That of course depends on how much that “other existing forum software” is complicated, and how many features need to be changed to fit our needs.
To answer your question, it is not “obviously better”; you just have to make a probabilistic estimate. Sometime people make the mistake of reinventing the wheel, when a good (or even superior) version already exists. But sometimes also people spend a lot of time investigating existing options, only to find out that all of them have some major problem. (And this process can also cost a lot of time, and when it fails, you are left empty-handed at the end.) I have seen both of these situations in real life. The “don’t reinvent the wheel” part gets very popular on internet, but people underestimate that even when a solution is already available, it can take a lot of time to learn how to use it properly (that means, not just use the software as an ordinary user, but also as an administrator, and even to be able to customize it; those are an order of magnitude more complex tasks).
I guess a reasonable first step would be to set a deadline, for example a month or two, to explore the existing alternatives. (But you have to set a deadline, because there will always be yet another obscure alternative you have not thoroughly examined yet.) Maybe a month would be enough, if we consider the fact that LW already runs for a few years and it was not replaced by an existing superior alternative.
Is rewriting everything from scratch obviously better than forking some other existing forum software (e.g. Discourse) and adding features?
That of course depends on how much that “other existing forum software” is complicated, and how many features need to be changed to fit our needs.
To answer your question, it is not “obviously better”; you just have to make a probabilistic estimate. Sometime people make the mistake of reinventing the wheel, when a good (or even superior) version already exists. But sometimes also people spend a lot of time investigating existing options, only to find out that all of them have some major problem. (And this process can also cost a lot of time, and when it fails, you are left empty-handed at the end.) I have seen both of these situations in real life. The “don’t reinvent the wheel” part gets very popular on internet, but people underestimate that even when a solution is already available, it can take a lot of time to learn how to use it properly (that means, not just use the software as an ordinary user, but also as an administrator, and even to be able to customize it; those are an order of magnitude more complex tasks).
I guess a reasonable first step would be to set a deadline, for example a month or two, to explore the existing alternatives. (But you have to set a deadline, because there will always be yet another obscure alternative you have not thoroughly examined yet.) Maybe a month would be enough, if we consider the fact that LW already runs for a few years and it was not replaced by an existing superior alternative.