Appreciate the kind words! Linear scaling in terms of speed and cost. We haven't yet optimized the prompts & choice of model to minimize token usage, so I'd recommend emailing us for advice if you want to run this on a large dataset