Yes. Rather a lack of control of the subject and intensity of focus. Which piss everyone off because they can't steer it from the outside, hence the "lack of focus" perspective.
This is very different to my experience and I am wondering why. Maybe because I come from chess and can't help myself to compare it with this frame of reference. Anyway I felt that my progress up to 5k was largely driven by a better understanding of principles of plays than tactical training. As a thought experiment, I feel that its possible to adopt a very risk averse style that negates tactical complexities to the expense of many points on the board and still largely win against weaker players. It's not my experience with chess. If you suck at tactics, your elo sucks too.
> Anyway I felt that my progress up to 5k was largely driven by a better understanding of principles of plays than tactical training.
Well yeah. But look at these tutorials. They're all local tactics. Nothing on early game or middle game strategy.
Despite dozens of tactical positions, I don't think a single one of these holds a one-point jump, two-point jump, horse move, diagonal move, 4-row move, 3-row move or similar pattern.
I'd say this stuff gets extremely important around 15kyu, where your tactical knowledge is passable (maybe not great, but passable). It becomes more important to move around the board with 2-point jumps, recognize sente vs gote and tenuki vs weak gote moves from your opponent.
Tactical knowledge will only get you maybe to 20kyu or 15kyu at the best IMO. Then you're forced to learn "squishy" and "opinionated" discussions that no one really knows how to teach. There's patterns (ie: 2-point jump or horse-move), but its not like there's any ground truth to knowing when these moves are appropriate.
----------
I guess people's point here is that if you choose a 2-point jump as a strategic connection when a horse-move was more appropriate, you might lose 1 or 2 points. But both of these moves are likely worth 5, 10, 15, or more points.
The more important bit is knowing when to play strong and connected (often because each move is strongly sente / has momentum and forces an opponent's response), vs when a sequence is gote and the opponent (or yourself) should consider a move elsewhere (ex: possibly start an invasion).
Correct me if I am wrong, but Docling can do both. It has also, among other strategies, a non-AI pipeline to determine the layout (based on qpdf I believe). So these projects are not that different.
While it has a PDF parser, my understanding is that it is mainly used to break a PDF document into chunks, which are then handed off to various specialized models. From its docs: "The main purpose of Docling is to run local models which are not sharing any user data with remote services."
reply