> because of the huge amount of resources search consumes I’ve been intermittently working on much the same idea as the OP, and I suspect this is actually a lo

> because of the huge amount of resources search consumes I’ve been intermittently working on much the same idea as the OP, and I suspect this is actually a lot less of a problem than it seems, since they’re focusing on a niche. Indexing everything the way Google does requires a lot of resources, but indexing the majority of useful material in a specific domain takes a lot less. (My ElasticSearch index for the entirety of StackOverflow is a mere 40 GB, for example.) By far the more expensive part is likely to be paying market rates for a developer (you need a decent number of users paying $10/mo to hit a mid-market salary), but in theory this scales relatively independently of userbase. Edit: I’ve just noticed I’m replying to the OP, who’s mentioned downthread that they’re using BigQuery and spending $200/week. I’ve gone the marginalia.nu route and run everything on a computer in my living room, which changes the calculus somewhat—it’s a lot cheaper, but probably involves more development time. For me it’s mainly about the learning experience but I’d be interested to hear your thoughts on the tradeoff.