2049 (Mines of Merge)

At a glance

2049, or Mines of Merge, is worth trying if familiar tile merging has become too predictable and you want a puzzle with hidden hazards. You are one numbered tile in a dungeon: moving slides the other tiles too, while unseen mines can block routes and nearby cells reveal clues. The browser opening worked without an account. Its appeal is the tension between building power and reaching the stairs, rather than simply chasing a large number.

Good fit

Players who enjoy working out how a whole-board move will affect both their character and the available routes. Equal-value merges build power; the early browser trial took the player from power two to four and increased the score. Mines act as barriers for ordinary tiles, so an apparently blocked slide can itself be information. The rules make the board a navigation problem as well as a merging puzzle.

Consider something else when

You want ordinary 2048, manual Minesweeper flagging or a thoroughly tested long campaign. The player has health, stepping onto a mine causes damage, and the documented goal is to survive eight depths. Flags are automatic when adjacent cells are probed; pressing F in the trial explained that rule. Only early depth-one movement and restart were exercised, so later balance, victory and replay value remain unassessed.

Start here

Open the repository’s Play in the browser link and wait for the Python runtime to load. Find the blue player tile, health, power and stairs marker, then use an arrow key to see how the board shifts. Watch neighboring clue numbers as well as merge values. The trial confirmed movement and one power-increasing merge. For a new run, press R twice promptly: the first press warns before abandoning the active run, and the second reset the board, power and score. Touch swipes are documented, not tested here.

Cost and requirements

The source is MIT-licensed. The public browser version needed no signup or payment in the trial and loads Pyodide from a CDN, so initial access needs those assets to be reachable. The native package declares Python 3.13 or later with no runtime dependencies and documents uv run game; that path was not installed or executed. The browser frontend stores a best score locally when a run ends. Local browser scores are not an account-backed save or a synchronized ongoing game.

Limits to know

The source separates a Python game core from a browser renderer and a JSON bridge. Selected movement/probing code supports the explanation, but the acquired core excerpt was truncated and is not a complete code audit. No native build, full run, death/victory sequence or score persistence was tested. The displayed BEST tracked the current score and returned to zero after the early restart; that is consistent with the frontend saving only on game over, not evidence of a persistent score failure. Mobile controls and offline replay were not exercised.

Still unclear

• Later-depth balance and a complete escape were not assessed.
• Touch controls and score persistence across completed runs were not tested.