This is a pretty unique product. Just full stacked.

As a non-dev, I never really expected to be able to train my own model without a lot of headache. If I want good data AI capabilities, I spend most of my time iterating prompts and I can only get it so far. I uploaded 150 leads (that I manually reviewed / categorized) and it was just as good as GPT 3.5 in categorizing my leads. Once I get to a 1,000+ I bet it's going to be stellar.

Really excited to see where this product goes. A couple comments / questions to the team:

- Would be great if I could indicate multiple data columns to train on. This caught me off guard with the onboarding (kept trying to select more columns). Sure, I can concatenate the columns into a single column but why? And wouldn't data accuracy be better if you trained on multiple distinct columns? This was the part that felt the most off to me.
- Also, I am using the same data set for predicting multiple categories (ie multiple data columns). Could this all be lumped together in one model eventually or do I need to upload the same data for each categorization? There seems like some potential efficiency here for your users and could avoid a big headache as well (especially when the data set gets big).
- I think your existing NLPs could be a hidden gem if you get some really on point ones. The list didn't really jump out to me but I can see this being big as your list grows
- Onboarding guide could use a little smoothing over.

And to that 1 taco review, wtf?