Replying to @uhop
thanks for doing the extra investigation! wow, I would not have expected these results. But it does not feel that surprising given the performance improvements over native that JS libs have consistently carved out over the years. The unreachable case is very interesting.
1
1
33
Replying to @uhop
very nice! I have a soft spot for Duff's Device. Did you happen to compare this to `Array#copyWithin()`? It might be apples to oranges in a sense. It feels like a relevant comparison, though.
1
1
67
Replying to @Vjeux
Does it correlate with context window size for you? I see lots of “forgetfulness” about rules when contexts get long (even before pruning)
1
1
69
Replying to @tmikov
Idk, just hold it off with your finger. It’s not that deep. After spending my childhood on beaches full of cigarette buds and, yes, bottle caps it’s clear that we can’t have nice things one way or the other.
5
120
Replying to @nonken
I find the other way round works better: Codex has a better implementation baseline and incorporates fewer bad ideas. Claude is reviewing more thoroughly.
2
1
40
Replying to @void_0 @cnakazawa
And maybe that’s a takeaway: some things have only been better because they required fewer keystrokes.
1
34
Replying to @cnakazawa
I always liked TS-in-JSDoc. It is more cumbersome to type, but that doesn’t matter anymore.
1
121
out of interest, how much do you stay aware of the code base’s structure and non-user-facing features while not reading the code? It seems useful to be able to make these assessments (“misses the big picture”) and I haven’t figured out a good way yet.
1
373


