First some background:
An 'input mechanism' is a tool to enact the moves you want to be played on a chessboard. The perfect input mechanism is probably "play whatever move I have in my mind immediately". But is this the fastest possible input mechanism? If you don't mind some perversion of the spirit of chess, you can get much faster: "play a random legal move", eschewing the need for you to come up with a move and verify its legality. Where the line is between this and outright computer assistance, "play Stockfish's top move", is unclear. We'll come back to this.
The only input mechanism possible in over-the-board chess is using your hand to move the piece and the same hand to stop his clock (FIDE Laws of Chess Articles 4 and 6). (Any other OTB mechanism, e.g., using your non-dominant hand, is unilaterally disadvantageous, and would be a handicap.) This mechanism is substantially slower than most online chess input mechanisms, simply because of the great distance the hand has to move. Hence the fastest time-controls playable OTB (1m+2 or 2m+1) are much slower than the fastest controls played online (15s+0 and 10s+0). Although we are getting further from the slow ponderous game many of us are used to, success in even the fastest controls correlates primarily with chess skill more than reflexes: the world's best hyperbullet players are Andrew Tang (a GM), Alireza Firouzja (World Top 15 classical player) and Nihal Sarin (a strong 2600+ GM at only 16yo).
'Keyboarding' on Lichess
Let's get into specifics. Lichess (like Chess.com) allows a number of input mechanisms -- drag-and-drop with mouse, click on source and target squares with mouse, touch-pad, keyboard input using short algebraic notation. Around one month ago a new recently available extension became available, Lichess-Keyboard-Input, ostensibly with a goal of speeding up input without abuse (see the Github Readme). This extension allows you to press a key to select a piece and then move your mouse over a square. If it's legal to move your piece to that square, the move is enacted.
This, and not the SAN keyboard input (which is slower than a mouse), has been deeply controversial. Proponents argue that those who are 'locked in' to point-and-click simply do not want to be unseated by a new faster mechanism to which adaptation would be necessary. However, it is easy to point to instances of potential abuse using the extension: in particular, in time-scrambles I can eschew the need to verify my moves' legality, since just by pressing a key and moving the mouse around at random it will enact some move as long as there exists a possible legal move for that piece. Furthermore, I can play as follows:
- identify a piece to be moved and hold down the relevant key
- 'circle' my mouse around the board non-stop
This means I don't need to react to my opponent moving before I move - as soon as, traversing its circle, the mouse finds a legal square, after the opponent's last move, I will move instantly. This means I do not need to check legality, or react to my opponent's move; I can choose a piece and generate an instant random legal move with it, so long as one exists. This is equivalent to an input mechanism 'play a random move with the piece I select' in time-scrambles. And if I choose my King, I can eliminate the need to even avoid capture with the piece selected.
Hence, the opponent of a 'keyboarder' has no option but to checkmate them: winning on time is almost impossible. Empirically, the keyboard extension provides an enormous ELO boost of 100-200 points at 15s+0, the fastest time-control on Lichess, compared to any other input method. But since Andrew Tang and Nihal Sarin win by playing good moves quickly rather than poor moves extremely quickly, they could not enjoy much of this benefit even if they were to adapt to using the extension: no wonder then they resist its use.
The debate rages on how much of the mentioned ELO boost lies in the abusive mechanisms I mentioned, which I think we can agree insult the spirit of chess, and how much is a legitimate advancement thanks to a reasonable and fair new method of inputting moves.