Very interesting. This is pretty much the opposite of the idea I've had for making Go more beginner-friendly.
I think some beginners could benefit from see the more complex nature of Go, by letting them focus on large scale matters, and ignore the tactics of battle. The way to do this would be to let the player choose which general region of the board they want to play in, but have their stone snap to the locally best intersection as their cursor moves across the board. (The snapping, I suppose, could vary in distance depending on how much stronger the strongest move is. A very strong/important move could snap from across a quarter of the board, but if it's a less important move they'd effectively have to hover on one of its neighbours for it to be suggested.)
So effectively the player would get unassisted practise at spotting which groups of stones are weak, and where influence needs to be strengthened, but simultaneously have AI assist in dealing with local battles and making life.
I suspect this would make Go play more like a traditional RTS video game: at lower levels, the player bosses over more general goals, and individual units make local decisions on when to shoot and how to move places through a primitive AI. At higher levels, the human player takes over more of the local decision making in improving their "micro".
Though, personally I'm not sure how far one could get and staying true to the spirit, or the deliciousness of go, namely the interweaving quality of different ever-status-changing groups and also considering that stones have somewhat global effects (ladder).
Perhaps the AI could also make helpful annotations to player why the move was important. Or suggest that this and that group should be strengthened. Or if player makes a move that first selects a region, then give some options to choose from.
I love abstract aspects of go (e.g. reading attack and defence, direction of play) but often situations involve local context that are too interesting to ignore, and thus tesujis. It would definitely would be fun to see how a game like you are suggesting would feel like.
One thing I really liked about go is that learning from a book, or trying out puzzles, is much easier than for example in Chess due to much less destructive change in the board.
(I'm ranked online perhaps 3-5 kyu, so take my take with a grain of salt).
I like the idea you linked to, too. In fact, the first version of that "have more stones on the board" is exactly how I explain the rules of the game to beginners. They pretty quickly realise the tediousness of filling in what's effectively already owned intersections, at which point they're ready to hear that "normally, experienced players don't actually go through the trouble of filling it in, they just count it as theirs. If there is any dispute about whose it is, play resumes until it becomes clear again."
I think some beginners could benefit from see the more complex nature of Go, by letting them focus on large scale matters, and ignore the tactics of battle. The way to do this would be to let the player choose which general region of the board they want to play in, but have their stone snap to the locally best intersection as their cursor moves across the board. (The snapping, I suppose, could vary in distance depending on how much stronger the strongest move is. A very strong/important move could snap from across a quarter of the board, but if it's a less important move they'd effectively have to hover on one of its neighbours for it to be suggested.)
So effectively the player would get unassisted practise at spotting which groups of stones are weak, and where influence needs to be strengthened, but simultaneously have AI assist in dealing with local battles and making life.
I suspect this would make Go play more like a traditional RTS video game: at lower levels, the player bosses over more general goals, and individual units make local decisions on when to shoot and how to move places through a primitive AI. At higher levels, the human player takes over more of the local decision making in improving their "micro".