Post #2837037

2026-05-25 14:37 UTC

it continues to be amazing to me that this is the “high impact” area they’re going with: even if their analysis systems are better (and frankly I still don’t buy this wholesale, there’s a whole rest of the owl being handwaved[0]), but-elimination is by definition diminishing returns so you can only fanfare like this the first time [0] - having fucking gigantic budgets to throw at running a parse of every single repo and every test condition/simulation you wish to certainly does help a hell of a lot, even moreso when you can shell out to a half-dozen second stage review corps…

Replies (1)

@fiat_lux@lemmy.zip 2026-05-25 23:00

I honestly can't think of anywhere else they can go with it. They need: * something with a binary pass/fail to claim solid numbers at all * something where copy paste is a viable strategy * sufficient public training data from which to derive that copy paste strategy, and, * scary enough consequences to frame any success as impact. Code security review is probably the *only* way you can realistically achieve all four. But they're not even coming close. Not even with access to "partner" black box repositories coupled with under-resourced open source packages. And they know they're not succeeding, because they wouldn't bury that 530 high+ sev number deep in the middle of the press release if they thought it were impressive. Luckily for them, the slop "news" blogs will parrot numbers like 10k, and their only strength - model collapse as a marketing strategy - can handwave the rest of that owl.

Open ##3211628