If you're only working on one problem that's very valuable to solve, then taking the time to train a classifier is great. The beauty of LLMs is that you can ru

If you're only working on one problem that's very valuable to solve, then taking the time to train a classifier is great. The beauty of LLMs is that you can run a ton of experiments, notebooks, demos etc because you can write classifiers and structure unstructured data so fast, in a reasonably accurate way (at the moment it seems roughly in line with say hiring an intern to label things)