x.com/RohanArun/status/2108679353504551188
1 correction found
The complex part has been stuck at the same score, 7.009e-4, since PR193. Every PR since then uses that exact number.
This was already false when the post was published: PR #200, opened earlier that day, reported a different complex saving of 0.000665489485337, so not every post-#193 PR used 7.009e-4.
Full reasoning
PR #193 does report the 7.0091844e-4 complex saving the post mentions. But the claim says every PR since then used that exact number, and that is contradicted by PR #200, which was opened on 2026-10-09T14:26:45Z (before this X post's timestamp of 2026-10-09T16:09:56Z).
In PR #200, the author gives a different complex saving for its own complex supplier: 0.000665489485337, not 0.0007009184. So the complex side was not literally "stuck at the same score" across every PR after #193.
This does not mean PR #200 was better than PR #193 on the complex side; it means the post's blanket statement "Every PR since then uses that exact number" is factually wrong, because at least one later PR used a different complex value.
2 sources
- GitHub API: CrocSwap/integer-mult-bounds pull request #193
created_at: 2026-10-09T13:06:47Z ... "The complex saving is b = 219037/312500000 = 0.0007009184."
- GitHub API: CrocSwap/integer-mult-bounds pull request #200
created_at: 2026-10-09T14:26:45Z ... "The refined complex profile ... its saving is 665489485337/10^15 = 0.000665489485337."