That's super cool. I hate the pricing. I typically know everything I need to for day-to-day usage of the shell and only do things that require discovery every few months. 100 queries wouldn't be enough in those months, so I'd have some months where I'm paying for nothing and the odd month where I don't get enough usage.
$9 per month also makes it costly enough that I wouldn't buy it as a "just to have" kind of tool. I don't think I'd get $100 of value vs searching online, especially since I attribute some negative value to tools that can be taken away from me. I don't want to pay forever and be dependent on something that could disappear tomorrow.
I don't get why something like that needs to be an online service. I don't know much about AI, so maybe it's a lack of understanding on my part, but why can't I simply have a copy of the trained model on my local machine where there's no ongoing cost (to you) whenever I run a command? Isn't an online API a complex solution to a problem that could be solved with a local app + data?
Maybe I just lack understanding and the models are too big or the compute required to make a query is huge. If you could give some insight I'd genuinely appreciate it.
Even though I'd never buy it as a subscription, it's the kind of thing I'd pay for as a perpetual app. I'd wouldn't hesitate to pay $50 if I could install it on my machines and forget about it until it would be useful. I'd also expect to pay for updated versions of the models whenever I need them.
Regardless, I think it's amazing as a discovery tool. I don't mind reading 'man' pages to figure out details, but I always feel like it's a hassle to discover what command I need for certain tasks.
Also, I'm probably an outlier since I make a lot of effort to avoid tools that rely on an internet connection to function. IE: I won't rely on GitHub. I'll use it, but only as a push mirror.