1. Testing isn’t always as easy as following a list of steps, and when it _is_ those get to be technical. Explaining those as text instructions could get harder than what most expect if they are to be useful.
2. This is a lesson learned from experience - unlimited plans rarely scale. Instead, they make those who would fit into an entry level (aka just a few tests) feel like they’re paying a premium and those who will fit a large plan not pay nearly as much as what they consume.
Not only that, but when the founders realize this they’ll have to change it and users will be upset, so they’ll grandfather them and we’re back to the beginning.
Side note: I love “breaking” new features personally and know it’s not a straight forward process. If all this aims to do is replace selenium I don’t see the need for it — selenium tests are very predictable where text-to-human isn’t. If it aims to go beyond, that can’t happen without real knowledge of the application.