Makes sense, we're also very small (pre-seed) so definitely no cash cow for you guys yet. We probably shouldn't be prematurely optimizing our prompting performance as it's not really a bottleneck, but a 4x improvement just by swapping an API would be too good not to act on.