Comment by notnullorvoid
4 days ago
I don't expect Scion's design to be very good, and the implementation is likely pretty bad, unless it is a near 1 to 1 copy of existing open source.
The amount of training data for incremental/derived data systems is quite sparse in the grand scheme of things. I've seen little to no improvement over the last year when testing new models in that area, they always make some pretty bad mistakes, and what is correct is very inefficient.
Thanks. Indeed, the Scion design is quite good because Claude was pointed directly to Mononoke and asked to analyze the code base and copy it. I am very proud of what our team built by making contrarian bets at exactly the right spots.