alt Hacker NewsYou are not wrong, except at scale it gets complicated quickly. For starters, to support large user numbers, you’re going to have to process your own grib2 data for radar and turn them into tiles at zoom levels.
It takes about 24 cores with a GPU to do CONUS, Canada, Alaska, Pacific and Caribbean data. This should be 2x for redundancy. Even being cheap with main processing in my basement (gen power, backup internet) the cloud costs to serve it are $200 month plus data transfer. The standby grib machine spins up should it not see the cheap primary or the NOAAPort receiver is offline.
There is no money to be made without whoring out your user’s privacy. People just won’t pay for a privacy focused weather app. I keep this going as a hobby.
Fair enough. Things are always more complicated at scale.
But then again, we don't know whether this company is maintaining this infra themselves, or if they're paying for API access. Besides, if anything, running their own servers is often the more cost-effective option, so the details you mention might not matter in practice.
My incredulity has more to do with the profitability of this type of software, considering that the free options are good enough for the average person, and that the features promoted in the article are hardly innovative.
> There is no money to be made without whoring out your user’s privacy.
Well, I do object to that. It's certainly possible to sustain a profitable business without selling out your users' data. It may not be as profitable as the advertising model, which is often too enticing for companies to ignore. This company explicitly says that their income comes directly from customers, so apparently I'm underestimating the amount of people who find these features valuable enough to pay for them.