[francylisboa]

I particularly dislike using the phrase "trained on your own data". We're not actually training anything. We're just showing a pre-trained and complex auto-completer how we'd like the outputs to talk about. Training involves a recalibrate the weights of those beast LLMs.