This is a very nice step What I appreciate most is that it’s a stable language model. It “understands” what you are trying to say and ask. It can even “understand” very specific/niche issues (though it often can’t solve them).
Without reading much of the previous comments, I will agree with the last one by @PiercedMissile, the model seems a bit too agreeable, and in my opinion that’s not something you’d want from a tool like this - you’re going to it as an authoritative source. That also makes me wonder about my statement regarding niche issues - perhaps the reason why it often can’t solve them is in fact because it doesn’t “understand” them well enough, but is rather only parroting your observations back to you, because of it’s agreeableness, making it seem like it “understands” what you’re talking about more than it actually does.
Why I came to comment initially is this: The tool references deprecated things (Instances, properties, methods, etc.) way too often. As an example, in the last few messages it referenced BasePart.Velocity
, Instance.new("BodyVelocity")
and Instance.new("BodyForce")
. When confronted about it, it corrects itself, but I don’t see a good reason why we should need to take this extra step in the long-run.
Another thing it does (less frequently) is make up methods, such as CFrame:ToQuaternion()
. Again, when confronted about it, it corrects itself, and will even provide a custom function (such as toQuaternion(cFrame)
) if asked, but I don’t see why it should confuse users by making up methods in the first place, when it could output something like this instead:
local qX, qY, qZ, qW = toQuaternion(someCFrame) -- your custom function
Alternatively, it could always automatically generate such custom functions. I don’t feel strongly about that (as long as it’s capable of generating it when asked), but that’s the direction I lean towards.