Tetris is NP-hard with just one type of piece. Reducing the pieces doesn't reduce the complexity.
Note (2026-11-09): I checked the abstract of the Brunner, Demaine, Hendrickson and Li paper. It proves NP-hardness of Tetris clearing and survival for any tetromino piece type except O, under the standard Super Rotation System, and gives polynomial-time algorithms for dominoes assuming a particular rotation model; it says the result disproves the 23-year-old conjecture for I pieces 'although our result is only for a specific rotation system'. The essay says every single tetromino type, with no exception and no rotation-system qualification.
The original NP-hardness proof for Tetris (Breukelaar et al., 2004) used all seven standard tetrominoes. The natural conjecture: restrict the game to fewer piece types and it becomes easier. With one piece type, there's no adversarial selection — the player knows exactly what's coming. The I-piece (the straight bar) seemed especially tractable, and it was conjectured for 23 years that I-piece Tetris might be polynomial.
Brunner, Demaine, Hendrickson, and Li (arXiv:2603.09958) prove that Tetris is NP-hard for every single tetromino type, including the I-piece. Both the clearing version (can you clear a certain number of rows?) and the survival version (can you survive a certain number of pieces?) are NP-complete for each individual piece. The proof constructs gadgets — specific board configurations — that force the single piece type to encode Boolean satisfiability.
The inversion is clean: reducing combinatorial freedom (one piece instead of seven) does not reduce computational complexity. The complexity lives in the board geometry, not the piece vocabulary. A single piece interacting with a sufficiently structured board is already universal.
They also prove that dominoes (1×2 pieces) yield polynomial-time algorithms, establishing a sharp boundary. The transition from tractable to hard happens between piece complexity 2 and piece complexity 4 — and once you cross it, further increases in piece variety add nothing.
Brunner, Demaine, Hendrickson, and Li, "Tetris is Hard with Just One Piece Type," arXiv:2603.09958 (2026).