> They mention a host of things that the autorouters are bad at as evidence that autorouters are useless, instead of recognizing the opportunity to improve auto

> They mention a host of things that the autorouters are bad at as evidence that autorouters are useless, instead of recognizing the opportunity to improve autorouters to make their jobs easier. When a bunch of professionals say that something is useless, pehaps you should pay attention? The problem is that most PCB boards are extremely constrained routing environments. The boards that KiCad is used to design are extremely so. Vias go completely through the board. Even worse, vias tend to be larger than tracks. Consequently, changing a layer is expensive and introduces more obstacles. Your subcircuit part is also an obstacle on your routing layer. And, generally, you only get 2 "routing" layers (top and bottom) since the other two layers are used to route power and ground. There is a reason why boards that have FPGAs on them tend to use blind vias which don't penetrate the whole board--they can't route to the balls otherwise. They also tend to be 8+ layer boards. This represents real money and if people could do otherwise (either by hand or autorouter), they would. However, let's back this out to somewhere that "autorouters" are actually useful--VLSI. As someone who has done VLSI design for a very long time, I can pinpoint the moment when VLSI autorouters went from useless to useful--when metal layer counts hit 4+ layers (local connect, horizontal connect, vertical connect, and power layer). VLSI does not have the same obstacle issues that PCBs have. Vias only go layer to layer unless you design otherwise. Vias also are smaller than tracks. Consequently, you can practically guarantee that you can interconnect things with enough space. Your subcircuit doesn't also interfere with one of your routing layers. And VLSI design has a LOT more money to throw at developing autorouters. Even so, VLSI autorouting and placement still looks like crap 25 years on. Nobody has yet been able to come up with an autorouter that can identify a datapath and lay things out in a coherent fashion. And there is good money in doing so. People like you then come along and are suggesting that the KiCad developers spend a lot of time, on a known very hard computational problem, that people failed horribly to solve in the past with a lot more money and will likely only work well on a minority of expensive PCB boards (8+ layers with blind/buried via). Color me unsurprised when the responses are less than hospitable.