PG is wrong on the subject of patents though: the problem with software patents is that if you are building a car the price of having a lawyer work out if you a

PG is wrong on the subject of patents though: the problem with software patents is that if you are building a car the price of having a lawyer work out if you are violating a couple patents is a lot less than 1%, but it would properly make it many thousand times more expensive for a startup to develop a piece of software (assuming that we don't count the founders time). PG has said in other essays that it often makes sense for a startup to not worry about whether what they are trying to do until they are sued. If they are still tiny it is unlikely they will be sued even if they are infinging, and if they are large then they will have the resources to either defend against the suit or make a reasonable settlement. But that is a practical matter. None of that changes the fact that it borders on logically inconsistent to argue for patents of machines while against the patenting of algorithms (and a program is nothing but an algorithm that has been properly formatted).