project. First players are myself and friends: tight feedback loop, low stakes. If the game proves solid and copyright isn't a blocker, I'll publish publicly for anyone who enjoys word puzzles. A secondary audience: other developers — the codebase is a BMad-built reference project.

Success Criteria

1. Playable in a web browser — full game loop: guess 5-letter words, get green/yellow/gray feedback, solve or fail within 6 attempts
2. Multiple solver algorithms running — at minimum, the solver pre-computes attempt counts for every word in the bank
3. Word ranking calculated — every word has an objective difficulty score derived from solver data
4. BMad workflow exercised end-to-end — brief → PRD → architecture → epics/stories → implementation, all completed

Scope

In (v1):

• Core Wordle gameplay: 5-letter words, 6 attempts, green/yellow/gray positional feedback
• On-screen keyboard with letter highlighting
• Multiple computer solver algorithms
• Word ranking data derived from solver output
• Browser-playable (web application)

Out (post-v1, prioritized by friend feedback):

• Adaptive difficulty system
• Percentile ranking
• Streak quality weighting
• Standardized hints on attempts 5-6
• Post-solve learning layer (definition, etymology, picture)
• Daily word + infinite after-play flow
• Computer-guess subgame
• Real-time 2-player mode
• Variable word length (3-7 letters)
• Position-blind hard mode

Vision

V1 ships the core game and solver engine. Then I put it in front of friends, ask which feature from the backlog they'd actually want, and build that first. Each iteration is guided by real feedback, not guesses. If the game gains traction after public launch, the full brainstorm vision is waiting — adaptive difficulty, learning layer, ranked play. If it stays a learning project, the codebase stands as a clean reference for future BMad work.