>>109574823
>nobody wrote 2.8 million lines of code in a year
In a month*
>>109574805
>Is there anything in the code that unequivocally proves he's using Claude far more than he claims?
The two major signs of AI coding that I know of are
1. Obnoxious AI writing style in the comments
2. Code works but has incomprehensibly bad architectural decisions
I opened a file and random and got this comment:
-- Shared local date/time presentation for engine UI and mods. Preferences
-- live in options.lua, never checkpoint progress. "device" uses the process
-- time locale when the platform supplies one and otherwise falls back to the
-- deterministic DD-MM-YYYY / 24-hour convention.
which is blatantly AI. But I guess that doesn't help much if he's claiming that the comments are AI generated but he wrote the code himself. And architecture is hard to judge unless you know the codebase really well and also hard to distinguish from just being a bad programmer unless you're actively talking to them (if it's AI generated they won't have any good answer to "why did you do it this way that makes no sense").
Some other things you could maybe look for that are harder to spot:
>Thinking leaking into comments. Occasionally the AI will plan to implement a function one way but change its mind halfway through, and leave a comment like "Wait, actually, we should do this like X instead of like Y" and then the rest of the implementation is X.
>Comments that talk about an old implementation for no reason. Like "we used to do X, but now we do Y", and the implementation is entirely Y, with no sign of X. AI often writes comments like this when editing code. If you edited the code yourself and had the AI add comments after the fact, it's probably less likely to do this (not definitive though).
Anyway, the only proof you need is the fact that he's claiming to have written 2,800,000 lines in a month. That comes out to 100 lines per minute, 16 hours a day, for 30 days
