Importantly, it uses a decision model (like Jev) behind the scenes to get around the gotchas normally associated with this, like "green-to-red scale, but high numbers for this column are actually bad", like cost increases, or batter strikeouts, etc.
Importantly, it uses a decision model (like Jev) behind the scenes to get around the gotchas normally associated with this, like "green-to-red scale, but high numbers for this column are actually bad", like cost increases, or batter strikeouts, etc.
philrenaud•1m ago
Quickly figured out that there's a problem with just applying "big numbers = green, small numbers = red" uniformly, though. On my fantasy football team page, "number of fumbles" should show green when a player has few / red when they have many.
When TypesafeAI's Jev decision model was released, it seemed like a really good fit for being able to solve this sort of problem. With some context (like "does `cost` here seem like a good thing or a bad thing, given the title of the table and adjacent columns), it is great (and quick!) at determining expected order.
Was already doing the work of parsing numbers from cells, so things like "$50-60k" becomes comparable and so do cells with adjacent text (like "32.4% Xhigh effort").
It's all open source[0] and you can BYOK (for Jev or Clef) or use it for free (with some healthy limits)
I hope you like it!
[0]: https://github.com/philrenaud/practical-data